Abstracts from the 2023 Annual Meeting of the Society of General Internal Medicine (2024)

Scientific Abstract - Acute Disease Management

ADMISSION VITAL SIGNS AND LABORATORY CHARACTERISTICS AS PREDICTORS OF DISPOSITION STATUS IN A DIVERSE COHORT OF HOSPITALIZED COVID-19 PATIENTS IN NEW MEXICO

Dominic Lundquist³; Alexandra Do⁸; Jordan West³; Molly Biggs⁴; Gaelan Montoya⁹; Clinton Onyango⁶; Samuel Bonuke⁵; Ivy Hurwitz¹; Douglas Perkins²; Jens Langsjoen⁷

¹Internal Medicine, University of New Mexico Health Sciences Center, Albuquerque, NM; ²Global Health, University of New Mexico Health Sciences Center, Albuquerque, NM; ³School of Medicine, University of New Mexico Health Sciences Center, Albuquerque, NM; ⁴School of Medicine, University of New Mexico School of Medicine, Albuquerque, NM; ⁵Center for Global Health, University of New Mexico Health Sciences Center, Albuquerque, NM; ⁶Biomedical Sciences and Technology, Maseno University, Maseno, Nyanza, Kenya;⁷internal medicine, The University of New Mexico, Albuquerque, NM; ⁸Medical School, University of New Mexico School of Medicine, Albuquerque, NM; ⁹School of Medicine , University of New Mexico Health Sciences Center, Albuquerque, NM. (Control ID #3874848)

BACKGROUND: There is limited data on the admission vital signs and clinical laboratory values that are associated with patient discharge status [i.e., home, skilled nursing facility (SNF), or death]. Such information is particularly important in diverse ethnic/racial groups since data for the United States and the State of New Mexico illustrate substantially higher rates of hospitalization in American Indians/Alaska Natives (AI/AN). To foster improved preemptive care for future patients, we determined the impact of admission characteristics on disposition status in three hospitalized groups [AI/AN, Hispanic, and Non- Hispanic Whites (NHW)] at the University of New Mexico Hospital (UNMH).

METHODS: A retrospective observational study was conducted on COVID-19 patients (n=1,309) admitted to UNMH between March 2020 and July 2021. Self-reported race/ethnicity, vital signs, and clinical laboratory measures were captured within 48 hours of admission. Disposition status was determined upon discharge and was categorized as (a) home, defined as discharged home, COVID isolation shelters, correctional facilities without the need for nursing care, and nursing homes; (b) skilled-nursing facility (SNF), defined as discharge to any facility that provided additional care; or (c) death, defined as mortality during hospitalization. Univariate analyses for categorical and continuous variables were determined using Chi-Square analysis and Kruskal-Wallis analysis. Multinomial logistic regression analyses were performed using disposition status as the dependent variable with race/ethnicity (reference=NHW), age, sex at birth, vital signs, and clinical laboratory parameters as the predictor variables. Due to a large number of predictor variables, separate models were generated for vital signs, complete blood count, kidney and liver function, and others (i.e., coagulation, HgbA1C, and procalcitonin) with race/ethnicity, age, sex at birth, and admission to the intensive care unit (ICU) in all models.

RESULTS: Of the 1,309 patients included in the study, 777 (59.4%) patients were discharged home, 304 (23.2%) transitioned to SNF, and 228 (17.4%) died during hospitalization. Factors associated with adverse disposition (i.e., SNF or mortality) included increased age, admission to ICU, and AI/AN ancestry. Higher levels of mean arterial pressure, platelets, hemoglobin, Ca²⁺, Cl^-, CO₂, albumin, and HbA1C were associated with protective effects. Conversely, predictors of adverse disposition status were elevated BUN, K⁺, Na⁺, and alkaline phosphatase.

CONCLUSIONS: Admission characteristics for hospitalized COVID-19 patients, as well as race/ethnicity and age are important predictors of disposition destination. These factors should be taken into consideration by physicians when prognosticating patient care plans. Identification of such variables can help physicians take precautionary measures to try and achieve the best patient outcomes.

AN ASSOCIATION BETWEEN THE INPATIENT USE OF EXTENDED-RELEASE OPIOIDS AND INCREASED LENGTH OF HOSPITALIZATION FOR ACUTE PAIN EPISODES IN SICKLE CELL DISEASE

Madison Canning, John T. Menchaca, Morgan McLemore, Fuad El Rassi

Internal Medicine, Emory University, Atlanta, GA. (Control ID #3873560)

BACKGROUND: The intermittent acute pain syndrome associated with sickle cell disease is a poorly understood phenomenon with a complex pathophysiology mechanism. Given the broad spectrum of opioid analgesic formulations, including immediate-release (IR) and extended-release (ER) opioids, it can be challenging for physicians to decide on the most optimal combination of therapies for inpatient treatment of pain episodes. Notably, the American Society of Hematology 2020 guidelines do not comment on the preferred use of IR vs. ER monotherapy or combination regimens for inpatient treatment of sickle cell pain crisis. Based on a thorough literature review, no large-scale randomized trials have been conducted to evaluate the role of ER opioid formulations in improving pain control and decreasing the length of stay for hospitalized sickle cell patients. Therefore, the benefit of IR monotherapy vs. ER-incorporated regimens is largely unknown.

METHODS: This was a retrospective chart review (n=200) of patients with known sickle cell disease requiring hospitalization for acute pain secondary to vaso-occlusive crisis without an alternative secondary etiology for pain source. The review included patients aged 18-67 years old with 43.5% males and 56.5% females hospitalized from 2/1/2022 to 5/6/2022. The hospital course was evaluated for several factors including daily outpatient regimen of morphine milligram equivalents (MME), highest daily inpatient MME, choice of inpatient opioid agents, and length of stay. Specifically, the use of an ER opioid-incorporated regimen was compared against the length of hospital stay and evaluated using a Wilcoxon Rank Summary analysis model.

RESULTS: Out of the 200 patient cohort, n=128 were started or continued on an ER opioid (including Morphine Sulfate CR, Oxycontin, Xtampza ER, Methadone, etc.) as compared to n=72 on an IR monotherapy opioid regimen for the duration of their hospital course. The highest inpatient daily MME averaged 279 [156.0, 361.2] for the IR monotherapy cohort as compared to 380 [300.0, 470.0] for the patients treated with an ER opioid-incorporated regimen with a p-value of <0.01. The average length of stay was 5.0 days [3.0, 7.0] and 6.0 days [4.0, 9.0] for the IR monotherapy group and for the ER-incorporated regimen group, respectively, with a significant p-value of <0.01.

CONCLUSIONS: This study found that inpatient analgesia regimens which included an ER opioid were associated with a statistically significant higher daily inpatient MME and a longer length of hospitalization as compared to an IR monotherapy opioid regimen. A potentially confounding factor is that patients on home ER formulations were more likely to have failed IR opioid monotherapy due to opioid tolerance and therefore have a predisposed risk of prolonged hospital stay. Even with consideration of this factor, these results show that further research is needed to determine the benefit of ER opioids in regard to shortening the hospital course and minimizing inpatient MME escalation.

ASSESSING THE RISK OF CENTRAL PONTINE MYELINOLYSIS IN PATIENTS WITH DIABETIC KETOACIDOSIS AND HYPEROSMOLAR HYPERGLYCEMIC SYNDROME. Falah I. Abu Hassan, Sadia A. Tanami, Mustafa Al-Bayati, Md Rockyb Hasan, Kelly Mcmaster

Internal Medicine, Texas Tech University Health Sciences Center School of Medicine, Amarillo, TX. (Control ID #3852333)

BACKGROUND: Central pontine myelinolysis (CPM) is a dangerous complication of overcorrection of hyponatremia. Several cases have been described in the literature with CPM complicating hyperosmolar hyperglycemic syndrome (HHS), however, the exact risk is unknown.

METHODS: All adult cases with documented ICD-10 codes diagnostic of diabetic ketoacidosis (DKA), HHS, and sepsis were selected from the National Inpatient Sample (NIS) between the years 2016 and 2020. 1:1 propensity score matching without replacement was performed. The Greedy nearest neighbor algorithm was used to match DKA and HHS cases with sepsis cases based on CPM risk factors of hyponatremia, liver disease, malnutrition, and alcohol abuse. SPSS v25.0 fuzzy matching extension was used to perform the matching and the statistical analysis. Sample size was reduced from 2,651,553 to 685,764 following the matching procedure. Subgroup analysis with multivariate binary logistic regression was performed to assess the odds ratio of developing CPM in HHS, DKA and sepsis patients.

RESULTS: A sample of n = 685,764 weighted patients were included in the analysis. 47,762 patients had either DKA or HHS with sepsis, and thus the sepsis control group will refer here to patients with sepsis bu without DKA or HHS. Mean ages for the DKA, HHS, and sepsis were 46, 60, and 65 years respectively 50.8%, 55.3%, 52.2% of the DKA, HHS, and sepsis groups were males respectively. Hyponatremia was present in 23.7% of the DKA group, 32.2% of the HHS group, and 21.7% of the sepsis group. The controlled variables using the propensity score matching were comparable except for the prevalence of liver disease which was higher in the sepsis control group. Odds Ratio (OR) of developing CPM in HHS was 2.683 ± 1.314 (CI 95%, 1.369 - 5.258, p = 0.004), and 0.840 ± 0.341 (CI 95%, 0.499 - 1.415, p = 0.513) for DKA.

CONCLUSIONS: It seems that HHS is associated with increased odds of developing CPM. It is hypothesized that the extreme fluctuations of osmolarity associated with the treatment of HHS and hyperglycemia are responsible for the subsequent development of CPM. Our study matched controls to adjust for the presence of hyponatremia and the presence of other risk factors that can increase the risk of developing CPM, and despite correction, HHS patients had significantly higher odds to develop CPM. This effect was not observed with DKA patients. Unfortunately, our study was not designed to determine the exact etiology of CPM in the setting of HHS, but it may suggest that rapid correction of the high osmolarity and the associated pseudohyponatremia can be dangerous.

ASSOCIATIONS BETWEEN VOLUME OF INTRAVENOUS FLUID AND MORTALITY IN SEPTIC PATIENTS WITH AND WITHOUT HEART FAILURE

Alexander J. Beagle²; Priya A. Prasad¹; Margaret Fang²; Sandra Martinez-Oreper³; Yumiko Abe-Jones⁴; Kirsten Kangelaris⁴

¹Division of General Internal Medicine, Department of Epidemiology and Biostatistics, UCSF, San Francisco, CA; ²Medicine, University of California, San Francisco, San Francisco, CA; ³Hospital Medicine , University of California San Francisco, San Francisco, CA; ⁴Medicine, Division of Hospital Medicine, University of California at San Francisco, San Francisco, CA. (Control ID #3874537)

BACKGROUND: Surviving Sepsis guidelines recommend 30cc/kg intravenous fluid (IVF) resuscitation, but as a recent randomized controlled trial underscored, optimal volume goals are unclear. Some uncertainty may stem from comorbidities like heart failure (HF), a population that typically receives less fluid and in which volume-mortality associations are not well characterized. We sought to evaluate this association.

METHODS: Retrospective analysis of electronic health record (EHR) data from adult academic medical center patients (2012-2021) who met EHR-based Sepsis-III criteria. We excluded patients who presented hypervolemic (required intravenous diuretics). Logistic regression using restricted cubic splines (RCS) was used to model the fluid-mortality relationship, given the non-linearity of the volume-mortality relationship. Exposure: IVF received over the first six hours after patients met Sequential Organ Failure Assessment (SOFA) criteria for sepsis. Outcome: death in-hospital or discharge to hospice. We stratified this analysis by HF status, and adjusted for age and presenting SOFA score as potential confounders.

RESULTS: There were 9,149 patients; 1,375 had HF identified by ICD codes. HF patients received a median of 15 cc/kg of fluid (IQR 7-27), whereas those without HF received 23 cc/kg (IQR 13-37). Across all patients, the unadjusted relationship between volume and probability of death/hospice was U-shaped (Figure A). It was not U-shaped after adjustment for confounders in patients with HF (Figure B). In those without HF, receiving 0-5 cc/kg was associated with increased odds of death vs receiving 25-30 cc/kg (OR= 1.30, 95% CI 1.09-1.55).

CONCLUSIONS: Receiving 25-30 cc/kg was associated with reduced mortality in patients without HF. Nearly 75% of HF patients received less than this, but there was no association between IVF volume and mortality in this group in an adjusted RCS analysis. Large prospective studies could help identify ideal fluid resuscitation strategies in septic patients with HF.

A SURVEY OF PERCEPTIONS OF INPATIENT PAIN MANAGEMENT AMONG HOSPITALIZED PATIENTS WITH INFLAMMATORY BOWEL DISEASE

Ariel Jordan¹; Mehwish Ahmed²; Queen Saunyama²; Jami Kinnucan³; Megan Riehl²; Jeffrey Berinstein²

¹Internal Medicine, University of Michigan Michigan Medicine, Ann Arbor, MI; ²Gastroenterology, University of Michigan Michigan Medicine, Ann Arbor, MI; ³Gastroenterology, Mayo Clinic in Florida, Jacksonville, FL. (Control ID #3863586)

BACKGROUND: Patients with inflammatory bowel disease (IBD) are at risk for chronic pain, particularly during disease exacerbations, which can negatively impact quality of life. Providers often face challenges controlling pain due to limited safe pain management options, leaving IBD patients feeling their pain is not being addressed or treated. There has been limited data focusing on inpatient perceptions of screening and management of pain during IBD relapses. We aim to understand patients with IBD perception of pain during hospitalization in order to optimize pain management in the acute care setting.

METHODS: This was a prospective survey study in adult patients with IBD hospitalized at a single tertiary center for the indication of IBD-related relapse between December 2021- November 2022. Patients completed a survey focused on screening for pain level, inpatient interventions, and outpatient follow-up plan. Descriptive statistics were reported as means and standard deviations for continuous variables and as frequency and percentages for categorical variables.

RESULTS: Among the 40 respondents, 60.5% were female with a mean age of 45.2y (SD 17.1y). The majority were Caucasian (95.0%), 60.0% had Crohn’s disease, 37.5% had ulcerative colitis, and 2.5% had indeterminate colitis with a mean age of diagnosis of 32.2y (SD 20.0y). 67.5% of patients were on an immunosuppressive agent, 52.5% had extraintestinal manifestations, and 67.6% required steroids during admission.

The average pain score prior to admission ranged from 1-9, 27% rating their pain as 1-3, 27% rating their pain as 4-6, and 24.3% rating their pain as 7-9. Most common home pain medications were acetaminophen (38.7%), gabapentin (12.9%), and nonsteroidal anti-inflammatory drugs (12.9%). During the admission, 43.2% of respondents reported an average pain level of 4-6. The majority (71.1%) noted a discussion about pain during admission. 51.4% of respondents were started on new pain medications during admission, most commonly oxycodone (32.4%), acetaminophen (27%), and tramadol (10.8%). The most recommended non- pharmacologic interventions were exercise (25.6%), breathing techniques (15.4%), and dietary adjustments (15.4%). 55.3% of respondents felt offered inpatient interventions improved their pain during admission 23.1% felt they did not have an adequate pain management plan prior to discharge.

CONCLUSIONS: Our data shows inpatient pain management is adequate, though with significant room for improvement. Hospitalized patients at our center utilize a combination of pharmacologic and non- pharmacologic interventions to address their pain. Unfortunately, many patients received opioids during admission, which can increase risk of disease relapse and mortality. We encourage inpatient providers to maximize use of non-opioid medications and non-pharmacologic interventions.

CHANGES IN HOSPITALIZATION OUTCOMES AMONG PRIVATE EQUITY AND NON- PRIVATE EQUITY HOSPITALS DURING THE COVID-19 PANDEMIC

Zirui Song^1,2; Joseph D. Bruch³; Xiaoran Zhang⁴; Derek A. Haas⁴

¹Health Care Policy, Harvard Medical School, Boston, MA; ²Massachusetts General Hospital Department of Medicine, Boston, MA; ³University of Chicago Division of the Biological Sciences, Chicago, IL; ⁴Avant-garde Health, Boston, MA. (Control ID #3875446)

BACKGROUND: Hospitals are increasingly owned by private equity (PE) firms. To date, however, evidence relating hospital quality of care to PE remains scant. In particular, how patient outcomes differed between PE and non-PE hospitals during the COVID-19 pandemic, a time when hospital resources were stressed, is unknown. Using nationwide Medicare data, we examined the relationship between PE hospital ownership and changes in outcomes among COVID-19 and non-COVID-19 hospitalizations, focusing on mortality, before and during the COVID-19 pandemic.

METHODS: We used 100% traditional Medicare inpatient claims from January 2019 to February 2021. We identified hospitals acquired by PE before the study period (treatment group) using data from S&P Global Market Intelligence software, Irving Levin Associates, Securities and Exchange Commission, official press releases, and hospital websites. The control group comprised non-PE hospitals in the same markets (CBSAs) as PE hospitals.

We identified COVID-19 admissions using the COVID-19 diagnosis, with all others defined as non- COVID-19 hospitalizations. Primary outcomes included: in-hospital mortality, 30-day all-cause mortality, and mortality inclusive of discharges to hospice or of deaths during 30-day readmissions. Secondary outcomes included discharges to hospice and discharges to post-acute care.

Analyses of COVID-19 hospitalizations were cross-sectional comparisons between PE and non-PE hospitals, given the lack of COVID-19 prepandemic. Analyses of non-COVID-19 hospitalizations used an OLS difference-in-differences model (pre and post periods before and after March 2020, respectively), adjusted for age, sex, race, CMS-HCC risk score, share of admissions with ESRD, share with disability, CBSA-level COVID-19 infection rate, month fixed effects (FEs), admitting diagnosis group FEs, and hospital FEs, with standard errors clustered by hospital. We further assessed differential changes in outcomes by race and ethnicity.

RESULTS: We identified 187,382 and 114,166 hospitalizations before and during the pandemic respectively, among 122 PE hospitals, along with 2,287,717 and 1,454,353 hospitalizations before and during the pandemic, respectively, among 1,060 non-PE hospitals.

For COVID-19 hospitalizations, 30-day all-cause mortality was 2.3 percentage points higher among PE hospitals relative to their non-PE counterparts (p=0.02, control group mean 29.0%). Among non-COVID-19 hospitalizations, PE hospitals exhibited a differential increase in 30-day all-cause mortality of 0.90 percentage points relative to non-PE hospitals (p<0.001, control group mean pre-COVID 8.1%).

No differential gaps or changes in mortality by race or ethnicity were found among PE relative to non-PE hospitals.

CONCLUSIONS: All-cause mortality after COVID-19 hospitalizations was higher, and that after non- COVID-19 hospitalizations differentially increased, among PE relative to non-PE hospitals in Medicare. These findings stress the importance of understanding the impact of PE on care delivery.

CHARACTERIZING INTUBATION IN SEVERE COVID-19 INFECTION

Ogheneyoma Akpoviroro¹; Nathan Sauers³; Myriam Castagne²; Wasique Mirza¹; Jameson Woodard¹

¹Internal Medicine, Geisinger Wyoming Valley Medical Center, Wilkes-Barre, PA; ²Clinical and Translational Science Institute, Boston University, Boston, MA; ³Department of Engineering, Pennsylvania State University, State College, PA. (Control ID #3871770)

BACKGROUND: We sought to characterize the group of severe COVID-19(SC19) patients who required intubation and mechanical ventilation(IMV)

METHODS: Retrospective review of admissions with SC19. SC19 was defined as requiring admission to the progressive care unit(PCU). Other inclusion criteria: aged 18-85 years, at least a 10-day course of high- dose dexamethasone and 10 days of remdesivir, admission between 01/01/2021-12/31/2021, intubated during admission.

RESULTS: The study sample included 40 patients. Median age was 60 years; interquartile range(IQR):15.5. Males comprised 57.5%(23/40), 42.5%(17/40) were female. Caucasian patients comprised 95%(38/40), 5%(2/40) were Black. Non-Hispanic patients comprised 82.5%(33/40),10.0%(4/40) were Hispanic; 7.5%(3/40) had unknown ethnicity. While 65% had government-funded insurance, 35% had private insurance. Median BMI was 35.5(IQR:10.6). Median Charlson Comorbidity Index(CCI) was 4(IQR:4). Prevalence of comorbid conditions was as follows: chronic kidney disease:5.0%, diabetes:47.5%, hypertension:75.0%, congestive heart failure:12.5%, chronic lung disease(CLD):50% and of these, 45%(9/20) had COPD and 55%(11/20), obstructive sleep apnea. Patients with an ever-smoker status comprised 65.0%, and of these, 15.4%(4/26) were current smokers and 84.6%(22/26), former smokers. Bacterial pneumonia occurred in 65%(26/40), and AKI in 12.5%(5/40). Azithromycin was administered in 37.5%, hydroxychloroquine in 5%, baricitinib in 22.5%, tocilizumab in 37.5%, anti-diabetic medication in 65%(compared to prior-to-admission use in 15%), and statin medications in 40%. Fully vaccinated patients comprised 5%(2/40). Average admission duration was 22.5 days(standard deviation(SD):10.0 days). Non- invasive ventilation occurred in 90%(36/40). Average admission-to-IMV interval was 108.1 hours(SD:134.9). Average duration of ICU stay was 14.4 days(SD:8.7). Average laboratory values were as follows:albumin(2.9, SD:0.4), ALT(85.7, SD:185.4), AST(70.7, SD:75.9), total bilirubin(0.5, SD:0.2), creatinine(0.8,

SD:0.4), CRP(85.3, SD:58.6), ESR(31.3, SD:32.3), lactate(2.7, SD:3.1), ferritin(1139.0, SD:966.7), D-Dimer(5.1, SD:5.2), troponins(31.1, SD:57.8), WBC count(14.0, SD:4.2), PaCO2(45.4, SD:12.7), pH(7.37, SD:0.06), bicarbonate(25.4, SD:5.0), and respiratory rate(22.8, SD:3.5). In-hospital mortality occurred in 52.5%(21/40). Patients received an average of 14.7 days(SD:4.5) of dexamethasone, and 10.1 days(SD:0) of remdesivir. An average of 0.3 days(SD:0.7) elapsed between starting dexamethasone and remdesivir. CONCLUSIONS: More intubated patients with SC19 infection were male. Racial and ethnicity distribution were like the regional population. Most patients had government-funded insurance, moderate comorbidity- level per the CCI, obesity, hypertension, a smoking history, and half had CLD. Bacterial pneumonia occurred in more patients than not. Only 5% were fully vaccinated. Liver function labs and markers of inflammation were elevated. Over half(52.5%) experienced in-hospital mortality.

CHARACTERIZING THE DISTRIBUTION OF UNDERLYING ETIOLOGIES IN HOSPITALIZED ANEMIA PATIENTS

Christian Carrier¹; David O. Meltzer^1,2; Stephanie Cardenas¹; Micah T. Prochaska^1,2

¹Department of Medicine, University of Chicago Pritzker School of Medicine, Chicago, IL; ²Section of Hospital Medicine, University of Chicago Pritzker School of Medicine, Chicago, IL. (Control ID #3874801)

BACKGROUND: Anemia is an acute and/or chronic condition with a range of underlying etiologies present in 40-70% of hospitalized patients. Diagnostic algorithms for anemia exist, but they are rarely used in the inpatient setting because they were derived in ambulatory patients where acute anemia is less common, and a comprehensive outpatient derived workup may be inappropriate in patients who are acutely sick with other conditions. Therefore, the workup of anemia in the inpatient setting varies by provider. As a result, a hospital-based diagnostic and treatment algorithm would improve identification of the etiology of anemia and allow for intervention if appropriate. However, data on the clinically relevant incidence of anemia etiologies in an inpatient setting is needed to develop such algorithms. Therefore, the purpose of this study was to quantify the distribution of anemia etiologies in hospitalized patients into categories clinically relevant to inpatient providers.

METHODS: Any hospitalized general medicine patient with a hemoglobin (Hb)<10 g/dL during hospitalization was eligible. Anemia categories were derived from laboratory data in consenting patients that was abstracted from the electronic health record (EHR) during the index admission and included: 1) Anemia due to nutritional deficiency defined by a ferritin level <50 ng/mL or Vitamin B12 level <20 pg/mL or B12 <300 pg/mL with a corresponding methylmalonic acid >0.4 mmol/L or a serum folate level <3.4 ng/mL or a red blood cell folate <316 ng/mL. 2) Chronic anemia defined by a glomerular filtration rate <30 mL/min/body surface area or a ferritin >50 ng/mL. 3) Acute blood loss anemia defined by an admission Hb >11.5 g/dL. 4) Sickle cell anemia based on a documented history in the EHR and/or an International Classification of Disease 10 code. 5) Undiagnosed anemia defined by patients without laboratory values available or meeting criteria for any of the above anemia etiologies. Patients meeting criteria for more than one etiology were categorized as 6) multiple etiologies of anemia.

RESULTS: There were 945 participants. The mean age was 58, 57% were female and 72% were African American. The mean admission Hb level was 9.5 g/dL, and the mean nadir Hb was 7.7 g/dL. Chronic anemia was the most common etiology (532 patients, 56%), followed by undiagnosed anemia (250 patients, 26%), acute blood loss anemia (141 patients, 15%), anemia due to nutritional deficiency (83 patients, 9%), and sickle cell anemia (74 patients, 8%). 132 patients (14%) were found to have more than one etiology of anemia.

CONCLUSIONS: Chronic anemia is the most common etiology of anemia, but a substantial portion of hospitalized adults with anemia have undiagnosed anemia. The distribution of the underlying anemia etiologies in hospitalized patients differs from the outpatient setting, and this data can be used to develop an optimal standardized approach to anemia diagnosis and treatment in hospitalized patients.

CLINICAL CHARACTERISTICS, RISK FACTORS, AND OUTCOMES OF PATIENTS WITH COMMUNITY-ASSOCIATED CLOSTRIDIOIDES DIFFICILE INFECTION

Nicole Winchester¹; Curtis J. Donskey³; Amy S. Nowacki²; Abhishek Deshpande⁴

¹Lerner College of Medicine, Cleveland Clinic, Cleveland, OH; ²Quantitative Health Science, Cleveland Clinic, Cleveland, OH; ³Cleveland VAMC, Cleveland, OH; ⁴Center for Value Based Care Research, Cleveland Clinic, Cleveland, OH. (Control ID #3875277)

BACKGROUND:Clostridioides difficile results in half a million new infections and 15-30,000 deaths in the US each year. Half of patients are diagnosed as outpatients and one third have no recent hospital exposures and are defined as community-associated infections. Most studies of community-associated C. difficile infection (CDI) were conducted when standalone nucleic acid amplification tests (NAAT) were the diagnostic test of choice. However, standalone NAAT runs the risk of over diagnosis as the test cannot differentiate between symptomatic infection and asymptomatic carriage, and may therefore underestimate disease severity in this population.

METHODS: We conducted a retrospective cohort analysis of patients (ages >18) with an initial diagnosis of CDI at the Cleveland Clinic between January 1, 2018 and March 1, 2020, and no overnight stay in a healthcare facility within the preceding 90 days. CDI was defined as diarrhea plus a positive toxin B NAAT and toxin enzyme immunoassay (EIA). Patients with white blood cell count greater than 15,000 and creatinine greater than 1.5 mg/dl or 1.5 times their baseline level were classified as severe disease, according to IDSA/SHEA guidelines. Numeric and categorical variables were assessed for association with disease severity utilizing t-tests/Wilcoxon rank-sum tests and chi-square tests/Fisher’s exact tests respectively. P < 0.003 was considered statistically significant to adjust for 17 comparisons.

RESULTS: 340 patients met inclusion criteria for community associated CDI. The median age was 68.79. 63.8% of patients were female, 84.1% identified as white, 9.1% as black, 2.1% as multiracial, 0.6% as Asian, and 0.3% as American Indian or Alaskan Native. The median BMI was 26.58 (IQR- 8). 18.3% of patients had inflammatory bowel disease at baseline, 20.6% had diabetes, and 21.6% had chronic kidney disease. During the 90 days prior to CDI diagnosis, 30% of patients were taking a gastric acid suppressant, only 34.0% received antibiotics prior, and half of patients had no healthcare exposures of any kind within the 90 days prior to CDI diagnosis. 85% of patients were hospitalized, 17.4% had severe disease, and 11.2% developed sepsis. Older age (p=0.002), diabetes(p,0.001), chronic kidney disease (p<0.0001), and no prior antibiotic exposure (p=0.001) were strongly associated with severity.

CONCLUSIONS: In our facility, patients with community-associated CDI had low rates of outpatient healthcare exposure and antibiotic use preceding diagnosis. 17.4% of patients developed severe disease and older age, diabetes, CKD, and lack of prior antibiotic exposure were all strongly associated with severe disease.

CLINICAL CHARACTERISTICS AND OUTCOMES OF VETERANS HOSPITALIZED WITH ALCOHOL WITHDRAWAL

Matthew Ronan^1,3; Kirsha S. Gordon²; Michael F. Krug¹¹; Patrick O. Godwin¹²; Dan Heppe¹³; Matthew Hoegh¹³; Joel C. Boggan⁴; Jeydith Gutierrez⁵; Micah Pescetto²³; Peter J. Kaboli⁵; Michelle Guidry¹⁴; Peter Caldwell¹⁴; Christine A. Mitchell¹⁵; Erik Ehlers¹⁵; Nazima Allaudeen⁶; Jessica E. Cyr⁷; Andrea Smeraglio¹⁶; Yarbrough Peter⁸; Richard Rose^17,8; Anand Jagannath¹⁸; Jaclyn Vargas¹⁸; Paul B. Cornia⁹; Meghna Shah⁹; Matthew Tuck¹⁰; Cherinne Arundel¹⁰; James D. Laudate¹⁹; Joel Elzweig¹⁹; Benjamin A. Rodwin²⁰; Joyce Akwe²¹; Meredith Trubitt²¹; Mary K. Charles²²; Craig Gunderson²⁰

¹Medicine, VA Boston Healthcare System, West Roxbury, MA; ²School of Medicine, Yale Univesity, New Haven, CT; ³Harvard Medical School, Boston, MA; ⁴Medicine, Duke Medicine, Durham, NC; ⁵Internal Medicine, Iowa City VAMC, Iowa City, IA; ⁶Internal Medicine, VA Palo Alto Health Care System, Palo Alto , CA; ⁷Internal Medicine, VA Pittsburgh Healthcare System, Pittsburgh, PA; ⁸Internal Medicine, University of Utah, Salt Lake City, UT; ⁹Medicine, VA Puget Sound HCS & University of Washington SOM, Seattle, WA; ¹⁰Medical Service, Washington DC VA Medical Center, Washington, DC; ¹¹Medicine, Boise VA Medical Center, Boise, ID; ¹²Jesse Brown VA Chicago Healthcare System, Chicago, IL; ¹³VA Eastern Colorado Health Care System, Aurora, CO; ¹⁴Southeast Louisiana Veterans Health Care System, New Orleans, LA; ¹⁵VA Nebraska-Western Iowa Health Care System, Omaha, NE; ¹⁶Portland VA Medical Center, Portland, OR; ¹⁷VA Salt Lake City Health Care System, Salt Lake City, UT; ¹⁸VA San Diego Healthcare System, San Diego, CA; ¹⁹White River Junction VA Medical Center, White River Junction, VT; ²⁰VA Connecticut Healthcare System, West Haven, CT; ²¹VA Medical Center Atlanta, Decatur, GA; ²²Durham VA Medical Center, Durham, NC; ²³Kansas City VA Medical Center, Kansas City, MO. (Control ID #3855393)

BACKGROUND: Alcohol withdrawal (AW) is common in hospitalized patients, yet few studies describe current management in hospitalized settings or compare practices to the American Society of Addiction Medicine (ASAM) guideline. We sought to describe patient characteristics and determine factors associated with complicated withdrawal (CW) and treatment duration in hospitalized patients with AW.

METHODS: A cohort of hospitalized patients with AW from 19 Veterans Affairs (VA) hospitals across the US was used. Demographic and comorbidity data were obtained from the VA’s corporate data warehouse. Inpatient management and hospital outcomes were obtained by chart review. Treatment duration was the total number days of treatment (last date minus first date of receiving AW medication). CW was a composite of delirium and seizure. Using a Poisson model, factors associated with treatment duration were examined.

Logistic regression was used to examine factors associated with CW. Both models were adjusted for demographic, clinical characteristics, treatment type, and medications.

RESULTS: There were 594 patients with AW. Most were male (96%), white (74%) and unemployed (60%). More than half had an AW admission in the prior year, 28% had prior alcohol withdrawal delirium, and 33% had a prior withdrawal seizure. Fifty Six percent were managed with symptom triggered therapy (STT) alone, 28% with fixed dose plus STT, 13% with front load regimens plus STT, and 6% with fixed dose alone. The medication given most was lorazepam (87%) followed by chlordiazepoxide (33%), diazepam (14%), and phenobarbital (6%). Median duration of treatment was 3 days. Thirteen percent of patients developed delirium, 5% seizure, and 1% died. Factors associated with treatment duration included prior delirium, acute medical reason for admission, and prior intensive care unit admissions. STT alone and STT with front loads were associated with reduced treatment duration. Lorazepam, gabapentin and phenobarbital were associated with increased treatment duration. Factors associated with CW included race (Black compared to White, AOR 3.1, 95% CI 1.5-6.4), prior delirium (AOR 2.3, 95% CI 1.2-4.4), prior seizure (AOR 2.1, 1.1-3.9), admission alcohol level (AOR 1.0, 95% CI 1.0-1.0) and thrombocytopenia (AOR 2.2, 1.2-3.8). Neither treatment mode nor choice of medication were associated with CW.

CONCLUSIONS: Our results support recommendations by ASAM to manage patients with long-acting benzodiazepines using STT alone or with front-loaded medications. The most popular medication used was the intermediate-acting lorazepam, which was associated with increased treatment duration. Fixed-dose regimens were commonly used and were associated with increased treatment duration. Symptom triggered regimens were preferable to fixed dose regimens for preventing complicated withdrawal in univariate analysis, but the difference was non-significant in adjusted models.

CLINICAL OUTCOMES ASSOCIATED WITH RACE AND GENDER AMONG COVID-19 PATIENTS HOSPITALIZED IN 2020 IN THE USA

Joseph-Kevin Igwe^1,2; Bilen Kassu²; Kelechi Weze²; Precious Ogbonna²; Nicolas Bakinde²

¹Cardiovascular Research, Stanford University School of Medicine, Stanford, CA; ²Morehouse School of Medicine, Atlanta, GA. (Control ID #3875514)

BACKGROUND: COVID-19 disease (COVID-19) has been associated with significantly increased mortality and morbidity. However, the association between race and gender in COVID-19-related mortality has not been evaluated using a national inpatient sample. To determine these associations, we analyzed mortality and related outcomes associated with race and gender for COVID-19 patients hospitalized in the U.S. during the COVID-19 Pandemic using the 2020 National Inpatient Sample (NIS).

METHODS: We identified adults (≥18 years-old) from the 2020 (NIS) database with and without ICD-10- CM codes indicating COVID-19 disease. Pearson χ2 analysis, Analysis of variance (ANOVA), multivariable regression, and propensity-score matched analyses (PSM) of independent variables were conducted to analyze significant associations between variables. Propensity-score matching used age, primary insurance type, length of stay, hospital characteristics, comorbid conditions, medications, and population median- Elixhauser-sum to determine primary outcome: death; and secondary outcomes: acute pulmonary embolism (PE), cardiogenic shock, mechanical ventilation (MV), and ECMO (Extracorporeal Membrane Oxygenation) utilization. STATA/MP software (StataCorp. 2021. Stata Statistical Software: Release 17. College Station, TX: StataCorp LLC)) was used for all analyses.

RESULTS: We identified 1,698,560 patients with COVID-19 disease (6.31% of all admissions) with a mean age 63.19±0.09 years and predominance of males 885,110 (52.11%). Mortality in COVID-19 females (11.63%) was significantly lower than COVID-19 males (15.10%, p≤0.01).

The PSM of 339,719 COVID-19 observations demonstrated that female gender was associated with significantly lower odds of mortality (aOR: 0.83, [0.81-.85], p≤0.01), acute PE (aOR:.70, [0.65-.76], p≤0.01), MV (aOR: 0.82, [0.80-0.84], p≤0.01), and cardiogenic shock (aOR:0.76, [.69-0.84], p≤0.01); but no significant difference in ECMO utilization compared to male patients. However, within the matched female subpopulation, African American and Asian race was associated with higher odds of cardiogenic shock (1.41, [1.18-1.69] p≤0.01; 1.72, [1.28-2.31], p≤0.01, respectively) and ECMO utilization (aOR:2.10, [1.30-3.38], p≤0.01); (aOR:2.94, [1.48-5.85], p≤0.01, respectively) compared to white race COVID-19 patients.

The odds of death for females were significantly different between younger age groups compared to that of older age patients (Subpopulation Age 18-49: aOR: 0.72, [0.65-0.80], p≤0.01; Subpopulation Age≥50: aOR: 0.84, [0.82-0.87], p≤0.01).

CONCLUSIONS: In our study, COVID-19 outcomes among female patients were significantly better than male patients; and race or a race-related confounder plays a role in disease severity in COVID-19. The negative trend in odds of death associated with age was concordant with menopausal age among the female population. The anti-inflammatory effects of estrogen play a modulating role in COVID-19 disease. Additional studies are needed to expand on these findings.

CORONARY ARTERY DISEASE TESTING AMONG PATIENTS HOSPITALIZED WITH NEW- ONSET HEART FAILURE

Siamak Kohan³; Janet Lee¹; Huong Q. Nguyen¹; In-Lu Amy Liu¹; Mingsum Lee²; Cheng-wei Huang⁴. ¹Research and Evaluation, Kaiser Permanente Southern California, Pasadena, CA; ²Cardiology, Kaiser Permanente Los Angeles Medical Center, Los Angeles, CA; ³Internal Medicine, Kaiser Permanente Los Angeles Medical Center, Los Angeles, CA; ⁴Hospital Medicine, Kaiser Permanente Los Angeles Medical Center, Los Angeles, CA. (Control ID #3875255)

BACKGROUND: Coronary artery disease (CAD) testing may be underutilized in patients with heart failure with reduced ejection fraction (HFrEF). We aimed to identify factors associated with CAD testing among patients with new-onset HFrEF.

METHODS: We retrospectively studied adults (≥18 years) hospitalized with new-onset HFrEF between 2016-2021 within Kaiser Permanente. Heart transplant, hospice, pregnant, expired, and non-member patients were excluded. We evaluated the association between sociodemographic and clinical factors with CAD testing within 90 days of index hospitalization with multivariable logistic regression. Odds ratio (OR [95% confidence interval]) were estimated.

RESULTS: Among 2,908 patients, the median age was 70 (interquartile range: 58-80) and 61% were male. White, Hispanic, Black and Asian patients comprised 42%, 28%, 19%, and 10% of patients, respectively. CAD testing was performed in 54% of patients. Patients with ages 60-69 (1.35 [1.03-1.77]), reference: ages 50-59), ages 70-79 (1.43 [1.02-2.01], reference: ages 50-59), and valvular disease (1.26 [1.05-1.51], reference: absence) had higher odds of CAD testing. Patients with ages ≥80 (0.69 [0.48-0.98], reference: ages 50-59), atrial fibrillation (0.70 [0.58-0.84], reference: absence), and do-not-resuscitate order (0.41 [0.32-0.54], reference: full code) had lower odds of CAD testing.

CONCLUSIONS: Differences in CAD testing among patients with new-onset HFrEF may be explained by clinical judgment factoring in the perceived etiology of the HFrEF and goals of care. A universal CAD testing approach may not be appropriate.

DERIVATION AND VALIDATION OF A MODEL TO PREDICT BLEEDING IN MEDICAL INPATIENTS

Benjamin G. Mittman^2,3; Megan Sheehan^1,2; Lisa Kojima^1,2; Michael B. Rothberg²

¹Cleveland Clinic Lerner College of Medicine of Case Western Reserve University, Cleveland, OH; ²Center for Value-Based Care Research, Cleveland Clinic, Cleveland, OH; ³Department of Population and Quantitative Health Sciences, Case Western Reserve University School of Medicine, Cleveland, OH. (Control ID #3874718).

BACKGROUND: Venous thromboembolism (VTE) is the leading cause of preventable hospital death in the US. The 2018 American Society for Hematology guidelines recommend providing pharmacological VTE prophylaxis to acutely or critically ill medical patients at acceptable bleeding risk, but there are few validated risk assessment models (RAMs) for estimating bleeding risk in such patients. We developed a RAM and compared it with the IMPROVE model.

METHODS: We identified all medical patients admitted to one health system from 2017-2020. Bleeding events were identified by a combination of diagnostic and administrative codes and validated through chart review. Predictors included risk factors identified from previous studies and medications with known bleeding risk. Data were split into training (70%) and validation (30%) sets with pseudo-randomization of bleeding events to ensure proportional event rates in each set. Multiple logistic regression was used to generate and validate the model with all predictors retained. We also validated the IMPROVE model and compared its performance to our RAM in the validation set on the basis of calibration and discrimination. Finally, we compared models on bleed detection rates and the number of patients categorized as low vs. high risk using chi-squared tests.

RESULTS: Our training set included 22,351 patients, 86 (0.39%) of whom developed a bleed while in the hospital. Our final RAM included 15 variables: age, sex, race, platelet count, INR, PTT, GFR, bleeding event <3 months prior to admission, rheumatic disease, sepsis, ICU or CCU admission, central line, peptic ulcer, current cancer, and in-hospital receipt of antiplatelet medications (aspirin, clopidogrel, ticagrelor, or prasrugrel). The strongest predictors were peptic ulcer (OR = 7.2, 95% CI [3.7, 13.0]), central line (5.6, [3.4, 9.2]), and sepsis (4.3, [2.7, 7.0]). The validation set included 9,580 patients; 37 (0.39%) had an in-hospital bleed and individual predicted risks ranged from 0.006% to 28% (median = 0.12%), with good calibration. The C-statistic for our model was 0.77 (95% CI, 0.68-0.86) compared with 0.70 (95% CI, 0.61-0.79) for the IMPROVE model (p = .31). At a treatment threshold of 1.1% (equivalent to an IMPROVE score of 7), our model detected a similar percentage of bleeds as IMPROVE (32% vs 35%, p = .81), but categorized significantly fewer patients as high-risk (7.2% vs 9.0%; p < .001). Additionally, 69% of patients categorized as high-risk in our model nevertheless received in-hospital VTE prophylaxis.

CONCLUSIONS: Most hospitalized patients are at low risk of bleeding, although some are at very high risk. Our RAM had good calibration and similar discrimination to the IMPROVE model. Our RAM detected a similar number of bleeds as the IMPROVE model but identified significantly fewer patients as high-risk. Use of this RAM could help physicians accurately balance risks of bleeding against VTE when considering prophylaxis.

DEVELOPING A TOOL TO ASSESS THE RISK FOR THROMBOEMBOLISM IN PATIENTS HOSPITALIZED WITH SARS-COV-2

Zeina Kayali^1,2; Sarah Paglialonga^1,2; Wilhelmine Wiese-Rometsch³; Ke Ning⁴

¹Internal Medicine Residency Program, Florida State University College of Medicine, Tallahassee, FL; ²Sarasota Memorial Health Care System, Sarasota, FL; ³Graduate Medical Education, Florida State University, Sarasota, FL; ⁴Florida Cancer Specialists & Research Institute LLC, Sarasota, FL. (Control ID #3874205)

BACKGROUND: SARS-CoV-2 infection is thrombogenic. A standardized risk prediction tool to identify thromboembolic risk on admission has not been developed.

METHODS: Patient demographics and clinical and administrative data were extracted from electronic medical records under IRB exemption. Generalized regression with adaptive LASSO modeling and response surface was used to identify (p < 0.05) traits putatively associated with embolism, retaining those that optimized generalized RSquared for aggregated features. Univariate logistic regression for each retained trait created a receiver operating characteristic (ROC) curve with Youden Index estimating cut-point optimizing sensitivity and specificity associated with embolism. An analysis was performed using Boosted Tree (BT) computed at each vascular site thromboembolism percentage for the patient traits while differentiating between present on admission vs. delayed. The area under the ROC was optimized to represent the predictive synergy of the model.

RESULTS: Among 5724 patients hospitalized with SARS-CoV-2 and consecutively discharged from index hospitalization between March 14, 2020, and June 30, 2022, 424 (7.4%) experienced an embolism, including arterial embolism (AE, 3.4%), deep venous thrombosis (DVT, 1.8%), and pulmonary embolism (PE, 2.2%). The BT model identified 13 patient traits and their cutoff points and demonstrated that thromboembolic event signatures differed across embolism sites.

CONCLUSIONS: This single-institution study demonstrates a descriptive tool model to stratify the risk of developing thromboembolic events in patients hospitalized with SARS-CoV-2. We identified 13 different patient traits on hospital admission while establishing the proportion each trait contributes to thromboembolic sites. Future research with a larger sample size is needed to create a predictive tool to guide patient-centric treatment in this population.

DEVELOPMENT AND VALIDATION OF A MODEL TO PREDICT TRANSFUSION REQUIREMENT DURING ELECTIVE SURGERY

Matthew A. Pappas^1,2; Lisa Kojima⁵; Moises Auron³; Deborah Tolich⁷; Daniel I. Sessler⁶; Michael B. Rothberg⁴

¹Center for Value-based Care Research, Cleveland Clinic, Cleveland, OH; ²Department of Hospital Medicine, Cleveland Clinic, Cleveland, OH; ³Hospital Medicine, Cleveland Clinic, Cleveland, OH; ⁴Internal Medicine, Cleveland Clinic, Cleveland, OH; ⁵Cleveland Clinic Lerner College of Medicine of Case Western Reserve University, Cleveland, OH; ⁶Department of Outcomes Research, Cleveland Clinic, Cleveland, OH; ⁷Patient Blood Management, Cleveland Clinic, Cleveland, OH. (Control ID #3869236)

BACKGROUND: Patients completing surgery are often transfused at thresholds higher than supported by available evidence. Preoperative hemoglobin is related to transfusion requirements, but the system used to predict transfusion needs during elective surgery, the Maximum Surgical Blood Order Schedule (MSBOS), does not incorporate preoperative hemoglobin.

METHODS: We assembled a cohort of patients who visited a preoperative risk assessment clinic from 2008 through 2018. We used preoperative hemoglobin and scheduled Current Procedure Terminology (CPT) code to predict the lowest hemoglobin (dichotomized at 7 mg/dL) in the three days after surgery. When patients received intraoperative transfusion, we adjusted the nadir hemoglobin for transfusion volume received using the Lopez-Picado formula. We imputed preoperative hemoglobin when not measured on the morning of surgery.

We split the dataset into derivation (80%) and validation (20%) cohorts, stratified by CPT code. We then fit a multilevel logistic regression in the development set, clustered by CPT code and with preoperative hemoglobin as a restricted cubic spline. Then, in the validation set, we assessed discrimination using the area under the receiver-operator curve (AUROC) and calibration using the Brier score.

RESULTS: We identified 109,794 visits that were followed by major noncardiac surgery within 30 days; 104,267 had one or more preoperative hemoglobin measures. After imputation, our analytic sample included 107,003 surgeries of 2,363 types (87,998 for development, 19,005 for validation). The adjusted nadir hemoglobin was below 7 mg/dL in 5,967 (5.4%). The model's AUROC was 0.92 (95% CI: 0.91-0.93) and its Brier score was 0.04. A calibration plot is shown in the Figure.

CONCLUSIONS: A model using planned CPT and preoperative hemoglobin predicted whether a patient's nadir hemoglobin would fall below 7 mg/dL with excellent discrimination and calibration. This model could allow more efficient preoperative crossmatching compared to MSBOS guidance based on CPT code alone.

DIRECT ORAL ANTICOAGULANT FOR MANAGEMENT OF INTRA-CARDIAC THROMBI: A RETROSPECTIVE COHORT STUDY

Agara Srinivasa Gagan Kumar¹; Omar Hegazy¹; Niranjana Chellappa²; Ola Al-Sous¹

¹Internal Medicine, St Joseph Mercy Health System, Ann Arbor, MI; ²Pulmonary and Critical Care Medicine, Albert Einstein Healthcare Network, Philadelphia, PA. (Control ID #3875310)

BACKGROUND: Intracardiac thrombus (ICT) is a major risk for systemic embolization. Risk has been reported to be as high as 4-5% with left atrial thrombus and 10-15% with left ventricular thrombus in patients who are not anticoagulated. While strong evidence supports the use of direct oral anticoagulants (DOAC) over vitamin K antagonists (VKA) for thrombosis and embolization prevention, data for the management of established ICT has not been as concise. Current data have not only shown that DOAC have at least similar efficacy to VKA for the management of venous and arterial thromboembolic events but also have similar or fewer bleeding events. Given DOAC’s convenience, this data was extrapolated to manage ICT. VKA, however, remains the mainstay of treatment. This study aims to compare the effectiveness of DOAC vs. VKA for ICT management.

METHODS: This was a retrospective descriptive study. The primary outcome was a complete resolution of ICT after six months of treatment. Secondary outcome was serious complications, including major bleeding and cerebrovascular accidents (CVA). Data between January 1, 2015, and December 31, 2019, was obtained electronically from EMR, followed by manual extraction. Inclusion criteria were new diagnosis of ICT confirmed by echocardiogram. Exclusion criteria were loss to follow-up and no available follow-up echocardiogram. Unadjusted t-tests and Chi-square tests were used to assess for differences in baseline characteristics. Regression analysis controlled for baseline characteristics followed, assessing differences in ICT resolution in the first follow-up echo.

RESULTS: Out of 287 patients diagnosed with ICT, 38 patients were included. The VKA and DOAC arms included 22 and 16 patients, respectively. For the primary outcome, the unadjusted comparison revealed greater ICT resolution in the DOAC group vs. the VKA group (81.25% versus 40.91%, p-value = 0.027).

This effect was maintained after adjusting for co-variates, and regression analysis revealed lower odds of ICT resolution with VKA (OR = 0.13, p-value = 0.044). The composite secondary endpoint occurred in 13.16% and 7.89% in the VKA vs the DOAC arms, respectively (p-value =0.249).

CONCLUSIONS: Our study shows 87% decreased odds of ICT resolution with VKA compared to DOAC. Moreover, we did not observe a difference between the two treatment modalities in major bleeding and CVA. Our study’s main limitation is being underpowered due to the small sample size. Nevertheless, our results are consistent with recent evidence supporting DOAC as an alternative to VKA for the treatment of ICT. Large clinical trials are, hence, warranted to confirm these findings. Meanwhile, this study adds to the growing body of evidence, and we hope it contributes to future meta-analyses.

EFFECT OF EMPIRIC ANTIBIOTICS FOR METHICILLIN-RESISTANT STAPHYLOCOCCUS AUREUS ON MORTALITY FOR PATIENTS HOSPITALIZED WITH BACTERIAL INFECTIONS George B. Carey^1,2; Jurgen Holleck^2,1; Samer Ein Alshaeba²; Ritujith Jayakrishnan¹; Kirsha Gordon^3,2; Alyssa Grimshaw⁴; Craig Gunderson^5,2

¹Medicine, Yale School of Medicine, New Haven, CT; ²Medicine, VACT Healthcare, Yale School of Medicine, WestHaven, CT; ³Internal Medicine, Yale University School of Medicine, New Haven, CT; ⁴Cushing/Whitney Medical Library, Yale University, New Haven, CT; ⁵Internal Medicine, Yale School of Medicine, New Haven, CT. (Control ID #3873740)

BACKGROUND: Many hospitalized patients are treated empirically for methicillin-resistant Staphylococcus aureus (MRSA), though most do not have MRSA. The benefit of this practice is unclear. We performed a systematic review and meta-analysis to estimate the effect on mortality of initial empiric MRSA coverage, both in patients admitted with MRSA infection and in patients admitted with common infectious syndromes, for whom the causative pathogen may not be MRSA.

METHODS: A systematic literature search was conducted using 7 databases from database inception through April 2022. Studies were included if they reported mortality with active and inactive empiric antibiotics in hospitalized patients with culture-proven MRSA infection. The primary outcome was adjusted odds ratio of mortality with active empiric antibiotics. We used adjusted odds ratios from our meta-analysis with MRSA prevalence and base mortality rates from the literature to estimate the absolute mortality reduction with initial empiric MRSA coverage for common infectious syndromes.

RESULTS: Of 2136 titles, 37 studies (11661 participants) were included. Fifteen studies (6066 participants) reported adjusted odds ratios of mortality, mostly in bacteremia (11 studies, 4266 participants). The pooled adjusted odds ratio was 0.64 (95% confidence interval, 0.48-0.84), favoring active empiric antibiotics in patients with MRSA infection. We found no evidence of benefit in MRSA pneumonia (2 studies, 295 participants; adjusted odds ratio 1.56; 95% confidence interval, 0.74-3.27). We estimate an absolute mortality reduction of 0.1% or less with initial empiric treatment for MRSA for non-critically ill patients admitted with the 3 most common infectious syndromes (community-acquired pneumonia, skin and soft tissue infection, urinary tract infection). Higher absolute risk reductions were estimated in other conditions, such as septic shock (0.6; 95% confidence interval, 0.2-1.0).

CONCLUSIONS: We estimate that for most hospitalized patients, the absolute mortality benefit of initial empiric antibiotics for MRSA is likely 0.1% or less. Meaningful benefit may be seen in conditions with higher prevalence of MRSA and baseline mortality, such as septic shock.

EFFECT OF WEEKEND VS. WEEKDAY ADMISSION ON ENDOCRINE EMERGENCIES RESULTS FROM NATIONAL INPATIENT SAMPLE DATABASE 2016-2019

Bidhya Poudel¹; Biraj Shrestha²

¹Internal medicine, AMITA Health Saint Francis Hospital Evanston, Evanston, IL; ²internal medicine, Tower Health, Reading, PA. (Control ID #3875473)

BACKGROUND: Limited studies exist to understand the effect of weekend vs. weekday admission on Endocrine emergencies.

METHODS: We queried the National Inpatient Sample database from 2016-2019 for hospitalizations with a primary diagnosis of Endocrine emergencies (Diabetic Ketoacidosis (DKA), Hyperosmolar Hyperglycemic State (HHS), Thyrotoxicosis, Myxedema coma and Addison's Crisis). We used NIS-predefined admission days to find whether the admission was on weekends or weekdays. We assessed the two groups' demographics, in-hospital mortality, length of stay, and cost of hospital admission. In-hospital mortality was adjusted for potential confounders using multivariable logistic regression analysis adjusting for baseline characteristics significantly different in univariate analysis (liberal threshold of p<0.20).

RESULTS: From 2016 to 2019, we found 934,685 patients admitted for endocrine emergencies, of which 85.03% were for DKA, 9.63% were for HHS, 3.80% were for thyrotoxicosis, 1.12% were for Adrenal Crisis, and 0.41% were for Myxedema coma. Patients in the Weekday admission were older (43.76 vs. 43.25 years, p<0.001) with higher comorbidities like heart failure, obesity, valvular heart disease, and a higher Elixhauser comorbidity index (-3.09 vs. -3.19, p<0.001). In-hospital mortality was not statistically different in univariate analysis between weekend admission vs. weekday admission (0.55% vs. 0.49%, OR: 1.11, 95% CI: 0.97-1.28, p=0.136); however, when adjusted for the difference in comorbidities, we found slightly higher mortality in weekend admission compared to weekday admission (adjusted OR: 1.16, 95% CI: 1.01-1.34, p=0.039). We didn't find any cost difference between weekend and weekday admission ($8,043.50 vs. $8080.67 USD, p=0.437). We also didn't find any differences in length of stay between weekend and weekday admissions (3.38 days vs. 3.42 days, p=0.052).

On Subgroup analysis of endocrine emergencies, we found a higher unadjusted (unadjusted OR: 1.82, 95% CI: 1.05-3.14, p=0.03) and adjusted (adjusted OR: 1.80, 95% CI: 1.02-3.16, p=0.042) in-hospital mortality in weekend admission vs. weekday admission for Myxedema Coma. We also found a higher adjusted in- hospital mortality (adjusted OR: 1.88, 95% CI: 1.02-3.50, p=0.044) in weekend admission vs. weekday admission in patients admitted for thyrotoxicosis. We didn't find any difference in adjusted or unadjusted in- hospital mortality in weekend vs. weekday admission in patients admitted for DKA, HHS, and adrenal Crisis. For hospital cost and length of stay, we didn't find any differences in subgroup analysis for the various endocrine emergencies.

CONCLUSIONS: Weekend admission for endocrine emergencies was associated with a higher in-hospital mortality than weekday admission, primarily driven by myxedema coma and thyrotoxicosis. Further study is needed to understand the hospital, administrative and physician factors contributing to poorer outcomes in endocrine emergencies, especially in myxedema coma and thyrotoxicosis.

EXPLORING THE CHARACTERISTICS OF THE HOSPITALIZED PATIENTS ADMITTED FOR ACUTE PANCREATITIS WHO DEVELOP ABDOMINAL COMPARTMENT SYNDROME IN THE US HOSPITALS

Falah I. Abu Hassan, Andrea J. Zapata, Md Rockyb Hasan, Sadia A. Tanami, James ". Walker

Internal Medicine, Texas Tech University Health Sciences Center School of Medicine, Amarillo, TX. (Control ID #3872719)

BACKGROUND: Abdominal compartment syndrome (ACS) is a documented complication of acute pancreatitis and is thought to be secondary to aggressive fluid resuscitation. Recent studies advocated for moderate over large volume resuscitation in acute pancreatitis due to the high risk of fluid overload and lack of added benefit.

METHODS: All the cases admitted with a diagnosis of acute pancreatitis between the years 2016 - 2020 were selected by using the relevant ICD-10 codes from the national inpatient sample (NIS). ICD-10 code “M79A3” was used to determine the patients who developed ACS. Multivariate binary logistic regression was used to assess the odds of developing ACS in the presence of biliary pancreatitis, alcoholic pancreatitis, pancreatic necrosis, active alcohol withdrawal, malnutrition, liver disease (hepatitis, cirrhosis, or liver failure), Malignancy, and sepsis. Results were reported as the odds ratio (OR) with confidence intervals (CI) of 95%. Significant results had a p value < 0.05.

RESULTS: 371,729 patients with pancreatitis were found. 0.12% (n = 455) of the patients admitted with pancreatitis developed ACS. 52.7% of all pancreatitis admissions were males. 75.6% of patients with ACS were males. The average age of patients who developed ACS was 48 years as compared to 53 years in those who did not. Significantly higher odds to develop ACS were seen with the presence of sepsis OR = 11.32 (CI

9.16 – 13.99), pancreatic necrosis OR = 10.66 (CI 8.73 – 13.00), active alcohol withdrawal OR = 1.78 (CI 1.377 – 2.32), liver disease OR = 1.72 (CI 1.441 – 2.09), malnutrition OR = 1.57 (CI 1.27 – 2.93), and alcoholic pancreatitis OR = 1.48 (CI 1.1 – 1.88). Biliary pancreatitis, however, is associated with significantly lower odds of OR = 0.62 (CI 0.46 – 0.85). The presence of a malignant comorbidity was not significantly associated with the odds of developing ACS.

CONCLUSIONS: To our knowledge, specific risk factors associated with developing ACS in pancreatitis patients have not been previously reported. In pancreatitis, development of diffuse abdominal edema and fluid sequestration frequently occur, eventually leading to the evolution of ACS (>20mmHg) in rare cases. This leads to a cascade of decreased visceral perfusion and multiple organ dysfunction. Our results suggest that necrosis and sepsis play a major role in the progression of ACS and we hypothesize that the increased permeability and fluid sequestration associated with necrosis, and the larger volume resuscitation in septic patients are to blame. Interestingly, alcoholic pancreatitis and alcohol withdrawals during hospitalization are correlated with higher odds of ACS. We also hypothesize that this is related to the direct toxic damage induced by alcohol on the pancreatic tissue. Clinicians should be vigilant when treating such patients as further studies are needed to unveil unexplored risk factors associated with ACS in pancreatitis, such as necrosis.

IMPACT OF CORTICOSTEROIDS, ANTICOAGULATION, BOTH OR NONE ON PATIENT IN- HOSPITAL OUTCOMES FROM COVID-19 INFECTION

Alison Leslie¹; Katelyn Tessier⁴; Christopher Tignanelli⁵; Scott A. Chapman³; Sameh Hozayen²

¹University of Minnesota Medical School Twin Cities, Minneapolis, MN; ²Department of Medicine, Division of Hospital Medicine, M Health Fairview University of Minnesota Medical Center East Bank, Minneapolis, MN; ³Department of Experimental and Clinical Pharmacology, University of Minnesota College of Pharmacy, Minneapolis, MN; ⁴Biostatistics Core, University of Minnesota Masonic Cancer Center, Minneapolis, MN; ⁵Department of Surgery, Division of Critical Care/Acute Care Surgery, M Health Fairview University of Minnesota Medical Center East Bank, Minneapolis, MN. (Control ID #3876157)

BACKGROUND: There has been increased research evaluating corticosteroids (CCS) and anticoagulants (AC) on clinical outcomes in patients hospitalized with Covid-19. CCS may reduce mortality and ventilation needs in patients hospitalized with Covid-19, but data are sparse outside clinical trials. Similarly, Covid- induced hypercoagulability and the need for AC in SARS-CoV-2 patients may need more practice-driven data in an area of possibly conflicting clinical trial results. This study includes patients hospitalized with Covid-19 who received AC and/or CCS and to evaluate the impact of these pharmacologic interventions on patients’ real life clinical outcomes.

METHODS: This retrospective cohort study included adult patients hospitalized with Covid-19 from March 1st, 2020 to April 30th, 2022 at Fairview, a United States Midwest health system. Included patients were grouped into four cohorts: received AC only, CCS only, both AC + CCS, or neither. In-hospital outcomes included mortality and thromboembolic events. Baseline demographics and in-hospital outcomes were summarized and compared by medication group. Unadjusted and adjusted logistic regression models were performed.

RESULTS: 5420 patients were included: AC only (N=2332), CCS only (N=129), both AC + CCS (N=1857), or neither (N=1102). There were statistically significant differences in baseline characteristics by medication group, some likely driven by sample size. Results are in the table below. For the unadjusted model, AC only was associated with significant reduction in mortality while CCS only and AC + CCS were associated with a significant increase in mortality. AC only remained significant in its mortality reduction when adjusting for comorbidities, while the increased mortality with CCS only was no longer seen. CCS only was associated with increased thromboembolism risk in both adjusted and unadjusted models.

CONCLUSIONS: AC only decreased the risk of mortality while CCS only increased the risk of thromboembolism in hospitalized patients with Covid-19. Increased risk with CCS only group may reflect a higher risk group at baseline that had more comorbidities and higher acuity presentation.

IMPLICATIONS OF USING LOCALLY VALIDATED RISK FACTORS FOR RESISTANCE WHEN CHOOSING ANTIBIOTICS FOR COMMUNITY-ACQUIRED PNEUMONIA

Benjamin G. Mittman^1,2; Hamlet Gasoyan¹; Abhishek Deshpande¹; Peter B. Imrey³; Ning Guo³; Michael B. Rothberg¹

¹Center for Value-Based Care Research, Cleveland Clinic, Cleveland, OH; ²Department of Population and Quantitative Health Sciences, Case Western Reserve University School of Medicine, Cleveland, OH; ³Department of Quantitative Health Sciences, Cleveland Clinic, Cleveland, OH. (Control ID #3874709)

BACKGROUND: The 2019 ATS/IDSA community-acquired pneumonia (CAP) guidelines recommend prescribing empiric extended-spectrum antibiotics (ESA) for MRSA or P aeruginosa only if locally validated risk factors are present, but the implications of this recommendation are unknown. We determined the percentage of US hospitals that had enough resistant infections to identify local risk factors and, among hospitals that did, compared theoretical use of ESA based on local risk factors or generic risk factors against observed use of ESA.

METHODS: We identified 177 hospitals in the Premier Healthcare Database and quantified resistant infections among adult patients hospitalized for CAP from 2010-2015. We assumed that hospitals with ≥80 patients with resistant infections could identify local risk factors. Among these hospitals, we compared the performance of local risk factors with 2 generic risk factors recommended by the guidelines (recent intravenous antibiotics or previous infection with or MRSA or P aeruginosa). Local risk factors were those associated with resistant infections at p < 0.05 in standard two-group comparisons. We then compared the percentage of all patients and those with resistant infections who would receive ESA with each approach. Finally, we determined actual rates of ESA use in each hospital.

RESULTS: Over 5 years, 11/177 (6%) hospitals had enough resistant infections to derive their own risk factors. Physicians covered 60-80% of patients with resistant infections by prescribing ESA to 37-58% of all patients (Figure). Using local risk factors, these hospitals would improve coverage of resistant infections to 81%-98.8%, but this would require using ESA for 54-93% of all patients with CAP. Using the 2 generic risk factors would reduce the use of ESA to <13% of all patients, but would cover only 11-28% of resistant infections.

CONCLUSIONS: Most US hospitals did not have enough patients with resistant infections to derive local risk factors. For those that did, choosing ESA based on local risk factors would increase rates of ESA, whereas basing it on the two generic risk factors would decrease use of ESA. Guidance is needed regarding the appropriate tradeoff between treating patients with resistant infections and those without.

IN-HOSPITAL MORTALITY IN PATIENTS WITH PULMONARY EMBOLISM MANAGED WITH ULTRASOUND-ASSISTED CATHETER-DIRECTED THROMBOLYSIS COMPARED TO SYSTEMIC THROMBOLYSIS

Kristina D. Menchaca¹; Catherine A. Ostos¹; Nemanja Draguljevic²; Virginia Velez Quinones¹; Erika Ostos¹; Robert Chait¹

¹Internal Medicine, University of Miami School of Medicine, Miami, FL; ²Internal Medicine, Univerzitet u Beogradu, Beograd, Beograd, Serbia. (Control ID #3875383)

BACKGROUND: Patients with high-risk or intermediate-risk pulmonary embolism (PE) require more aggressive therapy, such as systemic thrombolysis. Although parenteral systemic thrombolysis leads to faster clot resolution, it is associated with a high risk for significant bleeding and its complications. The use of catheter-directed thrombolysis (CDT) decreases systemic exposure to thrombolytics and reduces the risk of complications and mortality. Adding ultrasound to CDT can potentially decrease the dosing of local thrombolytics and make it even safer for the patients.

METHODS: This retrospective cohort study identified 1,007 patients diagnosed with pulmonary embolism (PE) across several facilities. Patients were stratified into two groups based on the treatment they received, a group that received systemic thrombolysis (ST), 432 (42.9 %), and a group that received ultrasound-assisted catheter-directed thrombolysis (EKOS), 575 (57.1 %). The aim was to explore immediate mortality from direct and indirect complications arising from PE and its management in patients treated with systemic thrombolysis vs. EKOS. We used logistic regression to analyze data.

RESULTS: The patients with PE and who were managed with ST were more likely to expire than those treated with EKOS (OR 2.893, CI 1.766~5.041, p<0.001). Patients who passed were more likely to experience tachycardia (OR 2.432, CI 1,299~4.556, p<0.001) and use of vasopressors (OR 13.331, CI 8.075~22.006, p<0.001). There were 70 (6.95 %) deaths in the ST group vs. 26 (2.58 %) in the EKOS group.

CONCLUSIONS: Patients with PE managed with ST are more likely to die than those with EKOS. This illustrates better outcomes of using EKOS vs. systemic thrombolysis for PE. However, a limitation of this study was the small sample of patients and possible cofounders, which warrants future studies.

IRON INFUSIONS, THEIR REACTIONS AND EFFICACY RELATED TO DOSING – A SINGLE CENTER RETROSPECTIVE STUDY

Tejaswi Venigalla¹; Manasa Anipindi²; Madiha Gilani³; David Ezdon⁴; Jermine Mousa⁴; Mitchell Goldstein³; Minal Dhamankar³

¹Internal Medicine, Albert Einstein Medical Center, Philadelphia, PA; ²Internal Medicine, Einstein Medical Center Montgomery, East Norriton, PA; ³Hematology Oncology, Einstein Medical Center Montgomery, East Norriton, PA; ⁴Pharmacy, Einstein Medical Center Montgomery, East Norriton, PA. (Control ID #3855016)

BACKGROUND: Intravenous (IV) iron infusions are associated with a variety of adverse drug reactions (ADRs). The Commonly used IV iron formulations are Iron sucrose (Venofer), Ferric carboxymaltose (Injectafer), Sodium ferric gluconate complex (Ferrlecit), and others.

We aim to study:

1. If altering the dose of Venofer will result in a lesser incidence of allergic reactions.

2. To determine if one preparation of IV iron has a higher incidence of adverse events than another.

3. To evaluate the efficacy after 6 to 8 weeks of completion of iron infusions, by comparing the response to various dosages and formulations.

METHODS: This is a retrospective chart review of 274 patients ages 18-92 who received Intravenous Iron in our Infusion Center. Patients were divided into two groups based on the Venofer dosing schedule. Group D1 patients received any of the above-mentioned IV iron formulations from January to May 2021. In this group, Venofer dosing was 400mg x 1 followed by 300mg x2. Group D2 patients received any of the above- mentioned IV iron formulations from June to October 2021, where Venofer dosing was changed to 300mg x 3 doses or 300mg x 2 followed by 400mg. We collected data on age, gender, race, preparation, strength, total doses received, pregnancy status, prior history of allergies and transfusion reactions, adverse drug reactions, the timing of reaction, pre/post-medications, outcomes of reaction, hemoglobin, iron studies before and after the infusions.

RESULTS: Out of 120 patients in group D1, 59.2% of the patients received Injectafer. Out of these, 17% had ADRs. Responders in this group were 39.4%. 14.1% were lost to follow-up. 40% of the patients received Venofer, out of which 22.9% had ADRs. Responders to the infusion in this group were 22.9% with 20% lost to follow-up. 0.8% received ferumoxytol, with no ADRs.

In the D2 group of 154 patients, 49.3% of the patients received Injectafer, out of which 18.4% had ADRs. Responders in this group were 38.1% and 13.2% were lost to follow-up. 50% of the patients received Venofer, out of which 11.7% had ADRs. Responders to the infusion in this group were 46.8% and 11.7% were lost to follow-up. 0.6% received Ferumoxytol with no ADRs [Table No.1]

CONCLUSIONS: Injectafer was overall associated with more ADRs when compared to Venofer, including anaphylactic reactions. The most common ADR was hypotension with Injectafer, and back pain with Venofer. We found that Venofer dosing with 400mg x 1 followed by 300mg x2 was associated with the highest incidence of ADRs. Both strategies of Venofer dose modification with 300mg x 3 and schedule modification, 300x 2 followed by 400mg had greater efficacy and lesser frequency of ADRs as compared to Venofer 400mg followed by 300mg x2. Other studies have shown the benefit of Venofer in non-renal disease patients with iron deficiency. Although our study was not planned to demonstrate this outcome, our data may suggest that Venofer is beneficial for this group of patients. This should be looked into with further studies.

ISCHEMIC STROKE OUTCOMES AFTER STRESS-INDUCED HYPERGLYCEMIA IN DIABETIC VERSUS NON-DIABETIC PATIENTS

Honglian Duan²; Ho Jun Yun¹; Gary Rajah¹; Yanna Tong²; Zhe Cheng²; Lipeng Cai²; Xiaokun Geng^1,2; Yuchuan Ding¹

¹School of Medicine, Wayne State University School of Medicine, Detroit, MI; ²Department of Neurology, Beijing Luhe Hospital, Beijing, China. (Control ID #3857779).

BACKGROUND: Approximately one-third of patients with acute ischemic stroke (AIS) develop hyperglycemia regardless of pre-existing history of diabetes. Hyperglycemia is associated with increased risk of hemorrhagic transformation and can be considered as a predictor of poor functional outcome after intravenous thrombolysis (IVT) and mechanical thrombectomy (MT). Additionally, risk of death, hemorrhagic complications, and functional dependency are found to increase with stress-induced hyperglycemia especially in those without history of diabetes mellitus. This study assesses whether stress- induced hyperglycemia is a predictor of poor outcome for patients with AIS treated by endovascular treatment (EVT) and they are impacted by previous blood glucose status.

METHODS: This retrospective study collected data from 576 patients with AIS of large vessel occlusion (LVO) treated by EVT from March 2019 to June 2022. The sample was composed of 230 and 346 patients with and without diabetes mellitus (DM), respectively, based on their premorbid diabetic status. Prognosis was assessed with modified Rankin Scale (mRS) score at 3-month after diagnosis of AIS. Poor prognosis was defined as mRS>2. Stress-induced hyperglycemia was assessed by fasting glucose-to-glycated hemoglobin ratio (GAR). Each group was stratified into 4 quartiles of GAR (Q1–Q4). Any significant association between stress-induced hyperglycemia and poor outcomes in both the DM and non-DM groups was assessed. Multivariate logistic regression analysis was used to identify any relationship between different GAR quartiles and clinical outcome after EVT.

RESULTS: In DM group, GAR level was measured to be 1.27±0.44 and poor prognosis was seen in 122 (53%) patients. These variables were higher than non-DM group and the differences were statistically significant (p<0.05, respectively). Patients with more severe stress-induced hyperglycemia showed greater occurrence of poor prognosis at 3 months (in DM, Q1=39.7%, Q2=45.6%, Q3=58.6%, Q4=68.4%, p=0.009; in non-DM, Q1=31%, Q2=32.6%, Q3=42.5%, Q4=64%, p<0.01). After logistic regression, the highest quartile of GAR was independently associated with 3-month poor outcome (OR 0.34, 95%CI 0.16-0.70, p=0.004).

CONCLUSIONS: The outcome of patients with LVO stroke treated with EVT appears to be influenced by premorbid diabetes status. However, the poor prognosis in diabetic patients is not independently associated with stress-induced hyperglycemia. This could be due to the long-term damage resulted from persistent hyperglycemia and/or the adaptive alterations of diabetic patients responding to stress-induced hepatic glucose metabolism after acute ischemic damage to the brain.

IS HIGHER BMI A RISK FACTOR FOR WORSE OUTCOMES IN COVID-19?

Hossny Alaws¹; Gagan Kumar²

¹Internal Medicine, Northeast Georgia Health System Inc, Gainesville, GA; ²Critical care, Northeast Georgia Health System Inc, Gainesville, GA. (Control ID #3874843)

BACKGROUND: Obesity is a growing epidemic in the United States and has been described as a risk factor for severe COVID-19. Most studies, however, have used discreet cut-off points on BMI as their covariates and report poorer outcomes in patients with higher BMI categories. We used restricted cubic splines to examine BMI as a continuous variable to examine its association with COVID-19 outcomes.

METHODS: We performed a retrospective analysis of adult COVID-19 patients aged ≥ 18 years admitted to a large community hospital in northeast Georgia between March 1, 2020, and April 5, 2022.

Outcomes studied were - in-hospital mortality, the rate of intensive care unit (ICU) admissions, invasive mechanical ventilation (IMV), acute kidney injury (AKI), and the use of vasopressors. The Chi-square and the Wilcoxon rank tests were used for categorical and continuous variables, respectively. For all analyses, statistical significance was deemed as a p-value ≤ 0.05.

Restricted cubic splines with 5 knots were used for our analysis. Logistic regression was used to examine the probability of outcomes. All statistical analyses were performed using STATA MP 16.0 (Stata-Corp, College Station, Tx).

RESULTS: Of the 9698 patients admitted in the study period, BMI data were available for 9327 patients. Of these, 2056 had BMI < 25, 5936 had BMI between 25 and 39, and 1335 had a BMI≥40. The severity of COVID-19, as determined by the 4C score, decreased as BMI increased. There was a non-linear relationship between BMI and the outcomes studied. The probability of in-hospital mortality decreased with increasing levels of BMI until a BMI of 30 and then did not change with BMI after that (figure 1). Thus, the odds for in- hospital mortality were 0.55, 0.34, 0.31, and 0,30 at a BMI of 20, 30, 40, and 50, respectively, when compared to a BMI of 14.

Rates of AKI and pressor use increased till a BMI of 30, then flattened out; however, the rates of ICU admission and IMV increased with BMI without any elbow.

CONCLUSIONS: ICU admission and IMV increased with BMI, while rates of vasopressor use, and AKI did not change significantly once BMI was > 30.

The probability of in-hospital mortality decreased with increasing levels of BMI till BMI of 30.

IS UNCONTROLLED DIABETES PRIOR TO HOSPITALIZATION A RISK FACTOR FOR WORSE OUTCOMES IN COVID-19?

Hossny Alaws¹; Gagan Kumar²

¹Internal Medicine, Northeast Georgia Health System Inc, Gainesville, GA; ²Critical care, Northeast Georgia Health System Inc, Gainesville, GA. (Control ID #3874011)

BACKGROUND: Diabetes Mellitus (DM) is a well-described risk factor for hospitalization with COVID-19. However, its unknown whether poor diabetes control prior to admission results in poorer outcomes. We examined whether higher levels of HbA1c prior to admission were associated with poorer COVID-19 outcomes.

METHODS: We performed a retrospective analysis of adult COVID-19 patients (age ≥18 years) admitted to a large community hospital in Northeast Georgia between March 1, 2020, to April 1, 2022.

Outcomes were in-hospital mortality, the severity of COVID-19, the rate of intensive care unit (ICU) admissions, invasive mechanical ventilation (IMV), the use of vasopressors, the rate of hospital-acquired infections (HAI), and rates of readmission. We used the chi-square test and the Wilcoxon rank test for categorical and continuous variables, respectively. For all analyses, we deemed statistical significance a p-value < 0.05.

We used restricted cubic splines with 5 knots for our analysis. Logistic regression was used to examine the probability of outcomes. We performed all statistical analysis using STATA MP 16.0 (Stata-Corp, College Station, Tx).

RESULTS: Of the 9698 patients admitted during the study period, 3894 (40.2%) had diabetes mellitus. Of these, HbA1c data were available for 2672 patients. The severity of COVID-19, as determined by the 4C score, was higher in those with DM (mean 8.6 ± 4.3 vs. 10.9 ± 3.6, p < 0.001). There was non-linear relationship of HbA1c levels and outcomes studied. The rates of ICU admission, mechanical ventilation, HAI, or vasopressor use did not change significantly with higher levels of HbA1c. The probability of in- hospital mortality decreased with increasing levels of HbA1c (figure 1). Thus, the odds for in-hospital mortality were 0.43 (95% confidence interval (CI), 0.29-0.66) at an HbA1c level of 13 when compared to 7. The rates of readmission and AKI decreased with increasing levels of HbA1c.

CONCLUSIONS: HbA1c levels, prior to admission, were not associated with increased severity of COVID-19 as judged by ICU admission, vasopressor use, or rates of invasive mechanical ventilation. Higher HbA1c levels were negatively associated with in-hospital mortality and readmissions.

MEDIUM-TERM MORTALITY IN PATIENTS PULMONARY EMBOLISM MANAGED WITH ULTRASOUND-ASSISTED CATHETER-DIRECTED THROMBOLYSIS COMPARED TO SYSTEMIC THROMBOLYSIS

Catherine A. Ostos¹; Kristina D. Menchaca¹; Virginia Velez Quinones¹; Can Jones¹; Erika Ostos²; Nemanja Draguljevic³; Robert Chait¹

¹Internal Medicine, University of Miami School of Medicine, Miami, FL; ²St Petersburg College - Clearwater Campus, Clearwater, FL; ³Univerzitet u Beogradu Medicinski fakultet, Beograd, Beograd, Serbia. (Control ID #3875021)

BACKGROUND: Pulmonary embolism is one of the leading causes of morbidity and mortality in the United States^1-2. PE is classically treated with anticoagulation and systemic thrombolysis (ST)³. Over the last few decades, there have been important advances in therapeutic options for PE management such as catheter- directed thrombolysis (CDT) with or without addition of ultrasound assistance². However, the use of this devices is still controversial, given the lack of long-term evidence for its outcomes^2-3. We conducted a retrospective study to assess medium-term mortality comparing patients who received treatment with EKOS® Ultrasound-Assisted Catheter-Directed Thrombolysis (US-CDT) compared to Systemic Thrombolysis.

METHODS: This retrospective cohort study identified 1007 patients across several facilities diagnosed with Pulmonary Embolism (PE). Patient were stratified into groups based on the treatment they received. Patients were excluded if they had pregnancy, stroke, or history of recent bleeding. We observed 575 (57.1%) patients treated with EKOS® Ultrasound-Catheter-Directed Thrombolysis (US-CDT) and 432 (42.9%) patients treated with Systemic Thrombolysis (ST). The aim was to assess characteristics of patients and 60-day mortality after treatment, comparing patients who received US-CDT vs ST. Logistic regression was used to predict the 60-day mortality.

RESULTS: Patients’ characteristics were analyzed with 54.22% being male and 45.78% were female. Baseline comorbidities (hypertension, coronary artery disease, chronic obstructive pulmonary disease) were similar among both groups of treatment The patients with PE who were managed ST were more likely to expire within 60 days of the event compared to the ones managed with US-CDT (OR= 3.037, p<0.001, CI 1.856~4.971). Patients who expired were also more likely to experience tachycardia (OR= 2.432, p<0.001, CI 1.299~4.556), and use of vasopressors (OR= 13.331 p<0.001, CI 8.075~22.06) during their hospitalization. A total number of 70 deaths (6.95 %) were identified in the ST group vs. 26 (2.58 %) in the US-CDT group

CONCLUSIONS: Ultrasound-Assisted Catheter-Directed Thrombolysis is an option for treatment that provides less dose of anticoagulation, potentially reducing side effects among other outcomes in patients with PE². Although there has been not enough evidence to differentiate standard CDT vs US-CDT^1-3, our study demonstrated that patients with PE who were managed with ST were more likely to expire within 60 days of the event compared to the ones who had US-CDT (OR= 3.037, p<0.001, CI 1.856~4.971), making US-CDT a safer option for patients with PE. Limitation to our study includes small patient sample. More research is needed to assess their differences.

METFORMIN FOR TREATMENT OF COVID-19: SYSTEMATIC REVIEW OF CLINICAL TRIAL DATA AGAINST SARS-COV-2

Spencer Erickson¹; Sarah Fenno²; Carolyn Bramante³; Jenna Bartley⁴; George Kuchel⁴

¹University of Minnesota Medical School Twin Cities, Minneapolis, MN; ²University of Minnesota Medical School, University of Minnesota Medical School Twin Cities, Minneapolis, MN; ³Medicine, University of Minnesota Twin Cities, Minneapolis, MN; ⁴UConn Health, Farmington, CT. (Control ID #3876554)

BACKGROUND: Observational and preclinical data suggest metformin may prevent severe COVID-19 outcomes. We conducted a systematic review of randomized clinical trials of metformin treatment for COVID-19 to determine whether metformin affects clinical or laboratory outcomes in individuals infected with SARS-CoV-2 and a structured summary of pre-clinical data.

METHODS: Two independent reviewers searched PubMed on September 11, 2022 from 2020-2022 for randomized controlled trials which randomized adults with COVID-19 to metformin versus control and assessed clinical and/or laboratory outcomes of interest. The Cochrane Risk of Bias 2 tool was used to assess bias. The synthesis plan was developed a priori and was guided by SWiM guidelines. Summary tables and narrative synthesis were used.

RESULTS: Three randomized trials met inclusion criteria and were included. Each study utilized a different formulation of metformin; 2 of the trials found metformin improved clinical outcomes, and the 3^rd had a different sample population but did have a direction of benefit in the per-protocol group. The largest of the trials enrolled during the delta and omicron waves, but did not exclude vaccinated individuals. Many preclinical studies have found metformin to be effective against SARS-CoV-2. No safety concerns were identified in any of the trials.

CONCLUSIONS: The certainty of evidence that metformin improves clinical outcomes in patients with SARS-CoV-2 infection was moderate per GRADE criteria and is consistent with preclinical data. Future trials will help define the role of metformin in COVID-19 treatment guidelines.

OUTCOMES AMONG HOSPITALIZED CORONAVIRUS DISEASE 2019 (COVID-19) PATIENTS WITH RHABDOMYOLYSIS: A NATIONWIDE ANALYSIS

Bruce Adrian Casipit¹; Hussein Al-Sudani²

¹Internal Medicine, Albert Einstein Medical Center, Philadelphia, PA; ²Einstein Medical Center Montgomery, East Norriton, PA. (Control ID #3871799)

BACKGROUND: Rhabdomyolysis is a significant cause of morbidity among hospitalized patients. Previous studies suggest that patients with rhabdomyolysis have a higher risk of in-hospital complications including acute kidney injury (AKI), which if severe could lead to renal failure requiring renal replacement therapy (RRT) and maybe even death. However, there are limited data exploring the influence of Coronavirus Disease (COVID-19) infection in this population. We aim to investigate the impact of COVID-19 infection among hospitalized patients with rhabdomyolysis in terms of risk for in-hospital mortality, AKI, and requirement for RRT.

METHODS: We conducted a retrospective cohort study utilizing the Healthcare Cost and Utilization Project National Inpatient Sample (HCUP-NIS) 2020 database which represents 20% of all payer admissions to hospitals in the United States in 2020. Univariate logistic regression analysis was used to calculate unadjusted odds ratios (ORs) for the primary and secondary outcomes. Multivariate logistic regression analysis was used to adjust for potential patient and hospital level confounders. Proportions were compared using the Chi Square test, and a Student t test was used to compare continuous variables. A p value <0.05 was considered statistically significant.

RESULTS: A total of 1,058,815 admissions for COVID-19 infection were identified in 2020, of which, 1.13% (n=11,995) had rhabdomyolysis. Of the total sample, 11,695 met our inclusion criteria. Overall, the inpatient mortality rate among COVID-19 patients with rhabdomyolysis is 23.98% (n=2804/11695). Using a stepwise regression model that adjusted for patient and hospital level confounders, hospitalized COVID-19 patients with concomitant rhabdomyolysis had higher adjusted in-hospital mortality rates (odds ratio [OR] 2.62; 95% confidence interval [CI], 2.34-2.93), risk for AKI (odds ratio [OR] 4.14; 95% confidence interval [CI], 3.75-4.56), renal replacement therapy utilization rates (odds ratio [OR] 3.19; 95% confidence interval [CI], 2.56-3.98), hospital length of stay (p value 0.00) and total hospitalization charges (p value 0.00).

CONCLUSIONS: In our analysis, we concluded that among hospitalized COVID-19 patients with concomitant rhabdomyolysis, there is a higher risk for in-hospital mortality as well as complications including the development of AKI requiring RRT leading to a longer hospital length of stay as well as higher hospitalization charges. This concurs with previous studies suggesting that COVID-19 increases the risk for the development of rhabdomyolysis and AKI possibly through direct cytopathic effects and cytokine storm leading to muscle injury and subsequent renal impairment by means of the nephrotoxic effects of the byproducts of muscle destruction. With this, attention should be focused on mitigating the effects of rhabdomyolysis among hospitalized COVID-19 patients by employing strategies for early recognition and treatment in order to prevent further complications and poor outcomes.

OUTCOMES OF FEBRILE NEUTROPENIA IN CIRRHOTIC VERSUS NONCIRRHOTIC PATIENTS

Reem Dimachkie¹; Zakaria Alameddine¹; Ahmad Mustafa¹; Mira Alsheikh¹; Chapman Wei¹; John Afif¹; Suzanne El-Sayegh¹; Liliane Deeb²

¹Internal Medicine, Staten Island University Hospital, Staten Island, NY; ²Gastroenterology, Staten Island University Hospital, Staten Island, NY. (Control ID #3875138)

BACKGROUND: Liver cirrhosis is associated with high mortality and morbidity due to the resulting portal hypertension. However, it is also well known that patients with liver cirrhosis are at increased risk for serious infections. Febrile neutropenia is defined as fever in a patient with an absolute neutrophil count <1500 cells/microliter. The impact of febrile neutropenia in cirrhotic patients is not well studied. The aim of this study is to highlight the outcome of febrile neutropenia in cirrhotic versus noncirrhotic patients.

METHODS: National Inpatient Sample (NIS) Database 2016-2018 was used to sample individuals presenting with febrile neutropenia, and baseline demographics and comorbidities were collected using ICD-10-DM codes. Patients with missing data were excluded. Individuals were stratified into two cohorts based on the presence or absence of liver cirrhosis. Logistic regression analysis was performed to analyze the outcomes between the two groups.

RESULTS: Out of 116,142 patients admitted with febrile neutropenia, 1418 had cirrhosis. Baseline characteristics and outcomes are summarized in the Table 1. On logistic regression analysis, cirrhotic patients with febrile neutropenia had higher mortality (adjusted OR: 1.3 [1.1-1.6]; p=0.013), gastrointestinal bleed (adjusted OR: 3.8 [3.2-4.6]; p<0.001), sepsis (adjusted OR: 1.2 [1.02-1.40]; p=0.02) and acute renal failure (adjusted OR: 1.4 [1.2-1.6]; p<0.001).

CONCLUSIONS: This study highlights that the presence of liver cirrhosis in patients with febrile neutropenia was associated with higher mortality, gastrointestinal bleed, sepsis, and renal failure. Thus, special care should be taken when managing patients with febrile neutropenia with an underlying diagnosis of cirrhosis. Further studies are needed to establish management strategies for this vulnerable population.

PNEUMOCOCCAL URINARY ANTIGEN TESTING IN PATIENTS WITH SEVERE COMMUNITY-ACQUIRED PNEUMONIA: AN OVERLAP PROPENSITY SCORE-WEIGHTED ANALYSIS

Priscilla Kim^2,1; Michael B. Rothberg⁴; Amy S. Nowacki³; David Gugliotti¹; Abhishek Deshpande¹

¹Medicine, Cleveland Clinic, Cleveland, OH; ²Education, Cleveland Clinic Lerner College of Medicine of Case Western Reserve University, Cleveland, OH; ³Quantitative Health Sciences, Cleveland Clinic, Cleveland, OH; ⁴Internal Medicine, Cleveland Clinic, Cleveland, OH. (Control ID #3871788)

BACKGROUND: Pneumococcal urinary antigen tests (pUATs) facilitate early diagnosis of pneumococcal pneumonia, enabling clinicians to administer targeted therapy and improve antimicrobial stewardship. Current ATS/IDSA guidelines make a weak recommendation to perform pUAT only in patients with severe community-acquired pneumonia (CAP), but few studies have investigated this association. Our objective was to evaluate the association between severe CAP and pUAT positivity.

METHODS: We conducted a retrospective cohort study of adults (>18 years) who were hospitalized with CAP to the Cleveland Clinic Health System from 2010-2019 and were tested with pUAT. Patients were classified by CAP severity (severe vs. non-severe) using three definitions for severe CAP: a CURB-65 score of ≥3, pneumonia severity index (PSI) risk class IV/V, or need for vasopressors, invasive mechanical ventilation, or ICU admission in the first 48 hours of hospitalization (“proxy” definition). Because patients with severe CAP are more likely to have underlying comorbidities, overlap propensity score weighting was performed to address potential confounding. Three separate analyses were performed using each definition for severe CAP, where the primary predictor was CAP severity, and the primary outcome was a positive pUAT result.

RESULTS: A total of 19,282 patients were admitted with CAP and underwent pUAT. The median (IQR) age was 69.1 (57.4-79.9) years, 49.8% were female, and 176.5% were white. Among all patients who underwent pUAT, 51% had a CURB-65 score of ≥3; 68.6% were in PSI risk class IV/V; and 39.6% required vasopressors, IMV, or ICU admission in first 48 hours of hospitalization. A positive pUAT result was observed in 394 (2%) of patients. Overlap propensity score weighting showed no significant association between severe CAP and pUAT positivity by the CURB-65 definition (adjusted OR, 1.11; 95%CI, 0.89-1.37), PSI definition (adjusted OR, 0.99; 95%CI, 0.88-1.23), or proxy definition (adjusted OR, 0.90; 95%CI, 0.72-1.13).

CONCLUSIONS: In this large multi-center sample of patients, severe CAP was not associated with pUAT positivity by any definition for severe CAP. Patients with severe CAP alone may not represent the ideal population for patients who should undergo pUAT. Additional predictors of pUAT positivity should be identified to improve test efficiency.

POST-ACUTE SEQUELAE OF COVID-19 (PASC) AMONG OLDER ADULTS: AN OBSERVATIONAL STUDY

Nancy Haff²; Brandon T. Suehs¹; Richard Sheer¹; Stuart Cowburn¹; Gauri Bhatkhande²; Julie Lauffenburger²; Brian Powers³; Katherine Goodrich³; Niteesh K. Choudhry⁴

¹Humana Healthcare Research, Humana, Louisville, KY; ²Division of Pharmacoepidemiology and Pharmacoeconomics, Brigham and Women's Hospital, Harvard Medical School, Boston, MA; ³Humana Inc, Louisville, KY; ⁴Division of Pharmacoepidemiology and Pharmacoeconomics, Harvard Medical School, Boston, MA. (Control ID #3873726)

BACKGROUND: Many individuals experience prolonged or new symptoms weeks to months after acute infection with COVID, a syndrome now known as post-acute sequelae of COVID (PASC). There is significant heterogeneity in findings that describe the prevalence and trajectory of PASC, likely due to different research methodologies used and populations studied. Additionally, less is known about the prevalence of PASC in older adults.

METHODS: We conducted a cohort study of individuals age ≥65 enrolled in Medicare Advantage plans offered by a large US insurer. We used medical claims, supplemented by COVID test result data, to construct separate cohorts of ambulatory and hospitalized cases of COVID between Apr 1, 2020 and Jun 30, 2021. These case sets were each propensity-score (PS) matched to two comparison groups: 1) contemporary controls with a non-COVID ambulatory visit or negative COVID test, or a non-COVID hospitalization, and 2) ambulatory or hospitalized influenza cases between Jan 1, 2015 and Dec 31, 2018. Patients were followed until disenrollment, death, or end of the study period (Aug 31, 2021). A set of conditions that could represent PASC were identified from existing reports and clinical knowledge and mapped to ICD-9 and -10 diagnosis codes. Conditions observed beginning 28 days after cohort entry, with no codes for the condition prior to cohort entry, were counted as potential post-acute sequelae. In additional analyses we required the presence of two codes to count a condition. Incidence rates were calculated for each potential PASC condition and used to generate incidence rate differences (IRDs) between matched groups.

RESULTS: We identified 107,189 ambulatory and 50,143 hospitalized COVID cases resulting in 88,554 ambulatory and 30,210 hospitalized contemporary non-COVID matched pairs and 28,497 ambulatory and 8,343 hospitalized historical influenza matched pairs. The most common condition was dyspnea, with a cumulative incidence of 10.1% (95% CI 9.8-10.4%) among ambulatory and 20.8% (95% CI 19.9-21.6%) among hospitalized patients in the COVID cohort matched to contemporary non-COVID controls. Fourteen conditions were consistently more common among COVID cases regardless of comparison: dyspnea, fatigue/malaise, diabetic ketoacidosis/hyperglycemia, deep vein thrombosis/pulmonary embolism, pulmonary fibrosis/interstitial lung disease, malnutrition, alopecia, coagulopathy, sleep dysregulation, myopathy, post- traumatic stress disorder (PTSD), thrombocytopenia, olfactory/gustatory dysfunction, and pulmonary function abnormalities.

CONCLUSIONS: In this analysis of older adults with COVID and their PS-matched controls, 14 conditions were consistently associated with COVID, regardless of the comparison group, and likely represent true PASC. Many of these conditions have been observed in prior studies of PASC and are supported by mechanistic data. These results can help inform the care of older adults with PASC and guide further studies of mechanisms and treatments.

PREVALENCE OF POORLY CONTROLLED CHRONIC DISEASES IN RECENTLY HOSPITALIZED PATIENTS

Alexander Chaitoff¹; Stephanie Mueller²; Nancy Haff²; Julie Lauffenburger³; Niteesh K. Choudhry⁴

¹Medicine, Brigham and Women's Hospital, Boston, MA; ²Medicine, Brigham and Women's Hospital Department of Medicine, Boston, MA; ³Division of Pharmacoepidemiology and Pharmacoeconomics, Brigham and Women's Hospital, Harvard Medical School, Boston, MA; ⁴Division of Pharmacoepidemiology and Pharmacoeconomics, Harvard Medical School, Boston, MA. (Control ID #3875506)

BACKGROUND: Evidence suggests that significantly disruptive health events, such as hospitalizations, may increase patients’ receptiveness to interventions that lead to healthier lifestyle choices and medication adherence. Opportunities may exist to harness hospitalizations to address a broad array of chronic diseases, but the scope of those opportunities is not well-described. This study aimed to characterize the prevalence of poorly controlled hypertension, hyperlipidemia, and diabetes among recently hospitalized community- dwelling adults.

METHODS: We conducted an analysis using the National Health and Nutrition Examination Survey (NHANES), which is a survey that samples non-institutionalized civilians to provide nationally representative estimates. We included adults enrolled in NHANES between January 2017 to March 2020 who reported an overnight hospitalization in the preceding year. Respondents were considered to have a potential opportunity for a chronic disease intervention if they had poorly controlled hypertension, hyperlipidemia, or diabetes at the time of being surveyed. Poorly controlled chronic diseases were defined as: 1) hypertension: self-reported diagnosis of hypertension and a blood pressure reading >140/90 mmHg during the exam portion of the survey, and on 1 or more anti-hypertensive medications; 2) hyperlipidemia: aged 40-70, ASCVD risk score >7.5%, and not on lipid lowering therapy; and 3) diabetes: hemoglobin a1c>7% and age <65 years. Frequencies and counts were used to describe the combined population of patients with a potential opportunity for chronic disease intervention during hospitalization. Chi-square and t-tests were used to assess for respondent characteristics associated with a potential opportunity for chronic disease intervention. Percentages are weighted while counts are unweighted unless otherwise specified.

RESULTS: Of all surveyed adults, 935 had a hospitalization in the previous year. Among those with a hospitalization, 23.6% (95%CI: 20.5%-26.8%, n=282), corresponding to a weighted count of 5,563,810 individuals (95%CI: 4,394,885-6,732,735), had poorly controlled hypertension, hyperlipidemia, or diabetes at the time of their survey and therefore could be potentially eligible for intervention. Those who were potentially eligible for intervention were on average older (63.6 vs 51.6 years, p<0.001), more likely to be male (29.1% versus 20.1%, p=0.003), and more likely to not have a high school degree (18.4% versus 11.8%, P<0.001), but there were no differences by race/ethnicity or family income-to-poverty ratio (both p>0.05)

CONCLUSIONS: We found a high prevalence of poorly controlled chronic diseases in the year following discharge. This supports a potential opportunity for hospital-based chronic disease management efforts, especially non-pharmacologic ones, which should be further studied using data sources that can clarify temporal associations between disease control and hospitalization.

SYMPTOMS CONSISTENT WITH POST-ACUTE SEQUELAE OF COVID (PASC) IN TWO HEALTH SYSTEMS

Ingrid A. Binswanger¹; Jason M. Glanz¹; Jennifer Barrow¹; Courtney Kraus¹; Komal J. Narwaney¹; Katia J. Bruxvoort³; Jason A. Lyons¹; Jessica A. Lam²; Darryl Palmer-Toy²

¹Institute for Health Research, Kaiser Permanente Colorado, Denver, CO; ²Kaiser Permanente Southern California Department of Research & Evaluation, Pasadena, CA; ³School of Public Health, The University of Alabama at Birmingham, Birmingham, AL. (Control ID #3875455)

BACKGROUND: The understanding of post-acute sequelae of COVID (PASC) is rapidly evolving. The objectives of this study were to (1) estimate the prevalence of symptoms consistent with PASC, including persistent, recurrent, and new symptoms in a population-based sample, and (2) among people with a history of documented COVID, assess the association between SARS-CoV-2 seropositivity and symptoms twelve or more weeks later.

METHODS: We recruited English and Spanish-speaking members ≥18 years old from Kaiser Permanente (KP) Colorado and KP Southern California who had viral and serological (nucleocapsid or spike protein- based) testing as of August 2021. Between October 2021 and April 2022, participants completed an online or mailed survey assessing socio-demographics, prior COVID infection and testing, and potential PASC symptoms using PROMIS items for physical health, mental health, cognitive function, anxiety, and depression. Survey results were linked to electronic health record data on lab results. We conducted descriptive analyses and generated T-scores standardized to the (pre-COVID) US population (T-score 50.0; higher score=worse function) by prior antibody and viral testing results (+/+, +/-, -/+, -/-). T-scores were adjusted by study site, age, race, ethnicity, and time since the serology test date and survey date.

RESULTS: Across the two regions, 3947 members consented, completed the survey, and were included in the analysis (25% response rate). The mean respondent age was 52 years old (SD 15.6), the majority were female (68%), and 28% identified as Hispanic. Among participants who reported having had COVID, 49.8% reported having at least one or more symptoms that lasted over 12 weeks, with the most common being fatigue (56.6%), brain fog or focus issues (48.3%), and shortness of breath (41.6%). Fifty percent of respondents who reported having had COVID reported persistent symptoms, 29% recurrent, and 23% new symptoms after the initial infection. After a median of 15 months after serology testing, self-reported physical health and cognitive function were significantly worse among people with prior infection and positive serology (+Ab/+PCR) compared to those who did not have prior infection and had negative serology (-Ab/-PCR) but overall close to the US population norms pre-COVID. There were no significant differences in mental health, anxiety, and depression observed.

CONCLUSIONS: High proportions of patients reported they had persistent, recurrent, and new symptoms after COVID. After a median of 15 months, differences between groups based on prior viral and serologic testing demonstrated small but significant differences in physical and cognitive health, but no differences in several domains of mental health and overall values similar to the general population pre-COVID.

THE ASSOCIATION BETWEEN ANTIBODIES TO SARS-COV-2 AND REINFECTION IN COLORADO A PROSPECTIVE COHORT STUDY

Ingrid A. Binswanger, Jennifer Barrow, Komal J. Narwaney, Courtney Kraus, Jo Ann Shoup, Claudia Steiner, Jason M. Glanz

Institute for Health Research, Kaiser Permanente, Colorado, Denver, CO. (Control ID

#3872604)

BACKGROUND: In April 2020, antibody tests to detect SARS-CoV-2 immunoglobulins were approved under FDA Emergency Use Authorization. However, the association between antibody test results and SARS-CoV-2 reinfection risk in community-based samples and spanning the Omicron phase has not been well quantified, and the CDC cautions against using the tests to influence vaccine decisions or other preventive behaviors. Within a population-based cohort of Kaiser Permanente Colorado members, we conducted a nested case-control study to examine the association between antibody test results and the incidence of reinfection.

METHODS: Participants were recruited between June 2020 and March 2021, surveyed, and encouraged to complete viral and antibody laboratory testing every 28 days for 10 months. Participants with reinfections (positive SARS-CoV-2 viral test ≥90 days after the first positive viral test) were matched to control participants without reinfections by age, sex, first date of a positive viral test, last antibody test date, and antibody test type (nucleocapsid or one of 3 spike tests). Using conditional logistic regression, we compared the odds of a seronegative antibody test result between case and control participants, adjusted for demographic and clinical covariates. To account for vaccination status, we used an interaction term to assess the association between seronegativity and reinfection by the timing of antibody test relative to vaccination: not vaccinated, antibody test on the same date/after the first vaccine, and antibody test before the first vaccine.

RESULTS: The cohort (n=4235) included 2033 participants with ≥1 positive viral test, 120 of whom had a reinfection (0.52/100 person-months). Among 80 case patients who could be matched, the last antibody test was negative in 12 cases (15.0%) whereas the last antibody test was negative in 77 of 1043 (7.5%) of control participants. Seronegativity (aOR, 2.73; CI, 1.34-5.54), Hispanic ethnicity (aOR. 1.86; CI, 1.13-3.08), and household size (aOR, 1.14; CI, 1.01-1.28 for each additional household member) were associated with reinfection. When stratifying by vaccination status, the association between seronegativity and reinfection was stronger among those who were not vaccinated (aOR 4.19; 95% CI 1.28, 13.76) and among who had a vaccination on the same day or before the antibody test (aOR 6.30; 95% CI 1.66, 23.94) than among those who were vaccinated after the antibody test (aOR 1.08; 95% CI 0.21, 5.48).

CONCLUSIONS: Seronegative antibody status was associated with a greater than two-fold risk of COVID reinfection, suggesting serological testing may be useful to assess the risk of reinfection. Given the morbidity and mortality associated with reinfection, leveraging serology to support preventive efforts should be considered.

THE EFFECT OF RESTRICTIVE RED BLOOD CELL TRANSFUSION ON COGNITIVE FUNCTION IN HOSPITALIZED PATIENTS

Stephanie Cardenas¹; Christian Carrier¹; David O. Meltzer²; Micah T. Prochaska²

¹Department of Medicine, University of Chicago Pritzker School of Medicine, Chicago, IL; ²Medicine, The University of Chicago, Chicago, IL. (Control ID #3876664)

BACKGROUND: Anemia is a risk factor for cognitive impairment as a result of the reduced delivery of oxygen to the brain from low hemoglobin (Hb) levels. In ambulatory patients even mild anemia has been associated with reductions in patients’ cognitive function (CF). However, in hospitalized patients it is standard of care to maintain Hb levels far lower, around 7g/dL, as a result of restrictive transfusion practices. This raises the possibility that the lower Hb levels tolerated in hospitalized patients may have a deleterious effect on patients’ CF. This is important because reduced CF negatively effects patients’ self-efficacy and functioning, can increase the risk of delirium and length of stay, and may ultimately impact discharge destination and recovery after hospital discharge. While transfusion trials in hospitalized patients have shown no difference in outcomes between restrictive vs. liberal transfusion thresholds, these RCTs did not examine the effect that lower Hb levels may have on patient’s CF. Therefore, the purpose of this study was to test for an association between Hb levels during hospitalization and patient’s CF.

METHODS: Hospitalized general medicine patients at an urban academic medical center with a Hb <10g/dL were approached for an interview. CF was measured using the Short Portable Mental Status Questionnaire (SPMSQ), which has a range from 0-10 with higher scores indicating worse CF. Patients’ Hb values and clinical data were abstracted from hospital administrative databases. Patients with a diagnosis of dementia were excluded from the analysis. Linear regression was used to test the association between CF (SPMSQ) as the dependent variable and patients’ nadir Hb and age as the predictor variables of interest, controlling for race, gender, Charlson Comorbidity Index, a diagnosis of depression, and receipt of transfusion. An interaction between patient’s nadir Hb and age was also included to test for changes.

RESULTS: A total of 4,418 patients had a Hb<10g/dL and completed the SPMSQ questionnaire. The average age was 56, 58% were female, and 74% were Black. In the regression models, lower nadir Hb levels were associated with reduced cognition (higher SPMSQ) (β=-0.22, p<0.01) overall. Additionally, the effect of lower Hb level on reduced cognition was greater within each decreasing Hb strata compared to patients whose Hb was between 9-10g/dL during their hospitalization (Hb: 8-9g/dL,β=0.03, p<0.01; Hb: 7-8g/dL,β=0.27, p<0.01; Hb: 6-7g/dL, β=0.55, p<0.01; Hb<6g/dL, β=0.78, p<0.01). The interaction between a patient’s nadir Hb and their age was also associated with significant reductions in CF (β=-0.004, p<0.01).

CONCLUSIONS: Lower Hb levels during hospitalization, particularly the Hb levels tolerated as a result of restrictive transfusion policies, are associated with significant decreases in CF. Clinicians should be aware of and attentive to the effect of anemia on CF, and should consider screening for and monitoring high risk patients with anemia for decreased CF.

TRENDS OF VITAMIN C IN ADULT PATIENTS

Hannah Hoang, Isabella Kelly, Adil Mohyudin, Pradeep Bathina

Internal Medicine, The University of Mississippi Medical Center, Jackson, MS. (Control ID #3871958)

BACKGROUND: Vitamin C deficiency (VCD), rare in developed countries, still affects patients, making it an important diagnosis to recognize. Manifestations of VCD cause collagen synthesis impairment due to vitamin C-dependent hydroxylases which form stable triple helices in collagen. Deficiency leads to fragile connective tissue which presents as mucosal bleeding, bruising, impaired healing, anemia, and poor immune response. VCD occur in people on diets deficient in fruits and vegetables. Diagnosis is largely based on history and physical examination. Serum testing of the vitamin C (<0.4 mg/dL) is consistent with deficiency. This study will evaluate the trends of vitamin C levels at UMMC.

METHODS: Cross Sectional Observational study was performed on adult patients (age >18) who received a Serum Vitamin C level at UMMC from January 1, 2013 to December, 2021. Patient Cohort Explorer was used to obtain de-identified patient data from EPIC. We obtained the data for patients on whom the serum Vitamin C level was ordered and used Microsoft Excel, Stata 17.0 for calculations.

RESULTS: Serum Vitamin C level was ordered 439 times on 371 patients. 262 (59.68%) tests were done in females and 177 (40.32%) in males. 187 (42.69%) tests were done in African Americans and 238 (54.34%) tests in Caucasians.

Of the 439 tests, 199 tests were abnormal. 196 (44.65%) came back low with level <0.4 mg/dl, the lower cut off value. Only 3 tests (0.68%) came back high with level >2 mg/dl, the maximum cut off value. Of the 196 tests that came back low, 89 (20.27%) tests came back with level <0.1 mg/dl, the minimum detectable level 106 of 262 tests (40.45%) in females resulted low while 90 of 177 tests (50.84%) resulted low in males. For African Americans 86 of 187 tests (45.98%) resulted low. In Caucasians 105 of 238 tests (44.11%) resulted low 140 tests were done on 128 patients while admitted in the hospital, while the rest were done as outpatient. Of the 140 tests done as inpatient, 72 tests were abnormal. 70 (50%) tests came back low with level <0.4 mg/dl. 2 tests (1.48%) came back high with level >2 mg/dl. Of the 70 tests that came back low 35 (25%) tests came back with level <0.1 mg/dl, the minimum detectable level.

Alcoholic cirrhosis with Ascites is a commonly listed primary diagnosis at 2.73%.

CONCLUSIONS: This is the largest known data sample at a university medical center. One in two serum Vitamin C sample levels was lower than the minimum cut off value, and one in four samples had undetectable levels. Though patients may not reflect the deficiency in the general population, the prevalence of VCD in tested patients was alarming. More studies are needed to study risk factors of VCD especially with chronic wounds.

YIELD OF BIDIRECTIONAL ENDOSCOPY FOR ASYMPTOMATIC IRON DEFICIENCY ANEMIA IN AN INNER CITY POPULATION

Arielle Greenberg¹; Lawrence J. Brandt²

¹Internal Medicine, Montefiore Medical Center, Bronx, NY; ²Montefiore Medical Center, Bronx, NY. (Control ID #3869811)

BACKGROUND: Internists are often faced with evaluating patients with iron deficiency anemia (IDA) and the decision to refer to Gastroenterology (GI) for endoscopic evaluation of occult bleeding. Current GI national guidelines recommend bidirectional endoscopy (BDE) for the evaluation of IDA. The prevalence of anemia is 1.8 billion people worldwide and is 3 times more common in African Americans than Whites. Iron deficiency is the most common cause of anemia. Our institution is a Mid-Atlantic academic medical center located in an inner-city. Among our population, greater than 43% of residents are Black and greater than 54% are Hispanic. The diagnostic yield of BDE in IDA without gross GI bleeding in an inner-city patient population like ours is unknown.

METHODS: The study was conducted retrospectively with IRB approval. Inclusion criteria included adults with IDA based on hemoglobin, ferritin, and percent saturation. Patients with frank GI bleeding were excluded. Data from 2020-2021 were reviewed and patients were grouped into 4 source categories: (1) likely/definitive source of IDA identified; (2) possible source identified; (3) no source identified; and (X) source identified by pathology alone with no endoscopic lesions.

RESULTS: 102 patients (35.3% Black, 42.2% Hispanic) underwent BDE. 42.2% of patients were in source categories 1 and 2 (definitive or possible source found). Among esophagogastroduodenoscopies (EGDs) 27.5%, 5.9%, 53.9%, and 12.7% of patients were in source categories 1, 2, 3, and X respectively. Among colonoscopies 8.8%, 0%, 91.2%, and 0% of patients were in categories 1, 2, 3, and X respectively. No source categories had significantly different hemoglobin, ferritin, or percent saturation values from our overall sample. Sources identified with EGD included erosive gastropathy, duodenal erosions, polyps, angiodysplasias, atrophic gastritis, and malignancy. Patients in category X had biopsy proven H. pylori, which can impair iron absorption. Sources of IDA identified with colonoscopy included angiodysplasias and large polyps.

CONCLUSIONS: In our inner-city population, 49% had a possible or definitive GI cause for their IDA. 46.1% had sources found on EGD and 8.8% on colonoscopy. There are limitations to this study. Our patients often underwent colonoscopy before EGD. Some patients with sources of IDA on colonoscopy might not have gone on to EGD or been included in our sample. Including such patients would have resulted in a higher yield. 6 patients had sources found on both EGD and colonoscopy, so despite findings on unidirectional endoscopy, BDE should be completed for a complete evaluation.

Scientific Abstract - Adolescent Medicine and Transitioning to Adult Care

IMPROVING THE TRANSITIONAL CARE PROCESS FOR ADOLESCENT PATIENTS WITHIN THE PRIMARY CARE SETTING

Esha M. Kapania, Gianna Bosco, Suwon Nopachai

Medicine/Pediatrics, Rush University Medical Center, Chicago, IL. (Control ID #3857032)

BACKGROUND: The transition from pediatric to adult-oriented care can be a challenging time for both patients and providers. While it is well recognized that a successful transition is critical for improving health outcomes and costs, there continues to be limited infrastructure and education for healthcare providers to aid with this process. The objective of this study was to identify what providers felt are barriers to providing successful transitions of care and implement subsequent interventions to improve the process. Our hypothesis is that by increasing education and support around transitions of care, that providers would feel more empowered and comfortable conducting this process. Our goal was that through this process, there would be an increase in the number of providers who address transitional care readiness and participate in coordinating care between pediatric and adult practices.

METHODS: We sent out a survey to the pediatric department at Rush University Medical Center assessing their current level of comfort with healthcare transition (HCT) and identifying potential barriers to successfully transitioning a patient to adult providers. Based on the results, the following four interventions were completed: education sessions were held with providers regarding key aspects of HCT, a formal list of adult sub-specialists was compiled and distributed to pediatric providers, an educational packet was created for patients, and the EMR template for adolescent healthcare visits was revised to prompt providers to discuss age appropriate aspects of transitions of care. After 6 months with the above interventions, the same survey was sent back out to providers to assess if the interventions impacted provider comfort level with HCT. A Chi-squared test was used to conduct statistical analysis of the pre-and post intervention data.

RESULTS: The results of the pre and post intervention surveys showed significant improvement in perceived familiarity with the transition of care process, as well as increased comfort level when discussing HCT with patients. Additionally, more providers expressed that they were addressing transitional care readiness with their patients and participating in coordination of care between pediatric and adult practices. There was however, no significant difference noted in the comfort level that providers had working with adolescents with chronic diseases of childhood onset.

CONCLUSIONS: Providing education and support influenced more providers to address transitional care readiness and participate in coordination of care between pediatric and adult practices

MASCULINITY AND TOXINS: EXPLORING CONNECTIONS BETWEEN ADOLESCENT MALE GENDER EXPRESSION, SOCIAL NETWORKS, AND ADULT SUBSTANCE USE

Nathaniel J. Glasser¹; Jacob Jameson²; Elizabeth L. Tung¹; Stacy T. Lindau^3,4; Harold A. Pollack^5,6

¹Section of General Internal Medicine, University of Chicago Division of the Biological Sciences, Chicago, IL; ²Harvard University T H Chan School of Public Health, Boston, MA; ³Ob/Gyn and Geriatics and Palliative Care, University of Chicago, Chicago, IL; ⁴Department of Medicine -- Geriatrics and Palliative Medicine, University of Chicago Division of the Biological Sciences, Chicago, IL; ⁵University of Chicago Crown School of Social Work, Policy, and Practice, Chicago, IL; ⁶Public Health Sciences, University of Chicago Division of the Biological Sciences, Chicago, IL. (Control ID #3872099)

BACKGROUND: Endorsem*nt of stereotypically male gender identity through behavioral gender expression (GE) is associated with risk-taking behaviors in adolescence and adulthood. We examine associations between adolescent social network characteristics, adolescent-to-adult changes in male GE, and adult substance use.

METHODS: This prospective cohort study uses 1994-95 (Wave I) and 2016-18 (Wave IV) data from the National Longitudinal Study of Adolescent to Adult Health (Add Health). Participants are a nationally representative sample of adolescents (ages 11-18), followed longitudinally to young adulthood (ages 24-32). Associations between GE, social network variables, and substance use are modeled using multivariable regression, adjusting for participant- and neighborhood-level covariates. Primary exposures are population- standardized male GE, calculated in both waves; adolescent friendship network characteristics; and school- standardized adolescent GE. Primary outcomes are self-reported young adult prescription drug misuse, recreational drug use, marijuana use, smoking, and heavy drinking.

RESULTS: 4,882 (48%) of 10,263 males enrolled in Wave I completed both Wave I and IV surveys and were eligible for inclusion. Higher adolescent male GE was associated with greater young adult cigarette use (β=0.10, p<0.05), marijuana use (β=-0.12, p <0.01), and recreational brug use (β=0.11, p<0.05). Adult male GE and adolescent-to-adult increases in male GE were significantly associated with all adult substance use behaviors. Significant adolescent-to-adult changes in GE were predicted by school-standardized GE. Adolescents whose GE scores were above or below their school’s average GE were significantly more likely to change GE between waves in the direction of their school’s average (β=-0.80, p<0.01), relative to the overall population of males in each wave.

CONCLUSIONS: Adolescent GE, adult GE, and increased adolescent-to-adult GE are associated with greater adult substance misuse among men. An adolescent male’s positioning in apposition to his high school social milieu is associated with change in GE over time, with marked regression to the school mean. Interventions that influence adolescent group gender norms may be valuable and unexplored pathways to reduce substance misuse across the life course.

REPRESENTATION OF GIRLS IN ATTENTION-DEFICIT HYPERACTIVITY DISORDER RESEARCH

Mary Rozelle¹; Bradley Peterson²; Joey Trampush³; Maria Bolshakova¹; Aneesa Motala⁴; Susanne Hempel¹

¹Population and Health Sciences, University of Southern California, Los Angeles, CA; ²Psychiatry, Children's Hospital Los Angeles Division of Adolescent and Young Adult Medicine, Los Angeles, CA; ³Psychiatry and Behavioral Sciences, Keck Hospital of USC, Los Angeles, CA; ⁴Keck Hospital of USC, Los Angeles, CA. (Control ID #3834322)

BACKGROUND: Girls are underrepresented in the diagnosis and treatment of attention-deficit hyperactivity disorder (ADHD). Girls also present ADHD differently than boys in symptoms, functions, and associated comorbidities. The less disruptive behaviors of girls with ADHD may account for lower diagnostic rates and referrals. Impulsivity and risk-taking behaviors place girls at increased risk of self-harm, early- sexual debut, perceived sexual exhibitionism, and vulnerability to stigmatized behaviors. We studied the extent to which ADHD intervention research studies included girls.

METHODS: The American Academy of Pediatrics nominated an Agency for Quality and Research (AHRQ) evidence report update on ADHD in children and adolescents to support new clinical guidance, addressing multiple questions regarding the diagnosis, treatment (pharmacologic and nonpharmacologic), and monitoring. The review identified 18,055 citations and involved searches in eleven databases. Interventions included medication, behavioral therapy, supplements and others, published since 1980. Two independent reviewers screened publications against prespecified eligibility criteria to identify studies in children and adolescents up to 18 years old diagnosed with ADHD with or without concomitant conditions. From 5,498 full-text publications screened, data were abstracted on the proportion of girls from 263 identified intervention studies. We evaluated data using descriptive analysis and random-effects meta-analysis.

RESULTS: The number of girls studied overall was small, and several identified studies were conducted exclusively on boys. The percentage of girls in the studies ranged from 0% to 64%. Most studies included less than a third of girls (median 24.3%, IQR 17.6-30.0), a pooled result of 23.7% girls (95% CI 22.5%, 24.9%). Compared to the population's prevalence of ADHD, girls in the research studies were underrepresented.

CONCLUSIONS: Intervention research in ADHD predominately includes boys. Interventions do not target girls, and the small proportion of girls hinders analyses of potential differential treatment effects for boys and girls. Clinicians should consider interventions that meet the interests and needs of girls diagnosed with ADHD, as their psychosocial needs differ from boys. However, current research practice does not support high-quality evidence about the appropriateness of interventions for girls.

THE CLINICAL AND ECONOMIC IMPACT OF CHILD AND ADOLESCENT FIREARM INJURIES ON SURVIVORS AND FAMILY MEMBERS

Zirui Song^1,2; Jose Zubizarreta^1,3; Mia Giuriato¹; Katherine Koh^4,5; Chana A. Sacks²

¹Health Care Policy, Harvard Medical School, Boston, MA; ²Massachusetts General Hospital Department of Medicine, Boston, MA; ³Statistics, Harvard University, Cambridge, MA; ⁴Massachusetts General Hospital Department of Psychiatry, Boston, MA; ⁵Boston Health Care for the Homeless Program, Boston, MA. (Control ID #3875581)

BACKGROUND: More US children and adolescents today die from guns than any other cause. Yet often forgotten is many more of their peers who are shot but survive. Even larger is the number of family members —parents and siblings—of these victims. To date, evidence on the health and economic impact of firearm injuries on children/adolescents and their family members remains scant. Existing literature is limited by small samples, lack of claims data, or lack of control groups. Indeed, most studies on guns exclude children.

METHODS: Using 2007-2020 Marketscan commercial claims, we defined 2 exposed groups: (1) people under age 20 with a firearm injury, and (2) their family members. Each exposed subject was matched to 5 control subjects without replacement using risk set (cardinality) matching—exact matched on year, month, sex, plan attributes, and MSA, while balanced on mean age and DxCG risk score.

We used an event study, difference-in-differences model to examine changes in spending, utilization, and morbidity from 1 year before through 1 year after firearm injury, with covariate adjustment to improve precision.

Sensitivity analyses used randomization tests and inverted the permutational t-test on the effect estimates. We tested sensitivity to unmeasured confounding using Rosenbaum bounds to calculate how large a confounder would need to be to explain away the effects.

We further conducted subgroup analyses by the type and severity of firearm injury, as well as the subgroup of family members of children/adolescents whose firearm injuries were fatal.

RESULTS: We identified 2,002 child/adolescent survivors and 9,931 matched controls, along with 6,040 family members of survivors matched to 29,963 controls, balanced on age, sex, risk score, plan attributes, and MSA (standardized mean differences 0 to 0.02).

Survivors had an increase in medical spending of $2,661 per survivor per month relative to control in year 1—a 6.2-fold increase from baseline (p<0.001). They also experienced a 112% increase in pain syndromes, 59% increase in psychiatric disorders, and 96% increase in substance use disorders relative to control (p<0.001). Family members of survivors had a 13.5% increase in psychiatric disorders in year 1 (p=0.008) with a 10.8% increase in ER visits (p=0.04) and 4.3% increase in office visits (p=0.007), relative to control. Family members of children/adolescents killed by guns experienced a larger 3.3-fold increase in psychiatric disorders and 72% increase in psychiatric medication days relative to control (p<0.001).

Intentional and more severe (ICU) firearm injuries drove the results. Estimates were robust to sensitivity analyses.

CONCLUSIONS: Firearm injuries in child and adolescent survivors increased their mental and physical disease burden as well as societal spending on their care. Family members also saw increased psychiatric illness, especially those of children/adolescents killed by guns. This illustrates the ripple effects of gun violence from young victims, through families, and to society.

Scientific Abstract - Career Development, Professionalism, and Wellness

APPLYING A POSITIVE DEVIANCE APPROACH: ANALYZING AAMC STANDPOINT SURVEY DATA TO ADDRESS FACULTY BURNOUT THROUGH SHARED ORGANIZATIONAL LEARNING

Robert G. Badgett¹; Jon A. Courand²; Jennifer L. Hartwell³; Catherine F. Pipas⁴; Valerie Dandar⁵

¹Internal Medicine, University of Kansas School of Medicine Wichita, Wichita, KS; ²Pediatrics, The University of Texas Health Science Center at San Antonio, San Antonio, TX; ³Surgery, University of Kansas School of Medicine, Kansas City, KS; ⁴Community & Family Medicine, Geisel School of Medicine at Dartmouth, Lebanon, NH; ⁵Medical School Operations, Association of American Medical Colleges, Washington, DC. (Control ID #3875072)

BACKGROUND: Workforce burnout and engagement are associated with organizational performance outcomes. The AAMC StandPoint Engagement Survey uniquely allows the measurement of differences (or heterogeneity) of burnout rates of academic departments within and across schools. If heterogeneity is found, then appreciative inquiry or positive deviance assessment of “bright spots” can guide organizational improvement efforts and ultimately, well-being and engagement outcomes. The importance of the organization was shown by Shanafelt’s finding that 47% and 11% of the variation in physician satisfaction and burnout, respectively, at an organization can be explained by the leadership score of their immediate supervisor.

METHODS: This study measures heterogeneity in rates of burnout across departments and schools and factors associated with low departmental rates of burnout. We studied AAMC StandPoint data from 22 academic medical centers (AMCs) surveyed from 2020-2022, which included 359 unique academic units containing 27 disciplines. Burnout was defined with the single-item Mini-Z.

RESULTS: 54% of full and part-time faculty completed the survey (11,044/20,584). Of those, 9,873 (2867 (29%) reported being burned out (estimated Maslach rate: 42%) and 81% reported dedication. Three of the 22 AMCs had lower-quartile burnout and upper-quartile dedication. Regarding burnout, there was "substantial" heterogeneity (I² = 75%; range 18% to 41%) across AMCs and "moderate" heterogeneity (I² = 48%; range 18% to 38%) across disciplines.

Regarding the variation in rates of burnout across departments, 15% was explained by the rates of burnout in the AMC and discipline (10% and 5%, respectively).

The variance in burnout scores within individuals was significantly explained by the mean burnout score in the respondent's department (R² = 6%) with no additional explanation provided by the medical center or discipline. After accounting for the department, identifying as a woman (R2 = 2%) and junior rank (R² < 1%) were significant, albeit lesser risk factors.

As an example, within an AMC selected for moderate heterogeneity (I² = 32%) in rates of burnout across departments, the strongest predictor of burnout was a lack of a sense of accomplishment (R² = 10%), followed by lack of a sense of autonomy (R² = 5%) and sense of belonging in a department (R² = 4%).

CONCLUSIONS: There is heterogeneity in the rates of burnout across organizations and disciplines. The strongest predictor of burnout in the individual is the mean burnout score in their department. The strongest predictor of burnout in a department was rates of burnout in the AMC, more than the discipline. Thus, we conclude that “local factors” dominate in driving burnout and should be the focus of proposed interventions.

The finding of heterogeneity in burnout rates can guide improvement via a positive deviance strategy. We invite schools to participate in this AAMC project.

A QUALITATIVE STUDY OF PROFESSIONAL IDENTITY FORMATION AMONG FEMALE INTERNAL MEDICINE TRAINEES

Samantha J. Cheng¹; Scheherazade Elkeshk³; Tamara Goldberg²

¹Neurology, Montefiore Medical Center, Bronx, NY; ²Internal Medicine, Mount Sinai Morningside Hospital, New York, NY; ³Psychiatry, Hackensack Meridian Hackensack University Medical Center, Hackensack, NJ. (Control ID #3873624)

BACKGROUND: Postgraduate medical training represents a crucial period in professional identity formation (PIF) for physician trainees. The ease of transition from medical student to physician can be challenging, particularly for those with pre-existing identities that have historically been marginalized within the profession. This study seeks to understand how gender informs the professional identity formation of female internal medicine residents over the course of residency training.

METHODS: Using a qualitative approach based in grounded theory framework, we pursued a nuanced understanding of the adaptive process of PIF among internal medicine trainees. All self-identified females at a single internal medicine residency program in New York City were invited via email to participate in a focus group session. Using a semi-structured moderator guide, PGY-specific focus groups were conducted among 11 total participants in December 2021. Focus groups were transcribed, and an iterative 3-phase coding process was employed to identify central themes for each focus group.

RESULTS: Based on focus-group dialogue, PIF for female resident participants was a 3-phase process impacted by peer support and mentorship. PGY-1 participants described “identity dissonance,” which was characterized by frequent tensions between personal ideals of inhabiting the physician role and the reality of how they are treated by others and what is expected of them. They expended much time cognitively processing dismissive and negative interactions, which impacted their professional confidence. PGY-2 women described “playing the role,” a performative trial-and-error approach towards internalizing their role and responsibility as leaders. PGY-2 participants were attuned to how they were treated compared to male peers, and were conscientious to provide support to other women. By PGY-3 year, participants had learned to “inhabit the role,” demonstrating internalization of their professional competence, but increased clarity and frustration towards gender power dynamics. Peer support and mentorship bolstered their confidence, encouraging them to confront biases.

CONCLUSIONS: Among female internal medicine residents at a single institution, professional identity formation encompasses a 3-phase process, aligned with postgraduate year, which is bolstered by peer support and mentorship. By elucidating this process, we hope educators and program directors can better tailor strategies to create inclusive learning environments for their trainees that support PIF.

A SYSTEMATIC REVIEW AND META-ANALYSIS OF GENDER DISPARITY IN ACADEMIC MEDICINE PROMOTION

Elizabeth Marhoffer^2,1; Samer Ein Alshaeba^2,1; Alyssa Grimshaw⁴; Jurgen Holleck^2,1; Benjamin Rudikoff¹; Lori Bastian^3,1; Craig Gunderson^2,1

¹VA-Connecticut, West Haven, CT; ²Internal Medicine, Yale School of Medicine, New Haven, CT; ³Internal Medicine, Yale University School of Medicine, West Haven, CT; ⁴Cushing/Whitney Medical Library, Yale University, New Haven, CT. (Control ID #3873101)

BACKGROUND: It is well established that there is a persistent gender gap in promotion in academic medicine despite an equal number of male and female medical students for the past 20 years. Possible mediators of this gender gap include differences between men and women in years on faculty, measures of productivity such as number of publications and grants, leadership positions and direct gender bias. We performed a systematic review of studies that report rates of promotion to full professor between men and women adjusted for potential mediators.

METHODS: A systematic search of Academic Search Premier, Business Source Complete, Cochrane Library, ERIC, Google Scholar, Ovid Embase, Ovid MEDLINE, PubMed Scopus, and Web of Science Core Collection was conducted from inception through August 18, 2022. Gender was treated as binary given the limitations of the literature. All studies were included that reported the proportion of faculty who were full professors by gender. The primary outcome was adjusted odds ratio for promotion to full professor stratified by area of medicine (Internal Medicine, Surgery, and other). Risk of bias was assessed using the Risk of Bias in Non-randomized Studies of Exposures (ROBINS-E) tool.

RESULTS: Three hundred and twelve studies met inclusion criteria. The unadjusted odds ratio for promotion to full professor was 0.38 (95% CI, 0.35-0.40). Fourteen studies reported adjusted analyses. The overall adjusted odds ratio for promotion of women to full professor was 0.60 (95% CI, 0.45-0.80). The adjusted odds ratio for promotion to full professor in Surgery was 0.55 (95% CI, 0.34-0.88) and 0.89 (95% CI, 0.65-1.21) for studies limited to Internal Medicine. Overall risk of bias was high due to lack of inclusion of important confounders. Statistical heterogeneity was high (I² = 83%, p<0.001). On meta-regression, 84% of the heterogeneity was due to whether the study was from the United States, which reported lower disparity (OR 0.75, 95% CI, 0.61-0.92).

CONCLUSIONS: Most studies continue to find a lower rate of promotion to full professor of women compared to men, even after adjustment for potential mediators such as number of publications, grants, or years on faculty. Gender disparity is particularly notable in Surgery and in studies from outside the United States. Studies limited to Internal Medicine did not find an adjusted difference in promotion. Our results suggest that differences in promotion are due both to differences in mediators of promotion and possible direct bias. Efforts to improve equity need to address both direct bias as well as the disparity in academic productivity.

A WORKSHOP ON SHAME IN MEDICINE AND ITS EFFECT ON BURNOUT AND HELP- SEEKING BEHAVIORS IN INTERNAL MEDICINE INTERNS

Lauren G. Navitsky, Brielle Spataro. Division of General Internal Medicine, UPMC, Pittsburgh, PA. (Control ID #3875803)

BACKGROUND: Burnout prevalence among US physicians continues to increase with the most recent national data suggesting that nearly two-thirds of doctors are experiencing symptoms of this psychological syndrome. Burnout has wide-ranging consequences on physician health, including impairment, stress, mood disorders, substance use disorders, and suicidal ideation, and therefore, efforts to minimize burnout and promote well-being have been prioritized. This has been operationalized in many ways, and one unique method is by examining shame in medicine. Shame, a self-conscious emotion, is thought to lead to diminished psychological and physical wellness in resident physicians. Bynum et al. found that providing training in the understanding and recognition of the self-conscious emotions and creating safe environments for discussing shame experiences might promote well-being among learners. However, this study did not assess for behavioral change. Our study evaluates whether attendance at a one-time workshop on shame in medicine leads to differences in burnout scores and help-seeking behaviors in internal medicine (IM) interns.

METHODS: Based on prior work by Bynum et al., we developed a three-part workshop on shame in medicine. The workshop included a didactic on the self-conscious emotions, a storytelling session during which IM residents and faculty shared their own shame experiences, and faculty-led small group discussions. At a single academic institution, half of the incoming IM interns were randomized to attend the workshop during intern orientation. All interns were invited to complete a survey assessing burnout scores and help- seeking behaviors at orientation and again at the end of the year. A single item was used to measure burnout on a 5-point Likert scale where higher scores indicated more severe burnout. Interns were asked whether or not they had reached out to peers, mentors, family members/friends, and therapists in the last six months. We compared burnout scores using a two-tailed t-test and help-seeking behaviors using chi-squared tests.

RESULTS: Overall, 30/50 (60%) and 37/52 (71%) interns completed the orientation and year-end surveys, respectively. The average burnout scores were 1.9 at orientation and 2.7 at the end of the year. We found no statistical difference in year-end burnout scores between the intervention and control groups (p=0.95). Overall, 31/37 (84%) interns reached out to someone for support in the preceding six months on the year-end survey. We found the acts of reaching out to peers, mentors, family members/friends, and therapists to all be independent of workshop attendance (p=0.55, p=0.48, p=0.97, and p=0.54, respectively).

CONCLUSIONS: While average burnout scores increase over the course of intern year, attendance at a one-time workshop on shame in medicine neither minimizes burnout scores nor promotes help-seeking behaviors in the long-term in IM interns. The impact of a longitudinal curriculum is unknown and remains a potential area of future study.

CURRENT PARENTAL LEAVE POLICIES FOR MEDICAL STUDENTS AND INTERNAL MEDICINE RESIDENTS IN ILLINOIS NEW YORK AND CALIFORNIA A COMPARATIVE STUDY

Stephanie Erickson¹; Savina Sahgal³; Usama Ahmad²; Caspian K. Folmsbee^1,2

¹Internal Medicine, Rush University Rush Medical College, Chicago, IL; ²Internal Medicine, Rush University Medical Center, Chicago, IL; ³Rush Medical College, Rush University Medical Center, Chicago, IL. (Control ID #3868798)

BACKGROUND: Medical training coincides with childbearing years, with the average matriculant into medical school reported as 24 years old. Consequently, medical students and residents must make difficult decisions regarding family planning that may impact health later in life. Programs may use a combination of medical leave, vacation and short-term disability as a form of parental leave. This propagates the archaic belief of pregnancy and child rearing as a pathologic state, dehumanizing trainees and undermining gender equity in the long term. Ideally, parental leave guidelines should adhere to the national guidance from the American Board of Medical Specialties4 and American Board of Internal Medicine; however, this has yet to be well-examined. This study sought to determine the state of parental leave policies in Illinois, New York, and California for medical trainees ranging from medical school to internal medicine (IM) residency.

METHODS: In October 2022, two independent researchers reviewed websites of the Accreditation Council for Graduate Medical Education’s 141 accredited IM residencies and 42 MD- or DO-granting medical schools in Illinois, New York, and California. These states were selected for their high number of medical training programs and geographic diversity. For residency programs, webpages, sample house staff contracts, and handbooks were searched using the following keywords: “pregnant” OR “pregnancy” OR “maternity” OR “paternity” OR “parent” OR “parental OR “family” OR “child” OR “birth.” The same keywords were utilized to search online student handbooks and school webpages. Information regarding medical leave, without inclusion of these keywords was not considered a parental leave policy. Data were analyzed using descriptive statistics.

RESULTS: Forty-two medical school programs and 141 IM residencies were evaluated. Of the medical schools, 57% (n=24) of programs discussed parental leave on their materials. 17% of programs (n=7) have specific parental leave policies not discussed under a pre-existing leave of absence or Title IX policy. Four policies referred to “maternity” only. Of the IM residency programs, 44% (n=62) have publicly available policies regarding parental leave discussed. Only 11% of residency programs (n=16) offer their residents dedicated paid time off for parental leave. Four programs offer dedicated paid time off for parental leave specific for “maternity”.

CONCLUSIONS: Navigating the complexities of family building during medical training requires institutional support. Parental leave policies are only available via public web search at approximately half of medical schools and IM residencies. Medical training programs should improve transparency of policies that clearly communicate the institution’s support and approach to its trainees building families. This could improve gender inequity and trainee well-being, while also serving as a recruiting tool for the institutions.

DEVELOPMENT OF A SHARED MENTAL MODEL FOR SUPPORT OF FACULTY, STAFF, AND TRAINEES WITH FAMILIES IN A DEPARTMENT OF MEDICINE

Whitney Harper, Lia M. Barros, Elizabeth K. Vig, Samantha Caldwell, Ryan Clodfelter, Susan E. Merel

Medicine, University of Washington, Seattle, WA. (Control ID #3865851)

BACKGROUND: Working caregivers in medicine balance the needs of their families, patients, and other responsibilities with limited resources. The context and climate of academic medicine reinforce gender inequities and lead to increased burnout, gender pay inequities, and decreased opportunities for career advancement for working caregivers. Stressors associated with work-life integration are highest in women with children and have been exacerbated by the COVID-19 pandemic. We sought the perspectives of faculty and staff to understand the barriers to supporting employees with families and to explore what an ideal model of support for faculty, staff and trainees with families might look like.

METHODS: We conducted five focus group interviews by videoconference in an academic medical center in the Northwest Region. Facilitators led participants in discussions about perceived barriers to achieving work-life integration, childcare needs, and ways to better support employees with families. Audio recordings were transcribed verbatim and analyzed using conventional content analysis.

RESULTS: 24 employees took part in focus groups. 17% were staff, 79% faculty, and 4% APP; 13% identify as male and 88% as female. See table for domains, themes, and subthemes identified in the qualitative analysis.

CONCLUSIONS: Participants identified barriers at home and work to feeling that they and their family life are supported as employees in the Department of Medicine. Systemic challenges identified within the Department included cultural and structural changes as well as perpetuation of societal inequities. In rich focus group discussions, participants were able to propose a number of solutions including aspects of necessary culture change. We plan to use these results to inform future discussions in our division of GIM as well as in the larger Department of Medicine.

IMPACT OF A NEAR PEER COACHING PROGRAM FOR NEW HOSPITALIST FACULTY AT A LARGE AND GROWING ACADEMIC MEDICAL CENTER

Zuzanna Czernik. Medicine, University of Colorado Denver School of Medicine, Aurora, CO. (Control ID #3873113)

BACKGROUND: Hospital medicine is a rapidly growing field and faces many challenges in academic medicine. One of those challenges is a lack of available senior mentors. The benefits of mentoring to success in academic medicine such as retention of faculty and faculty satisfaction have been shown. Coaching has been widely used in the corporate environment and is starting to be adopted as a tool in academic medicine as well, with early data showing benefits for decreasing emotional exhaustion, increasing well-being and increasing job satisfaction. In order to address the needs of a growing hospital the University of Colorado Division of Hospital Medicine has been hiring more than 10 new faculty members each year over the past 5 years. In an effort to develop systems to support all these new faculty a Near Peer Coaching Program was developed and deployed.

METHODS: This Near Peer Faculty Coaching Program was developed with the aims of: (1) creating professional connections within a large and growing group, (2) providing space for reflection, goals setting and future planning for new faculty and (3) providing structured support for addressing challenges and problems faced by new faculty. The program paired interested new faculty hires (“coachees”) with interested current faculty (“coaches”) for the duration of the new faculty’s first 12 months of employment. Coaches received basic training in coaching principles along with structured materials throughout the year on how to guide coaching sessions. Faculty pairs were prompted via e-mail to meet quarterly for coaching sessions. Participants were surveyed about their experiences upon completion of the year long program.

RESULTS: Between 2019 and 2022, fifty-four faculty members participated in the Near Peer Faculty

Coaching Program. Survey response rate across 3 years of data collection was 72%. Eighty-three percent of coaches and coachees reported that this program helped create a personal connection between faculty. Seventy-eight percent of coaches and coachees reported that this program provided an effective space for reflecting, goal setting, future planning and addressing challenges. Eight-five percent of coachees reported that this program eased their transition into their new role. One hundred percent of coaches and coachees recommended the program for other faculty.

CONCLUSIONS: This Near Peer Faculty Coaching Program has been well received by participants at our institution. Similar models could be used by other groups experiencing growth in order to make connection between their faculty, or by programs hoping to provide structured support to new faculty when there may be a lack of senior mentors.

PARENTAL LEAVE IN HOSPITAL MEDICINE: WE CAN DO BETTER

Angela Keniston³; Natalie Schwatka¹¹; Shradha A. Kulkarni⁷; Gopi Astik⁴; Anne S. Linker⁵; Matthew Sakumoto⁶; Marisha Burden²; Andrew Auerbach⁷; Kirsten Kangelaris⁷; Gregory Bowling⁸; Nila Radhakrishnan⁹; Benji Mathews¹⁰; Areeba Kara¹

¹Department of Medicine, Indiana University School of Medicine, Indianapolis, IN; ²Medicine, University of Colorado Denver School of Medicine, Aurora, CO; ³Division of Hospital Medicine, University of Colorado, Aurora, CO; ⁴Medicine, Northwestern University Feinberg School of Medicine, Chicago, IL; ⁵Medicine, Icahn School of Medicine at Mount Sinai, New York, NY; ⁶Medicine, University of California San Francisco, San Francisco, CA; ⁷Department of Medicine , University of California, San Francisco, San Francisco, CA; ⁸Medicine, The University of Texas Health Science Center at San Antonio, San Antonio, TX; ⁹Medicine, University of Florida, Gainesville, FL; ¹⁰University of Minnesota, Minneapolis, MN; ¹¹Colorado School of Public Health, Aurora, CO. (Control ID #3872130)

BACKGROUND: Parenting as a clinician is challenged by parental leave (PL) policies. Taking PL may negatively affect clinicians’ income, career, and relationships with colleagues. We describe PL experiences in hospital medicine.

METHODS: The Hospital Medicine Reengineering Network national collaborative supports research and improved care. We conducted virtual, semi-structured focus groups through the Network and described the data using rapid qualitative analysis.

RESULTS: There were 21 participants from 15 unique sites, 13 (61%) were women. Four themes emerged.

Preparing

PL was described as evolving towards more inclusion. An emphasis on better planning was noted- by staffing to needs and by facilitating earlier disclosure of pregnancy. Challenges included support for anticipatory budgeting, the complexity of hospitalists’ work affiliations and schedules. Hospitalists described having to self-advocate in the face of opaque policies.

Taking leave

PL was described as ‘disappointing’ with variations across sites. Many described continuing to work in their non-clinical roles during PL juggling parenting and academic expectations. Decreased resiliency by those covering for colleagues on PL was noted.

Returning

Upon return, support systems commonly support lactation including lactation credits and census reductions for lactating women. Childcare was described as increasingly challenging. Returning from PL was felt to disproportionately negatively impact women, with women reducing or altering their roles.

The way forward

As participants envisioned a way forward, the role of how our culture values parenting and shapes expectations emerged. Improving the workday of a hospitalist was identified as a powerful tool to support hospitalists at all stages of their life-cycle.

CONCLUSIONS: PL in hospital medicine poses unique and consequential challenges which disproportionately impact women. Concerted efforts to address our policies, procedures, staffing, culture and expectations that span preparing, taking and returning from leave are needed.

PEER COACHING IN ACADEMIC MEDICINE: BENEFITS AND PROFESSIONAL DEVELOPMENT FOR PEER COACHES

Kenzie A. Cameron¹; Adeboye B. Ogunseitan²; Kiran Nimmagadda³; Cybele Ghossein⁴; Gopi Astik⁵

¹Medicine/General Internal Medicine, Northwestern University Feinberg School of Medicine, Chicago, IL; ²Medicine/Hospital Medicine, Northwestern University Feinberg School of Medicine, Chicago, IL; ³Medicine/Gastroenterology and Hepatology, Northwestern University Feinberg School of Medicine, Chicago, IL; ⁴Medicine/Nephrology and Hypertension, Northwestern University Feinberg School of Medicine, Chicago, IL; ⁵Medicine/Hospital Medicine, Northwestern University Feinberg School of Medicine, Chicago, IL. (Control ID #3872666)

BACKGROUND: We implemented a peer coaching program for first-year faculty in a Department of

Medicine (DOM) and investigated the impact of the program on the coaches.

METHODS: In 2019-2021, we assigned all new DOM faculty a trained peer faculty coach. Coaches logged session details and voluntarily evaluated the program annually on 5-point Likert scales (Strongly Disagree – Strongly Agree), assessing program satisfaction, connectivity and career path, and confidence in coaching skills. Coaches participated in focus groups exploring their experiences and the impact of the program on themselves.

RESULTS: In 2019, 9 faculty were selected as coaches; an additional 5 faculty joined in 2020. All coaches were DOM clinician educator faculty at the rank of Assistant (n=6), Associate (n=1), or Full Professor (n=7); 64% were female. Coaches logged 109 meetings with 45 unique new faculty between September 2019-July 2021; meetings lasted 5-90 minutes (M = 46.8, sd = 18.3). Coaches reported high program satisfaction and perceived it to be a valuable career experience; their confidence in their coaching skills significantly increased from baseline to one year (p<.01), with an additional increase after two years. Although not statistically significant, coach perspectives of connectivity to and support from the institution were high at baseline and increased after 1 and 2 years in the program (Table 1). Anticipated benefits identified via focus groups included (1) opportunity to give back to the medical community, (2) developing useful skills, and (3) assisting others in their career development. Unanticipated benefits included: (1) increased connectivity through participation in a developing coaching community; (2) increased self-awareness, agency, and introspection, and 3) recognition of the applicability of coaching skills to professional and personal domains beyond the scope of the program.

CONCLUSIONS: A peer coaching program has unintended but powerful benefits for faculty coaches including increasing self-awareness and agency, personal growth, and an enhanced sense of community.

PREVALENCE OF WOMEN IN MEDICINE PROGRAMS AT UNIVERSITY-BASED INTERNAL MEDICINE RESIDENCY PROGRAM

Shinji Rho, Alyssa Rust, Koeun Lee, Lydia Zhong, Abby Spencer, Maria Q. Baggstrom, Rakhee Bhayani

Internal Medicine, Washington University in St Louis, St Louis, MO. (Control ID #3834888)

BACKGROUND: Gender disparity remains an unresolved issue in medicine. Women physicians face various forms of inequities during their training process that inhibit them from reaching their full potential. As a response, several academic institutions have established Women in Medicine (WIM) programs as a support system. Our objective was to investigate the prevalence of Women in Medicine (WIM) programs at university-based internal medicine residency programs as of December 2021.

METHODS: Four independent reviewers searched the websites of 145 university-based internal medicine residency programs for evidence of a WIM program. If a webpage for a WIM program was found, the reviewers assessed the site to determine if the program was specific to the department of medicine and if the program was targeted toward trainees. The proportions of programs that had a WIM program, had a Department of Medicine-specific WIM program, and had a trainee-specific WIM program were analyzed. RESULTS: Out of the 145 programs searched, 58 (40%) had a WIM program. Only 16 (11%) were specific to trainees (11 specific to internal medicine trainees and 5 for trainees in all programs). The remaining 42 programs were for faculty and trainees (5 specific to the Department of Medicine and 37 for all departments)

CONCLUSIONS: Few university-affiliated Internal Medicine residency programs have a WIM program specific to trainees. Given the paucity of women leaders at the highest of academic medical centers and evidence that supports early development of leadership skills and support networks, our findings highlight a possible gap in the residency training program infrastructure.

SHARED LEADERSHIP FOR SAFETY & HEALTH AT WORK

Natalie Schwatka^1,7; Angela Keniston²; Marisha Burden³; Gopi Astik⁴; Anne S. Linker⁵; Matthew Sakumoto⁶

¹Environmental and Occupational Health, University of Colorado, Aurora, CO; ²Division of Hospital Medicine, University of Colorado, Aurora, CO; ³Medicine, University of Colorado Denver School of Medicine, Aurora, CO; ⁴Medicine, Northwestern University Feinberg School of Medicine, Chicago, IL; ⁵Medicine, Icahn School of Medicine at Mount Sinai, New York, NY; ⁶Medicine, University of California San Francisco, San Francisco, CA; ⁷Center for Health, Work & Environment, University of Colorado, Aurora, CO. (Control ID #3873820)

BACKGROUND: Effective workplace safety and health (S&H) policies and programs require the engagement of all members of an organization, not just formal leaders. When all have influence, S&H policies and programs become more relevant and consistently implemented. In the hospital setting, shared leadership for S&H can be observed in the daily work of physicians and advanced practice providers (APPs) as they provide clinical care. In this presentation, we will discuss the themes that emerged from a multi- institutional quality improvement effort to understand shared leadership for S&H among hospitalists, including the perspectives of both formal leaders and those without formal leadership roles.

METHODS: We recruited hospitalists and APPs who were members of the Hospital Medicine Reengineering Network (HOMERuN), a collaborative of hospitals, hospitalists, and multidisciplinary care teams founded in 2011. It is a consortium of academic medical centers, with participating sites from the Northeast, Southeast, Midwest, West, and Northwest. We conducted three virtual focus groups with formal hospitalist leaders (n = 12) in Fall 2022 using a standardized focus group guide. We have three additional virtual focus groups planned with hospitalists not in a formal leadership role (n = 12, planned) in January 2023. We will use rapid qualitative methods to analyze the focus groups. The themes and subthemes identified will be used to understand shared leadership in the setting of hospitalist workplace S&H. We will have our final results by the time of the conference.

RESULTS: Preliminary themes from the focus groups with formal leaders include: Violence - The most pressing S&H issue they noted was violence, harassment, and verbal abuse from patients & families.

Advocacy - Everyone should be a leader for S&H, but everyone agreed it starts with formal leadership. They noted that it is hard to be a S&H leader if others around you are not role modeling.

Sharing responsibility - There is a sense that physicians do not need as much support as others, but hospitalist team members step up to share responsibility for each other. A key strategy that hospitalist teams use is communication of boundaries, limits, S&H problems that arise, etc.

Know your partners - You must know who to go to for specific issues that arise and sometimes that is outside of your team, e.g., patient advocacy, office of equity and inclusion, department of public safety. However, they are not always staffed enough to help.

Opportunities for improvement to begin sharing responsibility for S&H amongst hospitalist teams - role modeling, increasing S&H communication, and building safety and health into professional development as well as operations planning.

CONCLUSIONS: Typical leadership-based approaches to leadership development focus on those in a formal leadership role. We will use the findings from this study to inform a shared leadership development approach to advancing the S&H of those in the healthcare industry.

STEREOTYPE PERCEPTION AND BURNOUT IN INTERNAL MEDICINE RESIDENTS Erin L. Dyer¹; Carol Faulk²; Noor Al-Hammadi³; Rakhee Bhayani⁴

¹Internal Medicine, Washington University in St Louis School of Medicine, St Louis, MO; ²Medicine, Washington University in St Louis, St Louis, MO; ³Biostatistics, Washington University in St Louis, Saint Louis, MO; ⁴Medicine, Washington University in St Louis, St Louis, MO. (Control ID #3875211)

BACKGROUND: Burnout is a well-known phenomenon experienced by professionals working in high- stress and demanding fields, including healthcare. It is characterized by physical, mental, and emotional exhaustion that leads to cynicism and dissatisfaction in one’s work. Some research has proposed that women are at higher risk of burnout than men, as the experience of women in medicine may contribute to feelings of emotional exhaustion and reduced sense of personal accomplishment. Salles et al. demonstrated within a group of surgical residents that women were more perceptive of gender stereotypes than men, and women who perceived these stereotypes to a greater degree were more likely to have poorer psychological health. However, this association has not been demonstrated in a population of Internal Medicine residents. Our study will attempt to quantify burnout scores and perception of gender stereotypes within a population of Internal Medicine residents and determine whether there is an association between stereotype perception and burnout.

METHODS: In this prospective cohort study, a survey was distributed annually to the Internal Medicine residents at our academic medical center beginning in 2018. This survey included the Maslach Burnout Inventory (MBI) and questions to assess perceptions of gender stereotypes in the style used by Salles. Results of the MBI included three subscores: Emotional Exhaustion (EE), Depersonalization (DP), and Personal Accomplishment (PA). Participants had “burnout” if they scored in the top tertile for EE or DP or the bottom tertile for PA. Stereotype perception was assessed by asking whether faculty and/or society expect men or women to be better physicians, with responses scored 1-7 on a Likert scale. The Cochran-Armitage test was used to identify any trend in participants’ responses to the stereotype perception questions relative to their burnout scores.

RESULTS: We obtained 130 distinct survey responses over the course of three years. Within this group, 113 residents (86.9%) had burnout. The majority of participants chose the neutral response when asked about the faculty’s perception of male vs female physicians (80.0%), while the majority answered in favor of society expecting men to be better physicians than women (57.7%). There was no statistically significant association between responses to the stereotype perception questions and the presence of burnout, including when participants were were stratified by gender.

CONCLUSIONS: We found a high prevalence of burnout among our Internal Medicine residents, which was unexpected in a residency program that makes resident wellness a priority and has already implemented many initiatives to promote wellness and combat burnout. Our results did not show any statistically significant association between perception of gender stereotypes and presence of burnout. Further investigation is needed to determine whether perception of gender stereotypes has any role in the development of burnout.

THE ASSOCIATION OF WORK OVERLOAD WITH BURNOUT AND INTENT TO LEAVE THE JOB ACROSS THE HEALTHCARE WORKFORCE DURING COVID-19

Lisa Rotenstein¹; Roger Brown²; Christine Sinsky³; Mark Linzer⁴

¹Medicine, Brigham and Women's Hospital, Boston, MA; ²University of Wisconsin System, Madison, WI; ³Professional Satisfaction, American Medical Association, Chicago, IL; ⁴Medicine, Hennepin County Medical Center, Minneapolis, MN. (Control ID #3874347)

BACKGROUND: Burnout has risen across healthcare workers (HCWs) during the pandemic, contributing to workforce turnover. While literature has focused on physicians and nurses, current unprecedented levels of staffing shortages highlight a need to better identify predictors of burnout and work intentions in all HCWs. METHODS: This national, cross-sectional survey study conducted at 206 large healthcare organizations from April to December 2020 was administered to physicians, nurses, other clinical staff (e.g., pharmacists, nursing assistants, respiratory therapists, medical assistants), and non-clinical staff (e.g., housekeeping, administrative, and food staff). The survey assessed burnout via the Mini-Z single-item burnout measure and likelihood of leaving the job within two years. Participants also reported on work experiences, including work overload due to COVID.

We built separate maximum likelihood estimation models with standard errors clustered by organization determining likelihood of burnout and intent to leave (ITL) the job for each role type. Models included whether each respondent met criteria for work overload, and adjusted for respondent demographic characteristics, practice setting, state-level COVID burden at the time of survey completion, and other workplace and COVID-related experiences.

RESULTS: The sample of 43,026 respondents (mean response rate 44%) was comprised of 35.2% physicians, 25.7% nurses, 12.5% other clinical staff, and 25.8% non-clinical staff. Two-thirds (66.2%) of respondents identified as female and about half (47.7%) worked in inpatient settings.

The overall burnout rate was 50% (56% in nursing, 54% in other clinical staff, 47% in physicians and 46% in non-clinical staff; p<0.001 for difference). ITL was reported by 29% of HCWs, with nurses most likely to report ITL (41%), followed by non-clinical staff (33%), other clinical staff, (32%) and physicians (24%) (p<0.001 for difference).

The prevalence of perceived work overload ranged from 37% among physicians to 47% in other clinical staff. In adjusted models, work overload was significantly associated with burnout (OR 3.31 to 4.02) and intent to leave (OR 1.26 to 1.63) across role types.

CONCLUSIONS: High rates of burnout and intent to leave across healthcare roles are significantly correlated with workload. Addressing work overload may help with concerning staffing trends in HCWs. This will require a more granular understanding of sources of work overload across role types, and a commitment to matching work demands to capacity for all HCWs.

THE EMOTIONAL IMPACT OF INTER-HOSPITAL TRANSFERS ON HOSPITALIST PROVIDERS

Claire Westcott²; Lauren McBeth²; Eric Grimm²; Amy Yu¹

¹Medicine, University of Colorado, Denver, CO; ²Medicine/Hospital Medicine, University of Colorado, Aurora, CO. (Control ID #3876490)

BACKGROUND: Patients transferred between hospitals, in a process known as inter-hospital transfer (IHT), are at risk for higher rates of mortality, longer lengths of stay, and higher hospitalization costs compared to patients admitted from the emergency department. While patient outcomes have been studied, the impact of IHTs on providers who care for these patients is understudied. To characterize the effect of the IHT process on hospitalist providers, we examined the experiences of hospital medicine physicians and advanced practice providers (APPs) who care for IHT patients.

METHODS: We conducted a qualitative descriptive study using semi-structured interviews from 09/2021-12/2021 with hospital medicine physicians and APPs from an academic acute care hospital that accepts ~5,000 IHT patients annually.

Emotion categories were identified after reviewing and coding interview transcripts. The R package quanteda was used to split up transcripts into individual words, 2-word phrases, and 3-word phrases that were associated with the emotions. Words with different suffixes that share the same base word (e.g. “frustrated”, “frustrating”, “frustration”) were grouped together in order to count similar words as part of a single concept.

We then counted the number of times words and phrases associated with the identified emotions were used across interviews. Analyses were performed using R version 3.6.3.

RESULTS: We interviewed 30 hospitalists with 1-18 years of experience (mean 5.7 years). 17 (57%) of interviewees were physicians and 13 (43%) were APPs. 31 emotion categories were identified through qualitative coding and emotion words were mentioned over 220 times. Of the emotion categories, 23 were negative emotions, 5 were positive, and 3 were neutral or mixed. Positive emotion words including hopeful, appreciate, happy, grateful, incredible feeling were used 17 times. These emotions were most commonly used in the context of describing communication between clinicians, strategies to obtain clinical information, or patients’ feelings towards being transferred for additional care. Neutral/mixed emotion words such as trust, empathy, and surprise were mentioned 12 times. Negative emotion words were used 196 times. The most common emotion word was "frustration", used 102 times. The next most common was "stressed", used 18 times. Additional negative emotions included annoyed, concerned, upset, burned, and angry. Negative emotions were reported in relation to issues with incomplete or inaccurate information, difficulties with communication, differing expectations, lack of agency, cognitive overload, and patient safety issues. CONCLUSIONS: The IHT process brings up strong negative emotions in providers, which is worrying in a time when burnout is so prevalent among health care providers. Further research should focus on ways to ameliorate IHT process factors that lead to negative emotions and highlight elements of the IHT process that engender positive emotions.

THE FUTURE OF GIM RESEARCH IS BRIGHT, BUT WE HAVE WORK TO DO: RESULTS OF A SURVEY OF CURRENT AND FORMER GIM RESEARCH FELLOWS

Nisa Maruthur¹; Kira Ryskina²; Lauren Block³; Kristina M. Cordasco⁴; Deborah Kwolek⁵; Karin Nelson⁶; E. B. Schwarz⁷; Himali Weerahandi⁸; Donna Windish⁹; Michael Paasche-Orlow¹⁰; Mara A. Schonberg¹¹

¹Medicine, Johns Hopkins University School of Medicine, Baltimore, MD; ²Medicine, University of Pennsylvania, Philadelphia, PA; ³medicine, Donald and Barbara Zucker School of Medicine at Hofstra/Northwell, Lake Success, NY; ⁴Medicine, VA Greater Los Angles Healthcare System/ UCLA, Los Angeles, CA; ⁵Medicine, Harvard Medical School, Boston, MA; ⁶Medicine, University of Washington, Seattle, WA; ⁷Division of General Medicine, University of California, Davis, Sacramento, CA; ⁸Medicine/Population Health, NYU Langone Health, New York, NY; ⁹Internal Medicine, Yale University, Cheshire, CT; ¹⁰Medicine, Tufts Medical Center, Boston, MA; ¹¹Medicine, Beth Israel Deaconess Medical Center, Brookline, MA. (Control ID #3875746)

BACKGROUND: The research pipeline is key to the success of academic general internal medicine (GIM), but little is known about factors that influence investigators to select and continue in this career pathway.

METHODS: A team of 10 stakeholders (researchers, fellowship directors) developed and piloted a web- based REDCap questionnaire assessing: 1) GIM research fellowship experiences; 2) strengths/weaknesses of research training; 3) career paths; and 4) SGIM resource use. Respondents were GIM research fellows (in training programs with >50% of focus on research) between 2012 and 2022 identified from SGIM membership and direct outreach to GIM fellowship program directors. Program directors were identified from SGIM membership lists and government and health system/foundation funded training programs. We compared responses by gender (male/female), underrepresented in medicine (URM) status (yes/no), and SGIM membership (yes/no) using chi squared and Wilcoxon rank sum tests. Respondent open-ended comments were analyzed using thematic analysis.

RESULTS: 42% (160/379) of eligible respondents completed the survey. 77% of respondents were ≤40 years of age; 56% were female; 20% belonged to a URM group; 68% were SGIM members; and 76% had completed their research fellowship. Most reported having strong faculty mentors (96%) and sponsors (89%) and having received high-quality training in research methods (94%) during fellowship. Lowest rated domains of training were hiring/building a research team and negotiation. Experiences with training in grant- writing, diversity/cultural awareness, leadership and support of fellows as caregivers were variable. Most respondents (63%) had pursued/were planning to pursue a clinician investigator position after fellowship. Most respondents had participated in SGIM meetings (83%), but use of most other SGIM resources was low 31% reported discrimination during fellowship. While 87% of respondents reported symptoms of burnout, most felt positively about their careers (87%) and the future of academic GIM research (67%). Survey responses did not vary substantially by gender, URM status or SGIM membership. In open-ended comments, respondents expressed concerns about harmful academic culture, competitiveness and research funding. Recommendations included that SGIM have year-long programs to connect research fellows, better advertise their resources, and work to increase external funding opportunities in GIM research.

CONCLUSIONS: Over half of GIM research fellows pursue clinician investigator positions, but 33% do not. Issues identified included inadequacy of training, the research funding landscape, and discrimination during fellowship. SGIM can better support fellows through year-round career development, networking opportunities and advocacy for funding.

WOMEN FACULTY AND FULL-TIME CLINICAL FACULTY MENTORSHIP NEEDS NOT MET WITH TRADITIONAL MENTORSHIP STRUCTURES

Neha Deshpande³; Helene Starks¹; Paul B. Cornia²; Lauge Sokol-Hessner³; Geetanjali Chander³; Somnath Mookherjee³

¹Bioethics and Humanities, University of Washington School of Medicine, Seattle, WA; ²Medicine, VA Puget Sound HCS & University of Washington SOM, Seattle, WA; ³Medicine, University of Washington, Seattle, WA. (Control ID #3865073)

BACKGROUND: Mentorship is vital for academic medicine faculty career advancement, productivity, job satisfaction, and retention, yet specific mentoring needs and preferences are not well understood, particularly for full-time clinical faculty (FTC). There is a gender gap in advancement and promotion in academic medicine and mentorship is also vulnerable to bias. We surveyed our Division of General Internal Medicine (DGIM) faculty to determine gender and promotion pathway differences in 1) having mentors; 2) satisfaction with mentorship; 3) preferences for finding mentors; and 4) barriers to providing and receiving mentorship.

METHODS: DGIM includes 307 inpatient, outpatient and research faculty across multiple sites. We created a 24-item survey to assess the 4 domains above and examined the distribution of responses by gender and promotion track using chi-square tests. Faculty were invited to respond in October 2022.

RESULTS: 192 (63%) faculty completed the survey. 40 were not queried on satisfaction with mentors: 29 had <4 months in DGIM and we presumed their mentors were at prior institutions; 11 were unsure whether they had a mentor.

Frequency: 107/192 (56%) faculty reported having at least 1 mentor; the proportion of faculty with mentors was similar for women (n=108, 54%) and men (n=79, 59%, p=0.667). There were significant differences by track among who had mentors (40/90, 44% FTC vs 46/58, 79% Clinician-Scholar, and 18/23, 78% Physician- Scientist, p<0.001).

Satisfaction with current mentorship: 65% of faculty with mentors were satisfied. Similar proportions of women (67%) and men (64%) reported satisfaction with mentorship (p=0.895). Among faculty with mentors who were dissatisfied, 18% were CS, compared to 11% FTC and 6% PS (p=0.028). 28/52 (54%) faculty without mentors were dissatisfied with their lack of mentorship; most were FTC (n=17, 65%) or CS (n=9, 35%, p=0.002); all 4 PS with no mentor were satisfied.

Preferences: More women preferred assistance (facilitated connection or mentor assignment) with finding a mentor to support promotion (52% vs 44% for men, p=0.044) and career development (80% vs 62%, p=0.009). More FTC, compared to CS or PS, preferred assistance with finding mentors for promotion (80% vs 66% vs 50%, p=0.001) and career development (77% vs 59% vs 42%, p=0.001).

Barriers: Lack of time was the most prevalent barrier for both receiving (56%) and providing (58%) mentorship.

CONCLUSIONS: While many faculty reported satisfaction with current mentorship, we identified opportunities for improvement. Many faculty without mentors were dissatisfied by the lack of mentorship and FTC tended to lack mentorship. Women preferred facilitated connections with potential mentors. Mentorship programs must provide mentorship opportunities that address these differences and invest in protecting time for mentorship activities.

Scientific Abstract - Chronic Disease Management

20 YEAR TRENDS IN PRIMARY PREVENTION STATIN USE: STUCK IN A RUT SINCE 2013

Casey Kim¹; Micah Eades¹; Kenneth J. Mukamal¹; Jeremy B. Sussman²; Timothy Anderson¹

¹Medicine/General Medicine, Beth Israel Deaconess Medical Center, Boston, MA; ²Internal Medicine, University of Michigan, Ann Arbor, MI. (Control ID #3874079)

BACKGROUND: Statin therapy is a cornerstone of primary prevention of atherosclerotic cardiovascular disease (ASCVD) with historically lower rates of use for patients at increased cardiovascular risk. Little is known about recent trends in statin use for primary prevention of ASCVD and particularly following the 2013 American College of Cardiology (ACC) / American Heart Association (AHA) guideline changes to

pooled risk calculators.

METHODS: We analyzed trends in statin use using the National Health and Nutrition Examination Survey (NHANES) data from 1999-2000 to 2017-2018 in non-pregnant adults without an indication for secondary prevention who completed fasting laboratories. For participants using statins, baseline ASCVD risk was estimated by subtracting the individual statin mean cholesterol lowering effect. Survey-weighted multivariable logistic regression was used to identify predictors of primary prevention statin use including indication (ASCVD risk, diabetes, LDL>190 mg/dL), demographics, and regular source of care from 2015-2018.

RESULTS: Among the 21,691 participants (38.8% age 40-59, 51.8 % female, 67.8 % white) from 1999/2000 to 2017/2018, 31.6% (95% CI, 30.5 - 32.7) met criteria for primary prevention statin use. By indication, 3.6% of participants met criteria of elevated LDL, 10.2% of diabetes diagnosis, 9.9% of ASCVD risk score 7.5-19.9%, and 7.8% of ASCVD >=20%. While primary prevention statin use increased by 22.4% (95% CI, 16.0 – 28.7) from 12.1% (95% CI 8.2 – 16.0) in 1999/2000 to 34.5% (95% CI 29.5 - 39.4) in 2017/2018, growth in statin use plateaued following the 2013 ACC/AHA guidelines (-1.6% change from 2013/2014 to 2017/2018; 95% CI, -8.7 – 5.5). Primary prevention statin use varied greatly by indication and patient characteristics. For example, participants with diabetes were twice as more likely to use statins compared with those with ASCVD > 20% (OR 2.22; 95% CI, 1.33 – 3.70). Factors most strongly associated with not receiving statins were younger age (OR: 0.55 for age 40-59 compared to 60-75 years, 95% CI: 0.02- 0.15) and not having a regular source of care (OR 0.20; 95% CI, 0.07 – 0.63).

CONCLUSIONS: Growth in primary prevention statin use has plateaued since implementation of 2013 ACC/AHA guidelines with two-thirds of patients at increased ASCVD risk not using statins. Strategies to improve primary prevention statin use may include improving access to regular sources of care including primary care, increased attention to middle-age patients, and a population health focus on targeting the highest risk patients, such as those with ASCVD risk over 20%.

ALIGNING ANTI-HYPERTENSIVE PRESCRIBING PRACTICES WITH ACC GUIDELINES TO MITIGATE RACIAL DISPARITIES IN HYPERTENSION CONTROL AMONG BLACK PRIMARY CARE PATIENTS OF RESIDENT PHYSICIANS

Kevin M. Pearlman, Amy E. Wainright, Mim Ari

Internal Medicine, The University of Chicago Medicine, Chicago, IL. (Control ID #3860042)

BACKGROUND: Black patients with hypertension are less likely to have controlled blood pressure, and have increased complications including stroke, heart failure, renal disease, and death compared to non-Black patients. The purpose of this analysis was to assess current hypertension control among Black primary care patients empaneled to resident physicians at the University of Chicago Medical Center (UCMC), with specific analysis of ACC-guideline concordant prescribing practices. Additionally, knowledge and confidence of ACC guidelines was queried to help shape future interventions.

METHODS: Demographic, clinical, and pharmaceutical data was queried from the electronic health records of adult (age 18-84) primary care patients at UCMC with primary hypertension who were empaneled to resident physician PCPs between 3/1/2021 and 9/1/2022 (n =3,485). A voluntary and anonymized survey querying hypertension management practices was sent to all IM residents (n=49, response rate 43%). Questions assessed confidence in managing hypertension and knowledge of ACC guidelines.

RESULTS: Blood pressure was more likely to be uncontrolled (most recent systolic blood pressure reading >140) in Black patients (n=966/3027, 32%) than in non-Black patients (n=105/458, 23%; chi-square p<0.01). Only 9% (n=89/966) of Black patients with a systolic blood pressure (SBP) greater than 140 were prescribed single pill combination (SPC) therapy and 5.6% (n=167/3027) of Black patients were prescribed an ACC guideline-discordant RAS inhibitor as monotherapy.

Despite 86% (n=42/49) of residents feeling very or fairly confident managing a patient with hypertension, only 16% (n=8/49) were very or fairly familiar with the ACC guidelines. About half of respondents chose guideline-discordant options as appropriate monotherapy for a Black patient with stage II hypertension in a clinical vignette (ACE-I=44%, n=15/34; ARB=50%, n=17/34), and 53% (n=26/49) reported race does not affect their choice of anti-hypertensive. Most respondents (63%, n=30/48) indicated they never or rarely prescribe SPC therapy for patients with hypertension with only 45% (n=22/49) feeling very or fairly confident prescribing SPC.

CONCLUSIONS: Uncontrolled hypertension among Black patients is an important national and local healthcare disparity. Our analysis of resident empaneled patients revealed high rates of uncontrolled blood pressure, low rates of SPC therapy and ongoing guideline-discordant monotherapy prescribing. Limited confidence in prescribing SPC therapy and lack of ACC guideline knowledge are potential contributors. Educational interventions are being developed to address these factors and will be followed by additional outcomes evaluation to assess impact.

ALL IN THE FAMILY: BASELINE LIFESTYLE BEHAVIORS AND ADULT/PARENT EFFICACY FOR ACHIEVING LIFESTYLE GOALS AMONG ADULT-CHILD DYADS ENROLLED IN A FAMILY DIABETES PREVENTION PROGRAM PILOT STUDY

Maya S. Venkataramani¹; Nisa Maruthur²

¹Division of General Internal Medicine, Johns Hopkins University School of Medicine, Baltimore, MD; ²Medicine, Johns Hopkins University School of Medicine, Baltimore, MD. (Control ID #3875117)

BACKGROUND: Understanding similarities in adult and child lifestyle behaviors and adults’ perceived efficacy to effect behavioral change can inform the development of multi-generational lifestyle interventions. We examined baseline data from adult-child dyads enrolled in a pilot family-oriented Diabetes Prevention Program (Family DPP) to describe lifestyle behaviors and perceived efficacy for behavior change and explore adult-child correlations in these measures.

METHODS: The Family DPP Pilot Study recruited adults eligible for the DPP from an urban metropolitan area and one child they care for (aged 5-12 years). Adult participants were administered baseline surveys within 2 weeks of intervention start (October 2022). The surveys queried adult and child lifestyle behaviors (physical activity, sleep and dietary intake frequencies). Adults' self-efficacy for achieving lifestyle goals (related to physical activity, sleep, sugary drink and vegetable intake) and parenting-efficacy to support analogous child behaviors were measured on a Likert scale (1: “I know I cannot” to 5: “I know I can”). We performed descriptive statistics and examined correlations between adult and child measures (Fischer’s Exact Test).

RESULTS: Among the 12 adult-child dyads enrolled, mean adult age was 46.3 years (range 34 to 70). 83% of adults were female. Mean child age was 9.3 years. 33% of adults reported achieving their physical activity goal (150 minutes/week) in the past week, while only 18% (2 of 11 children >5 years) achieved child activity goals (60 minutes/day). 25% of adults reported achieving daily sleep goals (7 hours/night), compared to 83% of children (9 to 10 hours/night). Adult and child goal achievement was not correlated. While overall, adults and children had limited intake of fruits and vegetables (only 33% ate vegetables at least once daily, and only 50% of adults and 58% of children ate fruits at this frequency), dietary patterns were not significantly correlated. Adults were least confident about achieving sleep goals (42% endorsed a score of 4 or 5); the majority expressed high confidence in achieving physical activity goals (58%), eating vegetables regularly (83%) and limiting sugary drinks (100%). Adults reported the lowest levels of parenting-efficacy in reducing child sugary drink intake and supporting child physical activity (67% endorsed a score of 4 or 5). Correlation was observed between self-efficacy and parental-efficacy for physical activity (p=0.014).

CONCLUSIONS: At baseline, the majority of adults and children enrolled in a lifestyle intervention do not report meeting lifestyle guidelines, but adults endorse high self- and parenting-efficacy to change certain behaviors. Limited sample size prevented more robust examination of correlation of adult and child behaviors and self- and parenting-efficacy for behavioral change. Characterizing these factors in a larger sample is an important next step that can further inform the content and focus of family-oriented lifestyle interventions.

A PILOT STUDY TO DETERMINE EFFICACY OF BLUETOOTH-ENABLED HOME BLOOD PRESSURE MONITORING IN THE MANAGEMENT OF HYPERTENSION

Cody J. Paiva¹; Brinton Clark²

¹Internal Medicine, Providence Portland Medical Center, Portland, OR; ²Medical Education, Providence Portland Medical Center, Portland, OR. (Control ID #3876800)

BACKGROUND: Hypertension is a common diagnosis encountered in primary care settings, affecting more than 45% of the U.S. in 2017-2018. Hypertension is associated with increased risk of death and disability- adjusted life years. Current treatment strategies for managing hypertension have proven challenging, necessitating novel approaches for better blood pressure control. Prior research shows that the use of Bluetooth-enabled home blood pressure cuffs improves hypertension control. Patients engaged in digital home blood pressure monitoring are more often able to meet blood pressure targets. Our study assesses the use of Bluetooth-enabled home blood pressure monitoring on the management of hypertension in an internal medicine residency clinic.

METHODS: 97 adults with uncontrolled hypertension were identified at the Providence Portland Medical Center Internal Medicine Residency Clinic, defined by blood pressure greater than 150/90 in the 3 months prior to enrollment. Patients were provided with a Bluetooth-enabled blood pressure monitor at enrollment, and blood pressure data was transmitted via smartphone apps to the electronic health record. Home blood pressure readings were transmitted every 2 weeks, and PCPs continued to see patients routinely. A clinical pharmacist conducted consultations which were provided to the PCPs for consideration of medication changes. Blood pressure readings were collected for 3-6 months after enrollment to assess differences in blood pressure pre- and post-use of the home blood pressure cuff. In addition, data about number of medications, number of dose increases and decreases, number of pharmacist visits, and medical co- morbidities were reported.

RESULTS: 95 patients (97.9%) successfully transmitted data at enrollment time and 3-6 months after. 46.4% of patients were women. Average age at the time of enrollment in the study was 62.8 years. 29.9% had diabetes, 26.8% had sleep apnea, 53.6% had ASCVD risk equivalents, and 15.5% had CKD3 or greater. Patients enrolled had a reduction of 20.6 mmHg systolic (P < 0.05) and 8.4 mmHg diastolic (P < 0.05). Patients on average received two consultations by the clinical pharmacist. At the start of the study, patients were prescribed on average 1.77 antihypertensive medications compared to 2.01 medications 3-6 months after enrollment.

CONCLUSIONS: This is the first study in a residency-based clinic to demonstrates that Bluetooth-enabled home blood pressure monitoring has clinical benefit in the reduction of both systolic and diastolic blood pressure for patients with uncontrolled hypertension. By utilizing frequent blood pressure monitoring, clinical pharmacist consultations, and medication intensification we show benefit for optimizing hypertension management in the outpatient setting. Limitations of this study include small sample size and limited follow up. Additionally, we do not have data on adherence to medication regimens. Future directions of this study could include expanding the sample size and longer follow-up time.

A RETROSPECTIVE NARRATIVE REVIEW OF LIFESTYLE RISK FACTORS IN COVID-19. Simmy Lahori¹; Daniel Kashani³; Bright Thilagar²; Abhinav Singla²

¹Internal Medicine, Mayo Foundation for Medical Education and Research, Rochester, MN; ²General Internal Medicine, Mayo Clinic Minnesota, Rochester, MN; ³Hospital Internal Medicine, Mayo Foundation for Medical Education and Research, Jacksonville, FL. (Control ID #3871929)

BACKGROUND: Since the beginning of the COVID-19 pandemic, there has been evolving consensus on how certain lifestyle risk factors could be associated with an increased risk of infection. For eg., there has been an increase in trends with hospitalizations in class III obese population. We were keen to explore evidence that has been generated on how certain risk factors and comorbidities can affect the probability of contracting COVID-19.

METHODS: We included studies focusing on lifestyle risks for COVID-19 identified by the NCBI (National Center for Biotechnology Information) at the US National Library of Medicine (NLM). We searched cited databases systematic reviews/meta-analyses, randomized control trials, review articles, and retrospective cohort studies (Pic 1).

RESULTS: The findings of alcohol as a risk factor were found to be controversial. A study by Suzuki et. al showed a tendency toward an association between decreased mortality among infected and increased mortality in those uninfected. Among the lack of sleep studies, a meta-analysis, which included 149 studies, conducted by Liu et. al demonstrated that sleep disturbance was significantly associated with illness severity. Among the sleep apnea studies, Pataka, A. et al. demonstrated that OSA is a risk factor for COVID-19, while Gimeno-Miguel, A et al. demonstrated by adjusting the rest of the co-morbidities, COPD only remained significant in men and OSA only in women. When it comes to smoking, even though a study by Paleiron et al. showed that smokers had a lower risk of infection, eventually a lot of data flooded into the state otherwise. The Retrospective cohort study by Kompaniyets L et al. demonstrates that the risk of hospitalization, ICU admission, and death was lowest at normal BMIs of ~24 kg.m2 with a sharp increase with higher BMI. The exercise was a winner overall when it came to decreasing risk. In a large observational retrospective study, consistently inactive patients with COVID-19 infection had higher odds of hospitalization, ICU admission, and death.

CONCLUSIONS: There is plenty of evidence that certain predisposing factors, such as high BMI and sleep apnea, increase the risk of COVID-19 infection. Thus far, there is a lack of evidence that correlates COVID-19 risk and lifestyle, such as alcohol consumption, lack of sleep, and smoking. Exercise, overall decreased risk of COVID-19 and its complications.

ASSESSING MEDICATION DISTRIBUTION AND PATIENT NEEDS FOR UNDERSERVED COMMUNITIES AT URBAN STUDENT-RUN FREE CLINIC

Kathy Lu¹; Alex Wind¹; Nili Modi¹; Sara Kazyak¹; Mina Sitto¹; Jennifer Schmidt¹; Adlai Nelson¹; Sara L. Ma¹; Robert Sherwin²

¹School of Medicine, Wayne State University School of Medicine, Detroit, MI;²Emergency Medicine, Detroit Medical Center, Detroit, MI. (Control ID #3876292)

BACKGROUND: Student-run free clinics have been found to contribute to positive patient health outcomes for a variety of chronic conditions, including diabetes, hypertension, and cardiovascular disease. Uninsured or underinsured patients, many of whom facing houselessness or housing instability, face additional obstacles to obtaining quality healthcare, including access to prescription medications. This study examines an urban student-run free clinic in the Midwest offering medical care to un- or underinsured patients through immunization administration, food prescription, routine vision and hearing screenings, and provision of medications. In order to further understand the needs of the patient population and community, we sought to determine the main types of medications that were prescribed in an effort to better understand the chronic conditions impacting clinic patients.

METHODS: Walk-in student-run clinics were held weekly on Saturday mornings in an urban Midwestern city, with patients consisting primarily of underinsured and uninsured residents experiencing houselessness or housing instability. A list of medications prescribed by the attending physician were recorded weekly by student clinic coordinators in a spreadsheet for the duration of the year. Medication type, dosage, and quantity were recorded and data was compiled into graphical form.

RESULTS: A total of 103 different medications were prescribed at 45 clinics held from December 2021 to November 2022. Medications prescribed for the treatment of hypertension made up 45.7% and medication for diabetes management contributed an additional 15.8% of the total medications distributed. Medications for hypercholesterolemia, gastroesophageal reflux disease (GERD), and respiratory conditions were also distributed (6.3%, 4.1%, and 3.9%, respectively). 6.2% of total distributed medications consisted of antiplatelet drugs while 4.2% composed of vitamin and mineral supplements.

CONCLUSIONS: Many of the patients at walk-in free clinics have chronic, comorbid conditions and rely on these free, interdisciplinary services to regularly obtain medications. Gathering data on the most common medications allows us to curate best fitting lists for our underserved patient population when ordering new prescription medications. This is valuable for this clinic and other free clinics with similar patient populations to ensure community needs are adequately met. Future projects plan to identify major barriers that hinder underinsured patients from refilling prescription medications or the duration of time between medication supplies running out and patients refilling them. These projects would provide further practice-based understanding in areas that free clinics can improve their patient care.

ASSOCIATION BETWEEN FREQUENCY OF HOME BLOOD PRESSURE MEASUREMENT AND SUBSEQUENT BLOOD PRESSURE OUTCOMES AMONG PATIENTS ENROLLED IN A REMOTE PATIENT MONITORING PROGRAM FOR HYPERTENSION

Nadia Liyanage-Don^1,2; Brandon K. Bellows²; Kelsey B. Bryant³; Jessica R. Singer²; Siqin Ye²; Nathalie Moise^1,2; John Cheng¹; Maria-Jose Lopez-Sanchez¹; Adina Fraser⁴; Rakhi Kalra⁴; Nadine Dandan⁵; Ian Kronish^1,2

¹Center for Behavioral Cardiovascular Health, Columbia University Irving Medical Center, New York, NY; ²Department of Medicine, Columbia University Irving Medical Center, New York, NY; ³Department of Medicine, Icahn School of Medicine at Mount Sinai, New York, NY; ⁴IT, NewYork- Presbyterian Healthcare System Inc, New York, NY; ⁵Pharmacy, NewYork-Presbyterian Healthcare System Inc, New York, NY. (Control ID #3874751)

BACKGROUND: Remote patient monitoring (RPM) for hypertension (HTN) involves wireless transmission of home blood pressure (BP) data to the electronic health record, which can be used to guide HTN management. Such RPM programs have been shown to significantly improve BP control, especially when combined with nurse or pharmacist support. However, less is known about how the frequency of home BP measurement impacts subsequent BP outcomes.

METHODS: We implemented a HTN RPM program at an adult primary care network in NYC. Clinicians referred patients at their discretion and could choose between programs that offered remote nurse support alone or remote nurse support plus antihypertensive titration by a pharmacist. In both programs, patients were loaned wireless BP devices free of charge and were instructed to measure home BP 2x/day in the first week and at least 3x/week thereafter. Patients were taught how to properly measure their BP and received reminder calls if they had no home BP readings for >1 week. Descriptive statistics were used to quantify home BP trends and frequency of home BP readings during the first 6 months of enrollment. Linear regression was used to estimate the association between frequency of home BP readings and mean home BP by the end of the program.

RESULTS: From June 2019 to August 2021, 443 patients were referred to the program (mean age 67±13 years, 68% female, 24% Black, 62% Hispanic, 60% non-English). Of these, 329 (74%) were actively enrolled, with 221 (67%) in the nurse-only program and 108 (33%) in the nurse + pharmacist program. At baseline, mean home systolic and diastolic BP were 140±18 mmHg and 81±12 mmHg, respectively, improving to 131±15 mmHg and 78±11 mmHg, respectively, after 6 months. Patients measured their home BP an average of 14±7 times per month and 36±45 times total over the 6-month program duration. Increased frequency of home BP measurement was associated with lower systolic (β -0.05, SE 0.01, 95% CI -0.07-0.02, p<0.01) and diastolic (β -0.02, SE 0.01, 95% CI -0.04, -0.01, p<0.01) BP, even after adjusting for patient age, sex, race, ethnicity, language, baseline BP, and program type. Younger age (β 0.10, SE 0.04, 95%

CI 0.01, 0.19, p=0.03) and female sex (β -2.83, SE 1.21, 95% CI -5.22, -0.45, p=0.02) were also associated with lower systolic BP, though not with diastolic BP.

CONCLUSIONS: Increased frequency of home BP measurement was associated with lower systolic and diastolic BP among primary care patients after 6 months of enrollment in a HTN RPM program. This association persisted even after controlling for sociodemographic factors, type of program support received, and baseline BP. Possible explanations include greater patient awareness of BP trends, increased patient adherence to antihypertensive medications, or decreased clinical inertia in antihypertensive intensification. These findings suggest that strategies to promote sustained patient engagement in home BP monitoring are important to maximize the impact of RPM programs on BP outcomes.

CANNABIS FOR FIBROMYALGIA - WHAT IS YOUR SOURCE?

Abhinav Singla¹; Arya Mohabbat²; Jordan K. Rosedahl³; Lindsey M. Philpot³

¹General Internal Medicine, Mayo Clinic Minnesota, Rochester, MN; ²General Internal Medicine, Mayo Clinic, Rochester, MN; ³Community Internal Medicine, Mayo Clinic Minnesota, Rochester, MN. (Control ID #3871746)

BACKGROUND: Fibromyalgia (FM) is a chronic disease with a cardinal manifestation of chronic widespread pain. Patients with FM may experiment with cannabis as an alternative intervention for symptom management. Evidence suggests multiple benefits of using cannabis to help with symptoms, including pain.

METHODS: We conducted a survey (1472 respondents) to gather information regarding cannabis use in FM, including focus on where patients turn for cannabis use information. The survey was constructed using the Symptom Management Theory tool and deployed anonymously via web-based software to patients with a diagnosis of FM.

RESULTS: The overall response rate was around 30%, with the majority being female. We observed significant differences by age between those who reported cannabis use for FM versus those who did not(median age 45 versus 51). Around 50% of the patients used cannabis after receiving a FM diagnosis. Furthermore, 61% of patients never discussed their cannabis use with a healthcare provider (HCP). Picture 1 shows detailed data regarding patients’ primary source of information.

CONCLUSIONS: Cannabis is a popular choice to treat FM symptoms, with 50% of patients utilizing it after diagnosis. Patients mostly seem to get information regarding the benefits of cannabis from FM specialists followed by PCP and rheumatologists. Besides the aforementioned providers, the other sources of information (Figure 1) may have little to no awareness of FM as a disease in general, let alone the targeted benefits of cannabis. Most patients did not discuss with a healthcare provider before using cannabis. Accordingly, the instructions to use cannabis primarily came from vendors/dispensaries. One possible reason is due to the variable legalities associated with cannabis use across states.

CHANGES IN GLYCEMIC CONTROL AND BODY WEIGHT OVER THE COURSE OF THE COVID 19 PANDEMIC IN AN OUTPATIENT SETTING

Keerthana Haridas, Deborah Edelman

Medicine, Icahn School of Medicine at Mount Sinai, New York, NY. (Control ID #3850108)

BACKGROUND: The COVID-19 pandemic has altered health outcomes in populations through a host of downstream social, economic, and psychological changes, especially among those with chronic non- communicable diseases (NCDs). Some studies reveal worsened glycemic control and weight gain, while others indicate improved glycemic control and weight loss. Thus, evidence demonstrates conflicting results in this context. We aimed to conduct a study to explore changes in these metrics in an outpatient setting providing for an underserved population.

METHODS: We conducted a single-site observational study at a Federally Qualified Health Center (FQHC) in New York City to compare glycemic control and body weight, measured by Hemoglobin A1c (HbA1c) and body mass index (BMI) respectively, before and after the onset of the COVID-19 pandemic.

RESULTS: After the pandemic, there was a 103% increase in the annual change in average HbA1c from the years prior to the pandemic versus from early 2020 to 2021 (p < 0.005). Mean BMI increased during the pandemic, although this was not statistically significant. The slope for the change in BMI over five years prior to the pandemic is -0.09, while the slope of change in BMI before and after the onset of COVID-19 is 0.31. The difference between the two slopes is 0.48 (p = 0.37).

CONCLUSIONS: Our study reveals that the COVID-19 pandemic may have led to a worsening in the status of metabolic disorders due to decreased physical activity, worsened dietary habits, psychosocial stressors, and reduced access to healthcare, emphasizing the need for enhanced medical, pharmaceutical and emotional support. Concurrently, many individuals practiced healthier lifestyles via dietary and activity modifications, with an improvement in cardio-metabolic parameters. Given the need for limited data of this type in the United States and conflicting data from other parts of the world, it can be concluded that there is a need to explore changes in healthcare outcomes in patients with non-communicable diseases occurring due to downstream effects of the pandemic. There is indeed a pressing requirement to provide continued support to patients with chronic disorders through telemedicine, access to medications, improved access to laboratories or facilities for the collection of samples from home as well as continued emotional support during a pandemic to prevent acute and chronic complications and to decrease the burden on the healthcare system.

CHARACTERISTICS OF LONG COVID PATIENTS PRESENTING TO A DEDICATED POST- COVID-19 CLINIC

Rija Aziz²; William M. Brode²; Mary A. Kelley³; Nadia Siles Alvarado¹; Esther Melamed¹

¹Neurology, The University of Texas at Austin Dell Medical School, Austin, TX; ²Internal Medicine, The University of Texas at Austin Dell Medical School, Austin, TX; ³Neurology , Dell Seton Medical Center at The University of Texas, Austin, TX. (Control ID #3876646)

BACKGROUND: 10-30% of patients with acute SARS-CoV-2 infection may go on to experience post-acute sequelae SARS-CoV-2 infection (PASC), characterized by broad multisystem symptomatology. Currently, there are few publications on the clinical experience of PASC patients in a dedicated multidisciplinary outpatient clinic. In this study, we characterize clinical symptoms and epidemiological risk factors in patients seeking care for PASC at a tertiary care academic center.

METHODS: We assessed 252 patients treated at the Post-COVID-19 clinic, who presented between June 2021 and May 2022 via a cross-sectional study. Patients completed a standardized clinical questionnaire, Montreal Cognitive Assessment (MoCA) and mental health screenings at their initial visit as part of the standard of care. Data analysis was carried out in R and Microsoft Excel. Statistical significance was assessed with Fisher Exact Test and multiple hypothesis testing was completed with False Discovery Testing.

RESULTS: The median patient age was 44.5 years, 65% were female with a median BMI of 27.4. Patients presented at a median of 10.4 months from acute COVD-19 infection. 78% self-identified their race as white and 21% identified their ethnicity as Latino. Employment breakdown revealed 77% to be employed before March 2020, decreasing to 57% at the time of the initial clinic visit. Most patients (82%) had been vaccinated, and 26% developed PASC following a breakthrough infection. During the acute COVID-19 infection, 50% of patients had a moderate disease severity and 10.5% were hospitalized. The top three co- morbid conditions prior to SARS-CoV-2 infection included mental health conditions, hypertension and asthma. Patients reported a median of 18 new symptoms following COVID-19 illness. The most common symptoms were fatigue (89%), forgetfulness or ‘brain fog’ (89%), deficits in concentration and attention (77%), dizziness or lightheadedness (69%), and post-exertional malaise (68%). PROMIS Global 10 v1.2 assessment demonstrated a mean T score of 37.9 and 37.5 for Mental and Physical categories, with 65% of patients endorsing fair to poor mental health and 73% with fair to poor physical health. MoCA assessments demonstrated that 51% had normal scores and 46% had mild cognitive dysfunction. PHQ-9 scoring revealed 42% had moderate to severe depression, and 38% had moderate to severe anxiety on the GAD-7 assessment.

Symptom burden was similar across gender, age, and initial disease severity.

CONCLUSIONS: Patients presenting to a tertiary Post-COVID-19 program demonstrated numerous multisystem symptoms and functional impairment, independent of the initial COVID-19 disease severity. In addition to increasing research on the underlying pathophysiology and targeted treatments for PASC, further longitudinal studies are needed to define how PASC symptoms change over time. Post-COVID-19 programs also need to be scaled to meet the high patient volume toward delivering complex multidisciplinary care for this patient population.

CHARACTERIZING ONCOLOGY AND PRIMARY CARE INVOLVEMENT IN BREAST CANCER SURVIVORSHIP CARE DELIVERY: RESULTS FROM A POPULATION-BASED SURVEY

Archana Radhakrishnan¹; Allison Furgal²; Rachel Tocco¹; Sarah Hawley²; Ann S. Hamilton³; Kevin Ward⁶; Steven Katz⁴; Lauren P. Wallner⁵

¹General Medicine, University of Michigan, Ann Arbor, MI; ²University of Michigan, Ann Arbor, MI; ³Preventive Medicine, Keck School of Medicine, University of Southern California, Los Angeles, CA; ⁴UNIVERSITY OF MICHIGAN, Ann Arbor, MI; ⁵Internal Medicine, University of Michigan, Ann Arbor, MI; ⁶Emory University, Atlanta, GA. (Control ID #3873774)

BACKGROUND: There are 18 million cancer survivors in the United States (US) and the quality of their survivorship care is suboptimal, especially among vulnerable populations at risk for poor outcomes. While national organizations have called for shared care between primary care providers (PCPs) and oncologists, implementation of these models has been challenging across the diverse care delivery settings in the US. This is due, in part, to our lack of understanding of evolving provider roles during survivorship. We surveyed a large and diverse cohort of breast cancer survivors to characterize provider involvement in the delivery of survivorship care and management of survivorship issues, and explore disparities in provider involvement among vulnerable sub-populations.

METHODS: The iCanCare study is a population-based study of women with early-stage breast cancer diagnosed in 2014-2015 and surveyed 6 years into survivorship (2021-2022) (expected final N=1430, 60% current response rate). Respondents reported on: 1) which provider has been most responsible for providing your survivorship follow-up care (PCP-led, shared-care, oncologist-led); and 2) PCP management of survivorship issues (summary scale of 8-items asking respondents how often they discussed a survivorship issue, such as worry about recurrence and improving physical activity, with their PCP; overall response range from 0-40). We examined patient factors associated with PCP-led delivery and management of survivorship issues using multinomial logistic and linear regression, respectively.

RESULTS: In this preliminary sample, 23% reported PCP-led delivery of their survivorship care, 19% reported shared-care between the oncologist and PCP, and 58% reported oncologist-led delivery. Overall, PCP management of survivorship issues was moderate (mean summary score 19.3, SD 7.0). We did not find statistically significant variation in report of PCP-led delivery of survivorship follow-up care by patient factors. Women who were unemployed (vs. employed) were more likely to report greater PCP management of survivorship issues (mean 22.2 vs. 18.8, p <0.01)) as were patients with 2+ comorbidities (vs. 0 comorbidity) (mean 20.2 vs. 18.0, p<0.01). Results were confirmed in multi-variable analyses.

CONCLUSIONS: We found that in this cohort of breast cancer survivors who are more than five years out from initial treatment, reports of PCP involvement in survivorship was generally low. Notably, some sub- populations of vulnerable patients did report greater PCP management of survivorship issues. Lack of primary care involvement, especially later in survivorship, is a missed opportunity for achieving high-quality survivorship care. As we implement shared-care models as one strategy to improve quality, focusing next steps on engaging PCPs effectively and empowering them to be active participants in survivorship care delivery (e.g., clearly delineating roles and responsibilities, providing educational materials/resources to PCPs) will be critical.

CLINICAL AND GENETIC-BASED REFINEMENT OF THE WARFARIN DOSE

Noor Al-Hammadi^1,2; Kevin B Stephenson¹; Jie Hou³; Anne R. Bass⁴; Charles S Eby¹; Brian F. Gage¹

¹Department of Medicine, Washington University in St Louis School of Medicine, St Louis, MO;²Department of Health and Clinical Outcomes Research, Saint Louis University School of Medicine, Saint Louis, MO; ³Saint Louis University, Saint Louis, MO; ⁴Weill Cornell Medicine, New York, NY. (Control ID #3875225)

BACKGROUND: Warfarin remains a commonly prescribed anticoagulant, use of which is complicated by significant inter-patient variability in the therapeutic dose and a narrow therapeutic index. Algorithms to refine the warfarin dose are needed, especially during days 6-21 of warfarin therapy, when iatrogenic hemorrhages are common. This study is to develop clinical and genotype-guided dosing algorithms for days 6-21 of warfarin therapy.

METHODS: We analyzed data from the Genetic Informatics Trial of Warfarin Therapy to Prevent Deep Venous Thrombosis (GIFT). We randomly split data from the 1471 patients who achieved a therapeutic dose into training (80%) and validation (20%) datasets. We offered clinical, demographic, and laboratory variables into linear regression models. We also offered polymorphisms of the following genes VKORC1, CYP2C9, and CYP4F2 into the genetic models. We used prior research, biologic plausibility, and statistical significance (Akaike information criteria) to select variables. We quantified accuracy using correlation (R²) and mean absolute error (MAE) in the validation dataset.

RESULTS: The clinical model had an R² of 74.0% and MAE of 0.63 mg/day; the genetic model had an R² of 77.4% and MAE of 0.62 mg/day. The strongest predictors of therapeutic dose were the most recent warfarin doses and the current value of the International Normalized Ratio (INR).

CONCLUSIONS: Algorithms using clinical, demographic, and laboratory variables available from the initial 6-21 days of warfarin therapy can predict the therapeutic dose. These dose-refinements may be useful in guiding warfarin dosing.

COMPARATIVE EFFECTIVENESS OF CARE MODELS TO IMPROVE PAIN AND REDUCE OPIOIDS AMONG VA PRIMARY CARE PATIENTS ON LONG-TERM OPIOID THERAPY: A PRAGMATIC RANDOMIZED TRIAL

Erin E. Krebs^1,2; William Becker^3,8; David Nelson^5,2; Allyson Kats⁴; Patrick Hammett^5,2; Beth DeRonne⁵; Sean Nugent⁵; Agnes C. Jensen⁵; Karen H. Seal^7,6

¹Medicine, Minneapolis VA Health Care System, Minneapolis, MN; ²Medicine, University of Minnesota Medical School Twin Cities, Minneapolis, MN;³Internal Medicine, VA Connecticut Healthcare System, West Haven, CT; ⁴Public Health, University of Minnesota Twin Cities, Minneapolis, MN; ⁵CCDOR, Minneapolis VA Health Care System, Minneapolis, MN; ⁶Medicine, University of California San Francisco, SanAnselmo, CA; ⁷Medicine, San Francisco VA Health Care System, San Francisco, CA; ⁸Medicine, Yale University, New Haven, CT. (Control ID#3876626)

BACKGROUND: Patients prescribed long-term opioid therapy (LTOT) are at risk for unrelieved pain, poor quality of life, and serious adverse events. Guidelines recommend multimodal pain care and shared decision- making about opioid dose reduction, but primary care resources are often insufficient. The Veterans’ Pain are Organizational Improvement Comparative Effectiveness (VOICE) trial tested two collaborative care interventions for patients with chronic pain on LTOT. Aim 1 was to compare higher-intensity integrated pain team (IPT) versus lower-intensity pharmacist telecare collaborative management (TCM) for improving pain and reducing opioid dose. We hypothesized IPT would be superior to TCM.

METHODS: VOICE was a 10-site pragmatic randomized trial in VA primary care. Eligible patients had moderate-severe chronic pain and were prescribed opioids ≥20 mg/day. We randomized participants to 1 of 2 active 12-month interventions: 1) IPT, an interdisciplinary team intervention using motivational interviewing and multimodal therapies; or 2) TCM, a clinical pharmacist care management intervention using symptom monitoring and medication optimization. Both interventions provided individualized pain care and opioid tapering support. Masked assessors collected outcomes by telephone at 3, 6, 9, and 12 months. The primary outcome was pain response, defined as ≥30% decrease in Brief Pain Inventory (BPI) score (range 0-10, higher=worse) at 12 months. The main secondary outcome was ≥50% reduction in opioid daily dose. Primary analyses were intent-to-treat and used repeated measures logistic regression for comparisons of response rates.

RESULTS: We enrolled 821 patients and randomized 820. Participants’ mean age was 62.1 years (SD 10.6) and 13.5% were female. At baseline, the mean BPI was 6.7 (SD 1.5) in IPT and 6.6 (SD 1.6) in TCM. The baseline opioid daily dose was 80 mg (SD 73) in IPT and 75 mg (SD 57) in TCM. The primary outcome was assessed for 721 participants (88%) at 12 months. At 12 months, the mean BPI score was 6.1 (SD 1.8) in IPT and 6.0 (SD 1.9) in TCM. The pain response rate was 17% in IPT vs. 15% in TCM (p=0.61) and the model- estimated odds ratio of pain response for IPT vs. TCM was 1.11 (95% CI 0.74, 1.66). At 12 months, the mean opioid daily dose was 54 mg (SD 65) in IPT and 53 mg (SD 52) in TCM. The proportion with 50% dose reduction was 25% in IPT vs. 25% TCM (p=0.85) and the odds of dose reduction for IPT vs. TCM was 1.03 (95% CI 0.75, 1.42). Findings from 4 prespecified subgroup analyses also showed no difference between interventions.

CONCLUSIONS: Pain and opioid dose improvements were similar for two pain care delivery models augmenting primary care management of patients with chronic pain and LTOT. Local resources and preferences should drive decisions about which care model to implement.

COMPARATIVE EFFECTIVENESS TRIAL OF INTERDISCIPLINARY APPROACHES TO PAIN CARE: IMPACT ON CHANGE IN USE OF NON-DRUG THERAPIES IN VA PATIENTS WITH CHRONIC PAIN PRESCRIBED LONG-TERM OPIOID THERAPY

Karen H. Seal^1,2; Patrick Hammett³; William Becker^4,9; Elizabeth S. Goldsmith^3,5; David Nelson³; Allyson Kats³; Agnes C. Jensen⁶; Beth DeRonne⁷; Erin E. Krebs^8,5

¹Medicine, University of California San Francisco, SanAnselmo, CA; ²Integrative Health, San Francisco VA Health Care System, San Francisco, CA; ³Center for Care Delivery and Outcomes Research, Minneapolis VA Health Care System, Minneapolis, MN; ⁴Internal Medicine, VA Connecticut Healthcare System, West Haven, CT; ⁵Medicine, University of Minnesota Health Sciences, Minneapolis, MN; ⁶Center for Chronic Disease Outcomes Research (CCDOR), Minneapolis VA Health Care System, Minneapolis, MN; ⁷CCDOR, Minneapolis VA Health Care System, Minneapolis, MN; ⁸Medicine, Minneapolis VA Health Care System, Minneapolis, MN; ⁹Medicine, Yale School of Medicine, New Haven, CT. (Control ID #3875304)

BACKGROUND: Non-drug therapies (NDTs) are recommended as first-line treatment for chronic pain. Evidence is limited for approaches to increase uptake of NDTs for pain. The Veterans’ Pain Care Organizational Improvement Comparative Effectiveness (VOICE) trial compared two pain care models in veterans with chronic pain prescribed long-term opioid therapy (LTOT). This secondary analysis compared the effect of two pain care approaches on change in VOICE participants’ use of NDTs.

METHODS: From October 2017 to March 2021, 821 patients with chronic pain prescribed LTOT were enrolled in VOICE, a pragmatic trial at 10 VA sites across the U.S. Participants were randomly assigned for

12 months to either the Integrated Pain Team (IPT), an interdisciplinary intervention involving at least 3 clinical visits with medical and allied health providers (e.g., psychologist, physical therapist) and 8 motivational interviewing calls, or to Telecare Collaborative Management (TCM), consisting of at least 6 visits with a clinical pharmacist care manager supported by a consulting physician. Outcome assessors used the Pain Management Inventory to assess frequency of past-year use of NDTs at baseline, 6-, and 12-months. NDTs were categorized as manual therapies (acupuncture, chiropractic, and massage), exercise (yoga, Tai Chi, aerobic, stretching/strength training), and behavioral (relaxation, meditation, psychological). Thresholds for clinically meaningful use of NDTs were defined by expert (investigator and clinician) consensus. Multivariable repeated measures logistic regressions modelled the effect of intervention on change in meaningful use of ≥ 1 NDT at 6- and 12-months compared to baseline.

RESULTS: Of 820 randomized participants, 411 were assigned to IPT and 409 to TCM. Mean age was 62.1 +/- 10.6, 13.5% were female, and 24% identified as non-White race. Most participants in IPT (74%) and TCM (75%) reported meaningful use of ≥ 1 NDT at baseline. There were no significant between-group differences in change in meaningful use of ≥ 1 NDT at 6 and 12 months. In IPT, there were significant increases in meaningful use of ≥ 1 NDT at 6 months (82%, OR=1.58 1.16, 2.17) and 12 months (83%, OR= 1.64, 95% CI: 1.20, 2.23) compared to baseline. In TCM, there was a non-significant increase in meaningful use of ≥ 1 NDT at 6 months (77%, OR=1.09, 95% CI: 0.82, 1.44), and a significant increase at 12 months (85%, OR= 1.84, 95% CI: 1.31, 2.60) compared to baseline. In the overall sample, there were increases in meaningful use of ≥ 1 NDT(s) from baseline to 6- and 12-months; the largest gains were in meditation, relaxation, and strength training.

CONCLUSIONS: Meaningful use of NDTs increased in both IPT and TCM treatment groups over 12 months.

COMPARING GUIDELINES FOR STATIN AND ANTIHYPERTENSIVE TREATMENT AMONG U.S. VETERANS: A MICROSIMULATION STUDY

David Flood³; David Ratz²; Wyndy L. Wiitala²; TImothy Hofer^1,2; Rodney A. Hayward^1,2; Jeremy B. Sussman^1,2

¹Internal Medicine, University of Michigan, Ann Arbor, MI; ²VA Ann Arbor Healthcare System, Ann Arbor, MI; ³Internal Medicine, U. Michigan, Ann Arbor, MI. (Control ID #3874423)

BACKGROUND: While clinical guidelines for the use of anti-hypertensive medications and statins focus on a patient’s risk of developing cardiovascular disease (CVD) and specific risk factor treatment targets, they do not consider individualized estimates of treatments’ benefits and harms, called benefit-based tailored treatment (BTT). Furthermore, they address each risk factor in isolation, missing the chance to identify which treatment should be prioritized for a given patient. In this project, we examined the advantages of a BTT approach compared to current approaches.

METHODS: We conducted a microsimulation to compare the risks and benefits for U.S. veterans of different cholesterol and blood pressure treatment approaches for the primary and secondary prevention of CVD. Our model used all active primary care patients ages 40-79 years in the VA health care system beginning in 2011. We modeled each patient’s CVD health under three common guidelines (American College of Cardiology/American Heart Association, Veterans Affairs, and American College of Physicians/US Preventive Services Task Force) versus a BTT approach in which treatment decisions are guided by an individual’s predicted risks and benefits. The probability of key outcomes was estimated from personalized prediction models, then modified by the treatment recommended by each guideline. We then estimated the impact of each treatment approach compared to the status quo for outcomes including CVD events, deaths, and quality-adjusted life-years (QALYs).

RESULTS: The population included 3,682,371 individuals. Their median age was 65 years (IQR: (56 to 70), 91% were men, and 75% were white. At baseline, 18% had a history of CVD and 34% had a diabetes diagnosis. The BTT approach could prevent more CVD events and deaths and save more QALYs than the other guidelines. It would use less medicines than the ACC/AHA or VA guidelines but more than the ACP/USPSTF guidelines. For example, the BTT approach would prevent 29.2 CVD events per 1000 people over 5 years compared to 23.8 using the ACC/AHA guidelines, while using an average of 182 medications per 1000 people compared to. 214 for ACC/AHA. Findings were consistent across subgroups stratified by CVD history, age, sex, or race.

CONCLUSIONS: Among U.S. veterans, a benefit-based treatment approach, compared to ACC/AHA guidelines, could lead to major improvements in CVD outcomes while significantly decreasing overall medication use.

COVID-19 PANDEMIC DISRUPTIONS AND CHRONIC PAIN AMONG US MILITARY VETERANS ON LONG-TERM OPIOID THERAPY

Elizabeth S. Goldsmith^1,2; Patrick Hammett¹; Brian Martinson¹; Siamak Noorbaloochi^1,2; Barbara Clothier¹; Erin E. Krebs^3,1

¹Center for Care Delivery and Outcomes Research, Minneapolis VA Health Care System, Minneapolis, MN; ²Medicine, University of Minnesota Twin Cities, Minneapolis, MN; ³Medicine, Minneapolis VA Health Care System, Minneapolis, MN. (Control ID #3876114)

BACKGROUND: The COVID-19 pandemic may influence chronic pain outcomes through pandemic effects on systems and communities as well as through COVID-19 illness. We evaluated associations of COVID-19 illness and pandemic disruptions with pain-related functional interference within a cohort of US military veterans prescribed long-term opioid therapy for chronic pain.

METHODS: The Effects of Prescription Opioid Changes in Veterans (EPOCH) study is a national cohort that enrolled 9082 veterans in 2016. Participants have completed yearly surveys including Brief Pain Inventory Interference scale (BPI-I: 0-10 score, higher = worse; minimum clinically important change 1 point). Year 4 surveys (7/2020 – 2/2021) included NIH-VA-DOD Pain Management Collaboratory COVID-19 questions assessing pandemic disruptions. We compared pandemic indicators by prespecified demographic subgroups using chi-square tests, and estimated binomial logistic regressions with BPI-I change (worsened vs. stable/improved) as the dependent variable and pandemic experience indicators as independent variables, applying design weights and inverse probability weights to adjust for confounding.

RESULTS: Within the final analytic sample (N=3833), 75% (n=2876) of year 4 surveys were returned between July and November 2020, prior to the late 2020 US COVID-19 peak. Mean BPI-I was 6.3 (SD 2.2) at both year 3 and year 4. 5% of respondents reported COVID-19 illness before survey return. 59% reported the pandemic reduced their ability to get healthcare, with higher prevalence among veterans under 65 years old (66%) than 65+ (56%, p<0.001). 53% reported the pandemic worsened their mental and emotional health, with higher prevalence under 65 (60%) vs 65+ (50%, p<0.001) and among urban (56%) vs rural (49%) residents (p<0.001). Self-reported prevalence of all pandemic disruptions was higher among veterans under 65 years old than among veterans 65 and older. Self-reported COVID-19 illness did not significantly associate with BPI-I change between year 3 and year 4 (OR 0.90, 95% CI 0.53-1.54). Self-reported pandemic-related worsening of ability to get healthcare was associated with higher odds of worsening BPI-I vs. remaining stable or improving (OR 1.32, 95% CI 1.04–1.69). Self-reported pandemic-related worsening of mental/emotional health was associated with higher odds of worsening BPI-I (OR 1.72, 95% CI 1.34–2.21). Other pandemic disruptions were also associated with worsening BPI-I (worsened finances OR 1.37, 95% CI 1.07–1.74; worsened ability to meet basic needs OR 1.38, 95% CI 1.08–1.77)

CONCLUSIONS: In the first COVID-19 pandemic year, most veterans with chronic pain reported pandemic-related disruptions to healthcare access or mental/emotional health. Disruptions, but not reported COVID-19 illness, were significantly associated with worse pain outcomes. Health care access and support for mental/emotional health should be maintained during health emergencies.

COVID PANDEMIC EFFECTS ON GLYCEMIC CONTROL AND OTHER OUTCOMES IN PATIENTS WITH TYPE 2 DIABETES AT AN ACADEMIC MEDICAL CENTER

Lawrence Ma, Joseph Diaz, Jennifer DeConde, Melissa A. Wong, Bernice Ruo

Internal Medicine, University of California San Diego, La Jolla, CA. (Control ID #3867955)

BACKGROUND: Few studies within the United States have assessed the effect of the pandemic on the quality of chronic disease management for patients with type 2 diabetes (T2DM). We hypothesized that diabetes management would be adversely impacted, especially in individuals who are older, non-white, or living in zip codes with lower household incomes.

METHODS: This retrospective study compares diabetes outcomes pre-COVID (3/1/2019 – 2/29/2020) to during COVID (3/1/2020 – 2/28/2021) at an urban academic internal medicine practice. Adult patients with T2DM and at least one outcome measure in each time period were included. Patients with end-stage renal disease were excluded. Demographics including age, gender, race, ethnicity, insurance type, and zip codes were collected. Outcome measures included: hemoglobin A1c (Hgb A1c), systolic blood pressure (SBP), diastolic blood pressure (DBP), low-density lipoprotein level (LDL), and body mass index (BMI). T-test and ANOVA were used to determine statistical significance (p-value < 0.05).

RESULTS: 998 total patients were included, 52% were female. Mean age was 65 ± 11 years with 19% of individuals over the age of 75. 57% identified as White, 20% as Asian, 7% as Black, and 16% as other. 12% identified as Hispanic. 50% of the study population lived in zip codes with an estimated annual household income > $80,000. All patients were insured: 35% had commercial insurance, 51% had Medicare, 10% had Medicaid, and 4% had other forms of insurance. See table for diabetes-related outcome measures. There were no significant differences in any of the outcome measures during the two time points including when stratified by income, race, gender, ethnicity, age, or insurance type.

CONCLUSIONS: In this cohort of patients with T2DM, the COVID pandemic did not have a significant clinical or statistical impact on Hgb A1c, BMI, SBP, DBP, or LDL. Reasons for this may include: a cohort with well-controlled diabetes at baseline who had established primary care prior to the pandemic and the rapid implementation of a telehealth program. Limitations include a small sample size at a single academic center that consisted of a relatively healthy and affluent patient population. Further studies are needed to better understand the broader impact that the COVID pandemic has had on chronic disease management, and which populations may be more vulnerable to worsening health outcomes during a pandemic.

CUMULATIVE DOSE OF REPEATED CORTICOSTEROID INJECTIONS AND EMERGENCY DEPARTMENT VISITS/HOSPITALIZATIONS

Terin Sytsma¹; Shannon Thomas²; Laura Greenlund³

¹Internal Medicine, Mayo Clinic, Rochester, MN; ²University of Minnesota Twin Cities, Minneapolis, MN; ³Internal Medicine, Mayo Clinic, Rochester, MN. (Control ID #3869778)

BACKGROUND: Corticosteroid injections (CSI) are a common treatment for numerous causes of pain, particularly in medically complex patients who are poor surgical candidates. Despite decades of use, there is a paucity of literature regarding the systemic effects of cumulative doses of intraarticular/bursa CSIs. Systemically administered corticosteroids are known to cause immunosuppression, adrenal suppression, and significant hyperglycemia, all of which can result in emergency department (ED) visits and potentially hospitalization. We evaluated whether the cumulative dose of injected corticosteroids was associated with an increased number of ED visits/hospitalizations.

METHODS: Utilizing an institutional database, adult patients at our institution receiving joint, bursa and spinal CSI were followed from May 2018 to July 2022. Demographics, cumulative CSI dose, and number of ED visits/hospitalizations were collected. Based on cumulative triamcinolone equivalents received, patients were divided into “High”(≥ 600 mg) , “Medium”(160-599 mg), and “Low”(<160 mg) dose cohorts.

Descriptive statistics summarized the demographics. Logistic regression models ascertained odds of ED visits/hospitalization in relation to cumulative CSI dose.

RESULTS: 8817 patients were included: 247 in the High-dose, 2506 in the Medium-dose, and 6063 in the Low-dose cohort. Mean age was 68 years-old, and 61.9% of patients were female. 43,156 CSI occurred over the study period. The mean total CSI dose was 847 mg of triamcinolone equivalents (range 600-2140 mg) in the High-dose group compared to 276 mg (range 160-599 mg) and 70 mg (range 2.7-159 mg) in the Medium- and Low-dose groups respectively. The total number or ED visits/hospitalizations was 19,121 over the 4-year study period. Logistic regression modeling showed that patients receiving high and medium doses of CSI were more likely to have at least one ED visit/hospitalization during the study period with an odds ratio of 1.28 (CI 1.16-1.41, p<0.001) and 1.56 (CI 1.17-2.08, p=0,002) in the Medium-dose and High-dose groups respectively compared to the Low-dose cohort.

CONCLUSIONS: Patients receiving larger cumulative doses of injected corticosteroids were more likely to have at least one ED visit/hospitalization during the study period. This could represent systemic effects of injected corticosteroids and/or the complex healthcare needs of patients receiving CSI. Further work needs to be done to better isolate the effect of cumulative CSI dose on adverse health outcomes.

DETERMINING RISK FACTORS OF PHYSICAL INACTIVITY AMONG ADULTS WITH CHRONIC CONDITIONS SINCE THE START OF THE COVID-19 PANDEMIC

Nicholas Demetriou¹; Morgan Bonham²; Julia Yoshino Benavente²; Stephanie Batio³; Stacy Bailey²

¹Medicine, Northwestern University Feinberg School of Medicine, Chicago, IL; ²General Internal Medicine, Northwestern University, Chicago, IL; ³Health Literacy and Learning Program, Northwestern University, Chicago, IL. (Control ID #3876474)

BACKGROUND: Physical inactivity results in greater than 5 million global premature deaths annually. Physical inactivity is closely linked with chronic conditions such as cardiovascular disease (CVD) and type II diabetes and is essential to their management. The Covid-19 pandemic has significantly disrupted physical activity levels, with a particularly stark decrease among individuals with medical conditions. Understanding how this trend has progressed along with the pandemic is vital in equipping clinicians and public health officials with the latest knowledge to address physical inactivity. The goal of this study is first to determine how activity levels have trended since the start of the Covid-19 Pandemic. Second, we aim to identify adults with the highest risk of persistent inactivity.

METHODS: A cohort of 673 adults with 1+ chronic conditions from across Chicago enrolled in our cohort study in March 2020. Interviews were conducted with this cohort every four months for 28 months. The Behavioral Risk Factor Surveillance System Questionnaire, Section 10, assessed participants' physical activity levels to determine if the participants met the CDC’s adult physical activity guidelines targets: an aerobic target of 150 minutes of moderate-intensity physical activity per week and a strength target of two days of muscle-strengthening activity per week. Trajectory mixture models were used to identify clusters of participants following the same trajectories of physical activity. Multivariate logistic regression models were then applied to identify subgroups likely to be classified in different physical activity trajectories, adjusting for potential confounders.

RESULTS: Participants were on average 62 years old, with 60% females, 29% living under poverty, and 67% with 3+ chronic health conditions. Physical activity trajectory analysis for the two years following the onset of the Covid-19 pandemic identified two distinct groups; individuals who persistently met physical activity targets and individuals that did not. Within our study population, (45%) consistently met the aerobic target, (29%) consistently met the strength training target, and (21%) consistently met both. Subgroups less likely to meet physical aerobic, strength and overall physical activity targets included those identifying as Latinx (Aerobic: aOR: 0.45, 95% CI: 0.26,0.79; Strength: aOR: 0.46, 95% CI: 0.25,0.85; Overall: aOR: 0.35, Overall: aOR: 0.37, 95% CI:0.20,0.68).

CONCLUSIONS: In conclusion, our study has identified race as a key predictor of whether a person with chronic health conditions meets recommended physical activity targets. Other confounders, such as geography, specific disease states, and family and work commitments, likely confound the correlation between race and physical activity. Our next steps are to analyze other factors that may account for the correlations thus far identified.

DO RACIAL AND ETHNIC DISPARITIES IN SGLT2 OR GLP1 PRESCRIPTIONS DIFFER BY PAYER: A NATIONALLY REPRESENTATIVE STUDY

Lurit Bepo^2,1; Oanh K. Nguyen²; Anil Makam²

¹National Clinician Scholars Program at UCSF, San Francisco, CA; ²Internal Medicine, University of California San Francisco School of Medicine, San Francisco, CA. (Control ID #3877316)

BACKGROUND: Sodium-glucose cotransporter-2 (SGLT2) inhibitors and glucagon-like peptide-1 (GLP1) receptor agonists lower atherosclerotic cardiovascular disease (ASCVD) risk, morbidity, and mortality. Among persons with type 2 diabetes mellitus (T2DM), there is differential access to these medications by race and ethnicity. However, the extent to which insurance type and its interaction with race and ethnicity is associated with decreased SGLT2 and GLP1 use is unknown. This cross-sectional study examines the association between insurance type, race and ethnicity, and SGLT2 and GLP1 prescription rates in adults with T2DM.

METHODS: Design: Our study used nationally representative survey data of the U.S. civilian noninstitutionalized population from the 2018 and 2019 Medical Expenditure Panel Survey (MEPS). We used descriptive statistics to assess prevalence of SGLT2 and GLP1 prescriptions, sociodemographic factors, and clinical characteristics. We used sequential multivariate logistic regression to analyze the interaction between insurance type and race and ethnicity: 1) unadjusted, 2) adjusted for cardiovascular risk factors, and 3) adjusted for socioeconomic (SES) and health care access factors.

Participants: Adults ≥18 years old with T2DM.

Exposures: Primary insurance type defined as >6 months of private insurance, Medicare, or Medicaid. Race and ethnicity defined as Non-Hispanic White or Non-White and Hispanic.

Outcome: Rate of any SGLT2 or GLP1 prescription.

RESULTS: There were 4997 adults with T2DM in the 2018-2019 MEPS surveys who were 50% female with mean age 64 years. 34% had private insurance, 57% had Medicare, and 10% had Medicaid. Overall, 13% had any SGLT2 or GLP1 prescription (6% SGLT2, 9% GLP1). Individuals with private insurance had the highest prescription rates (17%), followed by Medicare (11%), and then Medicaid (10%).

In our unadjusted logistic regression, Non-Hispanic White individuals with private insurance and Medicare had statistically significant higher rates of any SGLT2 or GLP1 prescription than Hispanic and non-White persons (21% vs 10%, p=0.000 and 12% vs 9%, p=0.034, respectively). This disparity persisted in private insurance after adjusting for clinical (15% vs 7%, p=0.000) and SES/access factors (14% vs 7%, p=0.000) but was no longer statistically significant for Medicare. There were not statistically significant racial and ethnic disparities in Medicaid.

CONCLUSIONS: Overall SGLT2 and GLP1 prescription rates are low for adults with T2DM and differ by insurance type and race and ethnicity. Although privately insured individuals had the highest prescription rates, they also had the largest racial and ethnic disparities. Disparities in Medicare were attenuated by adjustment, suggesting that these findings may be due to differences in cost-sharing, prior authorization, and other formulary restrictions. Interventions targeting formulary policies may be a helpful first step to reduce inequities among privately insured individuals with T2DM.

ECONOMIC FACTORS ARE A MAJOR DRIVER OF MEDICATION NON-ADHERENCE FOR NON-COMMUNICABLE DISEASES IN RURAL KENYA

Pooja Gala¹; Jemima Kamano²; Manuel Vazquez-Sanchez³; Richard Mugo²; Vitalis Orango²; Sonak Pastakia⁵; Carol Horowitz⁴; Joseph Hogan⁶; Rajesh Vedanthan¹

¹Department of Medicine and Population Health, New York University Grossman School of Medicine, New York, NY; ²Medicine, Moi University, Eldoret, Uasin Gishu County, Kenya; ³Population Health, New York University Grossman School of Medicine, New York, NY; ⁴General Internal Medicine, Icahn School of Medicine at Mount Sinai, New York, NY; ⁵Purdue University, West Lafayette, IN; ⁶Biostatistics, Brown University, Providence, RI. (Control ID #3874431)

BACKGROUND: Poor medication adherence in low-and middle-income countries (LMICs) is a major cause of suboptimal hypertension (HTN) and diabetes (DM) control. We aimed to identify key factors that contribute to suboptimal medication adherence in western Kenya.

METHODS: We conducted a cross-sectional analysis of baseline data of participants enrolled in the BIGPIC study in western Kenya. All participants were age ≥ 35 with either DM or HTN who had been prescribed medications in the past three months. Baseline data included socio-demographic characteristics, economic status, and medication adherence. Predictors of medication adherence were separated into the five WHO dimensions of medication adherence: conditions-related factors, patient-related factors, therapy-related factors, economic-related factors (monthly income, cost of transportation, monthly cost of medications), and health system-related factors. A multivariable analysis, controlling for age and sex, was conducted to identify drivers of suboptimal medication adherence. Estimated predicted values of medication adherence at different combinations of transportation and medication costs were calculated, generated from the multivariable regression model.

RESULTS: The analysis included 1,496 participants (73.7% women) with a mean age of 60 years (range 35-97). Suboptimal medication adherence was reported by 71.2% of participants. Economic factors were associated with medication adherence. In a multivariable analysis that investigated specific costs, transportation costs were found to be associated with worse medication adherence (transportation costs Kenyan Shillings (KSH) 1-99 beta-coefficient 0.7 (95% CI 0.4, 1.1); transportation costs KSH 100+ beta- coefficient 0.5 (95% CI 0.1, 0.9)). In contrast, we found no association between medication costs and medication adherence. Figure 1 demonstrates that transportation costs were associated with higher (worse) estimated predicted adherence scores, while medication costs were not associated with adherence scores.

CONCLUSIONS: Suboptimal medication adherence is highly prevalent in Kenya, and primary associated factors include costs, particularly indirect costs of transportation. Addressing all economic factors associated with medication adherence will be important to improve outcomes for non-communicable diseases.

EFFECT OF OFFERING BUPRENORPHINE TO IMPROVE PAIN AND REDUCE OPIOIDS AMONG PATIENTS ON HIGH-DOSE LONG-TERM OPIOID THERAPY: A PRAGMATIC RANDOMIZED TRIAL

William Becker^1,2; Karen H. Seal³; Beth DeRonne⁷; Joseph Frank⁹; Agnes C. Jensen⁶; Allyson Kats⁵; David Nelson⁷; Patrick Hammett⁸; Erin E. Krebs⁴

¹VA Pain Research Informatics Multi-morbidities and Education Center, West Haven, CT; ²Yale School of Medicine, New Haven, CT; ³Medicine, University of California San Francisco, SanAnselmo, CA; ⁴Medicine, Minneapolis VA Health Care System, Minneapolis, MN; ⁵University of Minnesota School of Public Health, Minneapolis, MN; ⁶Center for Chronic Disease Outcomes Research (CCDOR), Minneapolis VA Health Care System, Minneapolis, MN; ⁷CCDOR, Minneapolis VA Health Care System, Minneapolis, MN; ⁸Research, Minneapolis VA Health Care System, Minneapolis, MN;⁹University of Colorado, Denver, CO. (Control ID #3875757)

BACKGROUND: Opioid tapering is challenging for patients and providers. Observational studies show promise for switching to the partial opioid agonist buprenorphine as an alternative to tapering. Buprenorphine (BUP) has similar analgesic properties yet is safer than other opioids. The Veterans’ Pain Care Organizational Improvement Comparative Effectiveness (VOICE) trial tested interventions to improve pain and reduce opioid use among VA primary care patients with chronic pain on long-term opioid therapy (LTOT). We examined whether offering the option of BUP, compared with no BUP option, would lead to improved pain and greater opioid dose reduction.

METHODS: VOICE was a 12-month, 10-site, randomized trial that enrolled VA primary care patients with moderate to severe pain despite LTOT. Participants were randomized to 1 of 2 pain care models; both models involved individualized opioid tapering recommendations. Participants prescribed high-dose opioids (≥70 mg/day) were additionally randomized to receive/not receive the option to discuss BUP and, if deemed appropriate, to switch to BUP. Masked assessors collected outcomes at 3, 6, 9, and 12 months. The primary outcome was change from baseline to 12 months in the Brief Pain Inventory (BPI: score range 0-10, higher=worse). The main secondary outcome was opioid dose in milligram morphine equivalents (MME). For MME calculations, BUP was considered to be 0. Primary analyses were intent-to-treat and used repeated measures logistic regression for comparisons of primary outcomes.

RESULTS: We randomized 207 to BUP option (n=104) or no BUP option (n=103). Mean age was 60.9 years (SD 10.2), 10.6% were female, and 13.9% were Black. At baseline, the mean BPI was 6.8 (SD 1.5) for the BUP arm and 6.8 (SD 1.6) for the no BUP arm. The baseline opioid daily dose was 157 mg (SD 75) for BUP and 165 mg (SD 88) for no BUP. Of 104 BUP option participants, 28 consented to switch to BUP. Across the 10 study sites, the proportion of patients assigned to the BUP arm who switched to BUP was 0-64%. At 12 months, the mean BPI score was 6.1 (SD 1.9) in the BUP arm and 6.3 (SD 1.7) in the no BUP arm; mean BPI change did not differ between groups (.62 for BUP vs .49 for no BUP, p=0.67). At 12 months, the mean opioid daily dose was 92 mg (SD 98) in the BUP arm and 105 mg (SD 89) in the no BUP arm; mean opioid dose reduction did not differ between treatment groups (65 mg for BUP vs. 61 mg for no BUP, p=0.55). CONCLUSIONS: Adding the option to switch from full-agonist opioids to BUP did not improve pain severity nor opioid dose reduction compared with individualized tapering alone. Minimal uptake of BUP at some sites limited ability to detect an effect. Qualitative analyses will explore patients’ experiences with the BUP option. To maximize patient-centeredness, prescribers may consider offering switch to BUP as an alternative to standard opioid tapering.

ENROLLMENT IN A CARE MANAGEMENT INTERVENTION AND CONTROL OF BLOOD CHOLESTEROL LEVELS FOR INDIVIDUALS WITH SERIOUS MENTAL ILLNESS

Karly A. Murphy¹; Elizabeth Sarker^3,2; Elizabeth A. Stuart³; Courtney Cook⁴; Stacy Goldsholl⁴; Gail L. Daumit⁴

¹Medicine, University of California San Francisco, San Francisco, CA; ²Biostatistics, Johns Hopkins University Bloomberg School of Public Health, Baltimore, MD; ³Mental Health, Johns Hopkins University Bloomberg School of Public Health, Baltimore, MD; ⁴Division of General Internal Medicine, Johns Hopkins University School of Medicine, Baltimore, MD. (Control ID #3866776)

BACKGROUND: Cardiovascular disease (CVD) is the leading cause of death for people with serious mental illness (SMI), contributing to 10-20 year premature mortality. Despite this risk, people with SMI often do not receive guideline-concordant care processes. We examined whether a care management intervention improved control of blood cholesterol for adults with SMI and identified factors associated with improved care quality.

METHODS: We used data from an 18-month comprehensive CVD risk reduction clinical trial for adults with SMI that was shown to reduce overall CVD risk. Trial participants had at least one CVD risk factor.

Intervention participants met with a health coach bi-weekly and a nurse for lifestyle counseling, medication adherence counseling, and care management, tailored to their CVD risk profile. This analysis examined participants recommended for cholesterol management: dyslipidemia, aged 40-75 years with diabetes, and/or with a 10-year ASCVD risk of >=7.5%. Our primary outcome was change in LDL cholesterol (LDL-C). We used t-tests and chi-square tests to describe differences in patient characteristics, intervention components, and changes in mean LDL-C levels; we also stratified by whether participants received an increase in cholesterol medication as a marker of active clinical care.

RESULTS: Of the 216 participants, 81% (N=174) of participants had dyslipidemia, 43% (N=92) had diabetes, 51% (N=110) had an elevated ASCVD risk score, and 62% (N=134) were not prescribed any cholesterol medications. During the study, 26% (N=27) of intervention participants and 16% (N=18) of control participants received cholesterol medication changes (p=0.1). Intervention participants were more likely to receive medication changes if they were older, female, Black, had a higher number of CVD risk factors, elevated ASCVD risk score, or diabetes at baseline compared with presence of depression or elevated ASCVD risk score in the control arm. Across the study, mean LDL-C levels declined more for the intervention group (-11.7 mg/dl, SD=29.9) than for the control (-3.0 mg/dl, SD=32.4) (p=0.046). In stratified analysis, LDL-C levels declined more for intervention participants who received medication changes (-24.0 mg/dl, SD=38.8) compared with those who did not (-7.2 mg/dl, SD=24.8) (p=0.01). In the control group, LDL-C levels did not differ when stratified by medication changes (-4.4 mg/dl [SD=41.7] vs -2.7 mg/dl [SD=30.9]) (p=0.8). Intervention participants with medication changes had more visits with their PCP when accompanied by the study nurse (3.3 visits [SD=2.9] vs. 1.2 visits [SD=1.4]) (p<0.001).

CONCLUSIONS: Mean LDL-C levels declined in participants who received a comprehensive CVD risk reduction intervention but not for participants who did not. Given the long-term risks of elevated cholesterol, interventions that sustain reductions in LDL-C are of great interest. Future population health interventions are needed to address cholesterol management for people with SMI.

EPIDEMIOLOGY, TRENDS AND UTILIZATION DISPARITIES OF CATHETER ABLATION AND ITS ASSOCIATION WITH CORONARY VASOSPASM AMONGST PATIENTS WITH NON- VALVULAR ATRIAL FIBRILLATION- A NATIONWIDE STUDY

Siva Naga S. Yarrarapu¹; Parth Shah²; Aditya Desai⁷; Andrea Mestre³; Adedolapo A. Ottun⁶; Shiv Shah⁴; Beshoy Iskander⁵; Abdelhakim Ouled Said⁹; Deepika Vunnam⁸; Vikramaditya Samala Venkata¹⁰; Renu Bhandari¹¹; Sahithi Kantamneni¹²; Anmol Bharti¹³; Ya-Ching Hsieh¹⁴; Urvish K. Patel¹⁵

¹Internal Medicine, Monmouth Medical Center, Long Branch, NJ; ²Internal Medicine, Reading Hospital, West Reading, PA; ³Universidad del Rosario, Bogota, Cundinamarca, Colombia; ⁴Government Medical College Surat, Surat, Gujarat, India; ⁵Case Western Reserve University, Cleveland, OH; ⁶Agogo Presbyterian Hospital, Agogo, Ashanti, Ghana; ⁷Internal Medicine, University of California Riverside, Riverside, CA; ⁸Dr Pinnamaneni Siddhartha Institute of Medical Sciences & Research Foundation, Vijayawada, Andhra Pradesh, India; ⁹IBN-SINA UNIVERSITY HOSPITAL, Rabat, Morocco; ¹⁰Cheshire Medical Center and Dartmouth- Hitchco*ck Keene, Keene, NH; ¹¹Manipal College of Medical Sciences, Pokhara, Nepal; ¹²Alluri Sitarama Raju Academy of Medical Sciences, Eluru, Andhra Pradesh, India; ¹³University College of Medical Sciences, Delhi, Delhi, India; ¹⁴Public Health, Icahn School of Medicine at Mount Sinai, New York, NY;¹⁵Public Health and Neurology, Icahn School of Medicine at Mount Sinai, Jersey City, NJ. (Control ID #3871811)

BACKGROUND: Non-valvular atrial fibrillation(NVAF), the most common supraventricular tachyarrhythmia, poses severe economic burden with high mortality rates. Catheter ablation(CA) is an important curative treatment for NVAF. Nationwide data on utilization and outcomes of CA is limited. Additionally, coronary vasospasm is a rare, life-threatening peri-operative complication of CA with limited literature in Caucasians.

METHODS: A retrospective cross-sectional cohort study was designed from NIS 2007-2017. Hospitalizations with NVAF treated by CA were identified using ICD 9 & 10 codes. CA induced vasospasm was identified. Univariate analysis was performed using chi square and unpaired t test to evaluate socio- demographic and geographic disparities. Cochran Armitage trend test was used to evaluate utilization trend. Multivariate survey logistic regression to identify predictors of coronary vasospasm. We calculated adjusted OR, 95%CI keeping p<0.001 as significant.

RESULTS: Out of 35,906,946 primary & secondary diagnosis of NVAF, 343641(0.96%) patients underwent CA. Utilization trend of CA declined from 1% in 2007 to 0.73% in 2017(p<0.0001). Patients aged 50-75yrs(aOR 1.6,95%CI:1.52-1.66), Native American race(1.39,1.24-1.55), CHF(1.06,1.05-1.08), higher annual income(1.15,1.12-1.18), private insurance(1.3,1.28-1.33) had higher odds of utilization. The Mid-West region (0.71, 0.69-0.73) fared lower than the South, the West and the North-East.(all p values < 0.0001). Prevalence of coronary vasospasm was higher among CA vs without CA(0.03% vs 0.02%; p<0.0001), however, non-significant association seen between CA & vasospasm(aOR1.29,0.83-2.03;p<0.2599). Predictors of vasospasm were male(2.04, 1.80-2.31), non-elective procedure(0.84, 0.70-0.99), substance abuse(3.36, 2.68-4.21), and hypertension(1.25, 1.09–1.45)(all p<0.0001).

CONCLUSIONS: The overall trend for catheter ablation utilization was found to be decreasing from 2007 to 2017. However, the outcomes of catheter ablation such as odds of death, discharge to non-home, severe disability, morbidity, and risk of death were found to have statistically improved outcomes compared to no ablation group. The Mid-West region had lower odds of utilization while the people of Native American race had higher odds compared to remainder of US population. Identification of factors associated with lower utilization of CA and its disparities will help to mitigate the burden associated with NVAF.

EVALUATING THE FINANCIAL SPENDING OF THE 2022 FISCAL YEAR AT AN URBAN STUDENT-RUN FREE CLINIC

Mina Sitto¹; Jennifer Schmidt¹; Adlai Nelson¹; Sara L. Ma¹; Kathy Lu¹; Alex Wind¹; Nili Modi¹; Sara Kazyak¹; Robert Sherwin²

¹School of Medicine, Wayne State University School of Medicine, Detroit, MI; ²Emergency Medicine, Detroit Medical Center, Detroit, MI. (Control ID #3876289)

BACKGROUND: Medical student-run free clinics (SRFC) provide the resources to support healthy outcomes for patients experiencing financial hardship. This study examines a SRFC in an urban Midwestern community that provides longitudinal healthcare to the underserved and uninsured populations. The clinic treats chronic conditions like diabetes, hypertension, and other comorbidities through regular visits, pharmaceutical interventions, and medical tests and screenings. This requires the clinic to take a prudent approach with the distribution of their funds in order to best serve the community and their needs. Few studies focus on the financial spending and budgeting aspects of operating a SRFC. The goal of this study is to analyze the yearly spending and the distribution of funds to efficiently operate as a SRFC in the community.

METHODS: Walk-in student-run clinics were held every Saturday morning in an urban Midwestern city. A list of supplies was recorded weekly within a spreadsheet by two student clinic coordinators. The costs and quantities were also recorded and divided into ten categories: Over-the-Counter Medications, Medical Tests, Electronic Medical Record System, Clinic Security, Prescription Medications, Food/Drinks, Office Supplies, Medical Supplies, Hygienic Items, Immunization Supplies, and Non-Profit Status and Website Subscriptions. The costs for each category was summed and the data was compared using percentages.

RESULTS: A total of $34,186.50 was accrued in the December 2021 to November 2022 fiscal year. Preliminary results show that 48.6% of the clinic’s funds were spent on prescription medications, followed by 10.5% on medical tests and screenings, 8.8% on immunization supplies, 8.4% on maintaining the electronic medical record, and 8.1% on securing the clinic.

CONCLUSIONS: This study found that the majority of the clinic’s funds were allocated to prescription medications, and medical tests and screenings such as lipid panels, hemoglobin A1C, blood glucose, and rapid COVID-19 tests. A likely explanation for this distribution is based on patient compliance and consistent monitoring through follow-up appointments for the management of chronic disease. These results can shape future budgeting and estimate spending costs for our clinic, as well as for other free health clinics. Future projects could focus on evaluating the fiscal spending for the past five years, using the trends to forecast future costs. Determining the estimated costs can assist the clinic in making informed decisions on their spending and grant submissions.

EVALUATION OF THE ROLE OF MILD COGNITIVE IMPAIRMENT AND FRAILTY IN 30-DAY READMISSIONS DURING ADMISSION FOR AN ACUTE EXACERBATION OF COPD

Juhi C. Gupta¹; Nicole Kappel²; Leah J. Witt³; Alex Spacht⁷; Megan Huisingh-Scheetz²; Steven R. White⁴; Vineet M. Arora⁵; Valerie G. Press⁶

¹Pritzker School of Medicine, University of Chicago Pritzker School of Medicine, Chicago, IL; ²General Medicine, University of Chicago Pritzker School of Medicine, Chicago, IL;³Geriatrics, University of California, San Francisco, CA; ⁴Medicine, University of Chicago, Chicago, IL; ⁵Medicine, University of Chicago Medical Center, Chicago, IL; ⁶Medicine, University of Chicago, Chicago, IL; ⁷Brigham and Women's Hospital Department of Medicine, Boston, MA. (Control ID #3876226)

BACKGROUND: Chronic Obstructive Pulmonary Disease (COPD) is the third leading cause of readmissions among patients aged 65 years or older in the US. Sixty-eight percent of patients with COPD have at least one comorbidity, and 30% admitted for acute exacerbations of COPD (AECOPD) have more than four comorbidities. Frailty, a syndrome of limited reserve to withstand/recover from stress (e.g., hospitalizations), affects many older patients with AECOPD. Our prior work showed that weak grip strength, a frailty marker, was associated with a 30-day COPD readmission risk. There is published evidence of an association between mild cognitive impairment (MCI) and frailty. Additionally, there is a greater prevalence of MCI in patients with COPD (versus normative data). MCI has been associated with poor outcomes post- AECOPD and difficulty with self-management (e.g., poor inhaler technique). To address this, we examined associations between impaired cognitive function in admission and all-cause 30-day readmissions among frail and pre/non-frail patients with COPD.

METHODS: Between July 2016 – December 2022 we prospectively identified adult inpatients admitted to an urban, academic hospital with AECOPD who were ≥ 18 years. Participants completed surveys and underwent physical assessments at one timepoint during their admission to assess demographics, frailty status (frail/pre-frail/non-frail) using Frieds Frailty Criteria (score= 0 non-frail, 1-2 pre-frail, and 3-5 frail), and MCI using the Montreal Cognitive Assessment (MoCA) (MoCA score <26, version 7.1 or 8.1). Descriptive statistics, chi-squared tests, and t-tests were used to analyze the data using STATA 15.1.

RESULTS: Enrolled participants (n=78) had a mean age of 63.7±9.4 years; 93% self-identified as black and 53% as female. Most of our population had MCI (84%). There were no significant differences in mean MoCA scores between participants who were frail versus pre/non-frail (19.4 ± 4.7 vs. 20.9 ± 4.7, p=0.2) or MCI (64% vs 77%, p=0.2). All-cause 30-day readmissions were not significantly different between frail and non-frail/pre-frail groups (15% vs. 13%, p=0.8). Similarly, there were no significant differences in 30-day readmissions comparing MCI vs. no MCI, though effect size for 30-day readmissions was nearly 2 times greater for the MCI group (MCI 15% vs. no MCI 8%, p=0.5).

CONCLUSIONS: Our data suggest that MCI is prevalent among patients admitted with AECOPD. Further, patients who demonstrated MCI during their inpatient evaluation had twice the rate of 30-day readmissions. MCI thus may be a potential risk factor for readmission among this population. Our future work will continue to evaluate the role of MCI and frailty in post-discharge functional trajectories regarding readmissions amongst older adults with COPD to determine their roles in risk-stratification and the potential for tailored interventions to reduce early readmissions.

EXAMINING THE INTERPLAY BETWEEN RACE/ETHNICITY, PATIENT-PHYSICIAN COMMUNICATION AND CANCER-RELATED POST-TRAUMATIC STRESS IN BREAST CANCER SURVIVORS WITH COMORBID DIABETES

Karisma Pantaleon¹; Christina Wang³; Lauren Guerra¹; Mita S. Goel²; Yael T. Harris⁴; Jenny J. Lin¹

¹Internal Medicine, Icahn School of Medicine at Mount Sinai, New York, NY; ²Northwestern University Feinberg School of Medicine, Chicago, IL; ³Gastroenterology, Icahn School of Medicine at Mount Sinai, New York, NY; ⁴Donald and Barbara Zucker School of Medicine at Hofstra/Northwell, Hempstead, NY. (Control ID #3874378)

BACKGROUND: Breast cancer is the most common cause of cancer in women in the United States. Although curable, breast cancer survivors (BCS) are at risk of developing significant psychological distress associated with their cancer. BCS are also more likely to experience comorbid diabetes mellitus (DM), which may worsen distress. Prior studies suggest that minoritized BCS commonly experience cancer-related post- traumatic stress (PTS), a condition that is associated with reduced quality of life. Effective patient-physician communication is associated with improved psychological well-being, higher quality of life, and may prolong survival. We aim to evaluate the relationship between race/ethnicity, patient-physician communication, and cancer-related distress in BCS with comorbid DM.

METHODS: BCS with DM were surveyed using the Impact of Events Scale-Revised (IES-R) to assess for cancer-related PTS and the Patient Reactions Assessment (PRA) to assess perceptions of patient-provider communications with their cancer (PRA-C) and DM providers (PRA-D). BCS with IES-R score >24 were characterized as having probable cancer-related PTS. Race/ethnicity, PRA scores, and IES-R scores were compared using Pearson’s correlation for continuous variables and the Kruskal-Wallis test for those with and without PTS.

RESULTS: A total of 162 female BCS with DM [mean (SD) age: 66.6 (7.1) years] who self-identified as White (40.1%), Black (31.5%), or Hispanic/Other (28.4%) were included. Non-White groups reported worse communication with their cancer providers [median PRA-C (IQR): White: 51 (14) vs. Black: 46 (9) vs. Hispanic/Other 45 (5), p=0.003] but not with their DM providers [median PRA-D (IQR): White: 48 (12) vs. Black: 46 (9) vs. Hispanic/Other: 46 (5), p=0.22]. PRA-C and IES-R scores were significantly correlated, such that improved communications with cancer providers (r=-0.23, p=0.004), but not DM providers (r=-0.13, p=0.08), were associated with lower cancer-related PTS. BCS with cancer-related PTS had lower PRA-C scores [median (IQR): 45(3) vs. 48(12), p=0.04] but no difference in PRA-D scores [median (IQR): 46(10) vs. 47(10), p=0.22] compared to those without cancer-related PTS. In subgroup analyses by race and ethnicity, the correlation between cancer-related PTS and PRA-C scores was strongest in Hispanic/Other BCS (r=-0.39, p=0.009) and weakest in Black (r=-0.12, p=0.42) and White BCS (r=-0.09, p=0.49). No significant correlation between cancer-related PTS and PRA-D scores were observed by race or ethnicity.

CONCLUSIONS: Minoritized BCS with comorbid DM report less satisfaction in communication with their cancer providers which may, in turn, result in higher levels of cancer-related PTS. Future investigations will evaluate whether differences in patient-provider communications mediate the association between race/ethnicity and cancer-related PTS.

EXPERIENCES TRANSITIONING FROM FULL-AGONIST OPIOIDS TO BUPRENORPHINE FOR PAIN: A QUALITATIVE STUDY

Sara Edmond³; Jennifer L. Snow²; Danielle Wesolowicz¹; Amanda Jankelovits³; Sophia Currie⁴; Manik Chhabra⁵; William Becker^2,3

¹VA Connecticut Healthcare System, US Department of Veterans Affairs, Washington, DC; ²PRIME, VA Connecticut Healthcare System, West Haven, CT; ³Yale School of Medicine, New Haven, CT; ⁴Yale University, New Haven, CT; ⁵VA Medical Center Corporal Michael J Crescenz, Philadelphia, PA. (Control ID #3874305)

BACKGROUND: Guidelines recommend tapering or discontinuation of long-term opioid therapy (LTOT) for chronic pain when harms of continued opioid therapy outweigh benefits, yet tapering is often challenging and may be associated with adverse events. An emerging alternative to tapering or discontinuing LTOT is switching to the partial agonist buprenorphine. The goal of this study was to understand the experiences of patients with chronic pain who transitioned from LTOT to buprenorphine.

METHODS: We conducted semi-structured qualitative interviews with 19 patients from two VA medical centers; eligible patients had previously been on LTOT and switched to buprenorphine (any formulation) for at least 30 days during the past three years. The interview guide included questions about perceptions of and experiences with buprenorphine. We performed team-based inductive thematic analysis including two independent coders and full research team meetings to form consensus on coding concepts and findings.

RESULTS: Participants primarily had chronic back pain and included 15 men and 4 women ranging in age from 38 to 81. Analyses identified five major themes; here, we focus on three. First, while there were diverse experiences, participants generally reported satisfaction with buprenorphine as a pain treatment that improved overall function. Most viewed buprenorphine as either equally effective to LTOT, or less effective for pain but superior overall given improved quality of life in part due to decreased side effects leading to clear headedness and feeling more even keeled. Next, a patient’s preconceptions about buprenorphine were often associated with subsequent reported experiences. One participant noted, “As long as you go into it with the right mindset… it works.” Finally, patient-provider communication, including a nonjudgmental approach and transparency about risks and benefits of buprenorphine, was identified as essential; one participant said: “That’s something I’ve always appreciated, there was no judgment.”

CONCLUSIONS: Veterans switching from LTOT to buprenorphine had mostly positive yet mixed experiences. While perceived analgesia varied, Veterans identified benefits beyond pain reduction as important to them. This highlights the need for research in this context that assesses a wide array of patient- reported outcomes beyond analgesia. Given the importance of preconceptions about buprenorphine and patient-provider communication, motivational interviewing targeting willingness to try buprenorphine along with patient education may contribute to improved patient outcomes

FASTING TO REDUCE CHEMOTHERAPY SIDE EFFECTS: AN UPDATED SYSTEMATIC REVIEW

Isabela Azevedo Ferreira de Souza¹; Cintia Gomes²; Caroliny Hellen Azevedo da Silva³; Isabele Ayumi Miyawaki⁴; Isabela Reis Marques⁵; Vittoria Caporal Salles Moreira⁶; José Eduardo Riceto Loyola Júnior⁷; Eduardo M. Padrao⁸; Anand Bhagat⁸

¹Faculdade de Medicina de Petrópolis, Petrópolis, Rio de Janeiro, Brazil; ²Maimonides Medical Center, Brooklyn, NY; ³Universidade Federal do Rio Grande do Norte, Natal, RN, Brazil; ⁴Universidade Federal do Parana, Curitiba, PR, Brazil; ⁵International University of Catalonia, Barcelona, Spain; ⁶University Israelita de Ciências da Saúde Albert Einstein, São Paulo, Brazil; ⁷Hospital Alemão Oswaldo Cruz, São Paulo, Brazil; ⁸Internal Medicine, University of Connecticut School of Medicine, Farmington, CT. (Control ID #3871757)

BACKGROUND: Chemotherapy (CTx) for the treatment of cancer may cause significant toxicity to patients and impair quality of life (QoL). Research in mice demonstrated that fasting minimizes CTx side effects and increases tumor sensitivity. Some clinical studies revealed that fasting might improve QoL in patients receiving CTx, although this remains controversial.

METHODS: We aimed to perform an updated systematic review of randomized controlled trials (RCTs) and non-randomized studies of intervention (NRSI) exploring the effects of fasting versus normocaloric diet in patients with cancer undergoing CTx. Pubmed, Embase, and Cochrane library were searched for studies published up to June 2022. The risk of bias was assessed by RoB-2 for RCTs and ROBINS-I for NRSI. The primary outcomes of interest were toxicity and QoL. The study was registered at PROSPERO CRD42022337752.

RESULTS: We found 4465 citations, of which 5 studies (4 RCTs) met inclusion criteria, with a total of 290 patients, 146 (50.3%) of whom received CTx with fasting. The population included gynecological malignancies with an average age of 49 to 57 years. Many types of fasting were used including fasting- mimicking diets, modified short-term fasting, short-term, and water only fasting. Time of fasting varied between 1 to 3 days before and after CTx. All studies found fasting to be safe and well-tolerated without significant side effects. Only two studies found that fasting improved patients’ tolerability to CTx with reduced symptoms such as stomatitis, headaches, weakness, and fatigue. One study found that fasting had beneficial effects only on hematologic toxicity and on DNA damage of healthy cells, such as lymphocytes and myeloid cells, with no benefits in non-hematological toxicity; however, another study verified that radiological and pathological tumor responses to CTx could potentially be enhanced by fasting. Four studies evaluated QoL with conflicting results. Two studies found statistically significant improvement in QoL. Two further studies yielded negligible improvement in QoL; however, one verified fewer CTx dose reductions and delays due to intolerable side effects.

CONCLUSIONS: In this systematic review of 5 studies and 290 patients, fasting at the time of CTx did not improve overall tolerability to CTx, however can reduce some significant side effects. It can potentially decrease hematological side effects and enhance the response of the tumor to CTx.

GUIDELINE-INDICATED PRESCRIBING OF SGLT2 INHIBITORS AND GLP1 RECEPTOR AGONISTS FOR TYPE 2 DIABETES AMONG PRIMARY CARE AND SPECIALTY CARE PROVIDERS

Michael E. Bowen^1,2; Kelsea Marble¹; Christine Mai¹; Jonathan Pak³; Mujeeb Basit¹; Ildiko Lingvay¹

¹The University of Texas Southwestern Medical Center, Dallas, TX; ²Parkland Health, Dallas, TX; ³Clinical Development & Medical Affairs , Boehringer Ingelheim Pharmaceuticals, Inc., Ridgefield , CT. (Control ID #3868156)

BACKGROUND: Although Sodium-Glucose Cotransporter-2 inhibitors (SGLT2i) and Glucagon-like Peptide-1 receptor agonists (GLP1RA) improve glycemic control, cardiovascular outcomes, and mortality, significant gaps in guideline-indicated prescribing remain. SGLT2i and GLP1RA medications can be prescribed by various specialties; however, it is unknown if seeing multiple specialties increases prescribing.

METHODS: Using EHR data from a large, academic medical center, we conducted a cross-sectional analysis of patients with type 2 diabetes and cardiorenal disease eligible for American Diabetes Association 2022 guideline-indicated treatment with SGLT2i and GLP1RA to determine their currently prescribed rates. Eligible patients were ≥18 years and had at least one completed outpatient encounter in either primary care, endocrine, cardiology, or nephrology from January 1, 2019 to October 30, 2022. All patients had type 2 diabetes and one or more of the following: atherosclerotic cardiovascular disease, heart failure, or chronic kidney disease. We describe current prescribing rates overall and by specialties (primary care, endocrine, cardiology, nephrology) who commonly manage these conditions.

RESULTS: A total of 10,819 patients (mean age of 67 years, 47% female, 41% with ASCVD, 25% with CHF, 47.9% with CKD stage 3-5) were eligible for guideline-indicated treatment with a SGLT2i or GLP1RA. Overall, 17.9% of eligible patients had an active prescription for at least one of these agents. The percent of patients (N=10,819) seen by the respective specialty and currently prescribed guideline-directed therapy were as follows: 55% and 20% in primary care, 48% and 20% in cardiology, 44% and 25% in endocrinology, and 18% and 18% nephrology. The increase in prescribing among patients seen vs. not seen within each specialty were as follows: endocrine 12.2%, primary care 5%, cardiology 4.7%, and renal 0.5%. Among patients seen only in a single specialty, prescribing rates were highest for endocrinology (N=1573; 17%), followed by primary care (N=2200; 11.3%), cardiology (N=1699; 10%), and renal (N=415; 3%).Among patients seen by two specialties, prescribing rates were highest for endocrine+cardiology (N=629; 27%), followed by primary care+endocrine (N=809; 26%), endocrine+nephrology (N=185; 26%), and primary care+cardiology (1169; 21%). The 816 patients seen by primary care+endocrine+nephrology had the highest prescribing rate (33%) and was similar to the 30% prescribing rate among the 325 patients seen by all 4 specialties

CONCLUSIONS: Overall rates of guideline-indicated treatment with SGLT2 inhibitors and GLP1 receptor agonists in a large academic medical center were low among both primary care and specialty providers. However, prescribing rates were higher among patients seen by multiple specialties who may prescribe the medications for guideline-indicated conditions independent of diabetes. System level interventions targeting multiple specialties and clinic types are needed to close treatment gaps.

HEALTHY BEHAVIORS IN CANCER SURVIVORS

Carter Baughman, Kathryn Norman, Kenneth J. Mukamal

Internal Medicine, Beth Israel Deaconess Medical Center, Boston, MA. (Control ID #3871967)

BACKGROUND: Primary care providers play an important role in continuity of care for people who have completed cancer treatment. The American Cancer Society (ACS) provides discrete lifestyle recommendations for these survivors, including avoiding obesity, engaging in regular physical activity, following a healthy diet, and avoiding alcohol. Adherence to these guidelines is unknown, as are factors that may affect adherence. Greater understanding of guideline adherence would better equip primary care physicians to support healthy behaviors among the increasing numbers of patients who have survived cancer

METHODS: We analyzed the 2017, 2019, and 2021 iterations of the Behavioral Risk Factor Surveillance System (BRFSS), a nationally-representative telephone survey conducted by the Centers for Disease Control and Prevention. A total of 17 states and 2 territories included modules on cancer survivorship. Adherence to the guidelines was evaluated in the four domains defined by the ACS: body-mass index <30 kg/m², exercise in the past 30 days, daily consumption of ≥2.5 servings of vegetables and ≥1.5 servings of fruit, and alcohol abstention. We summed the total number of healthy behaviors and used linear regression with survey weights to evaluate sociodemographic correlates of greater adherence.

RESULTS: A total of 10,020 adults reported having completed cancer treatment, representing 2.7 million.

Americans during the three surveyed years. Of these, 9,121 reported their adherence to all four metrics. The mean age of cancer survivors was 64±0.26 years and 56% were women. Adherence among cancer survivors was 72% (CI 0.71-0.74) for physical activity, 68% (CI 0.66-0.69) for healthy weight, 50.3% (CI 0.49-0.52) for alcohol abstention, and 12% (CI 0.11-0.13) for healthy diet. In total, 3.5% of cancer survivors adhered to all four guidelines, and the mean number of guidelines met was 2.0 (CI 2.0-2.1). Women reported 0.16 more healthy behaviors than men (p<0.001) and each decade of age was associated with 0.06 more healthy behaviors (p=0.003). Black adults reported 0.26 more healthy behaviors than white individuals (p=0.004). Greater educational attainment was associated with a roughly dose-response increase in behaviors (p=0.079). Region was also associated with adherence, with individuals in the West reporting 0.19 more healthy behaviors than those in the Midwest (p<0.001).

CONCLUSIONS: Healthy lifestyle practices are essential for cancer survivors. Our work suggests only 3.5% of cancer survivors fully adhere to current ACS recommendations. Independent factors associated with greater adherence include female gender, older age, Black race, greater education, and residence in Western states. Improved understanding of uptake of healthy behaviors will help primary care providers more effectively guide recommendations for their patients who have finished cancer treatment.

HELPING PATIENTS WITH CHRONIC CONDITIONS OVERCOME THE CHALLENGES OF HIGH-DEDUCTIBLE HEALTH PLANS

Alexandra Peirce²; Ying-Jen Lin²; Angela fa*gerlin^3,4; Michele Heisler^2,1; Jeffrey T. Kullgren^1,2

¹VA Ann Arbor Healthcare System, Ann Arbor, MI; ²University of Michigan, Ann Arbor, MI; ³University of Utah Health, Salt Lake City, UT; ⁴VA Salt Lake City Health Care System, Salt Lake City, UT. (Control ID #3877048)

BACKGROUND: Patients with chronic conditions who are enrolled in a high-deductible health plan (HDHP) experience high levels of cost-related access barriers and out-of-pocket (OOP) spending. These patients could potentially experience fewer cost-related access barriers and pay less for healthcare by using cost-conscious strategies such as having cost conversations with clinicians, saving for future costs, and using online price information. Our objectives were to develop a novel behavioral intervention to help HDHP enrollees with chronic conditions use these and other cost-conscious strategies and evaluate the acceptability, feasibility, and preliminary efficacy of the developed intervention.

METHODS: We used findings from a prior study of intervention preferences of HDHP enrollees with chronic conditions to develop a website with educational modules on discussing costs with clinicians, saving for future costs, comparing prices and quality, and steps to take before and after appointments. We developed email messages that directed individuals to each module and refined the website and emails through iterative user-centered testing. We then conducted a 5-week, single-arm proof-of-concept pilot study of the refined intervention among 36 HDHP enrollees with chronic conditions. Participants completed baseline and post- intervention surveys that measured our main outcomes of health insurance literacy (20-80 scale) and confidence in using cost-conscious strategies (0-10 scale). After the intervention we used purposive sampling to interview 10 participants with different levels of baseline health insurance literacy and intervention engagement.

RESULTS: 31 participants completed baseline and post-intervention surveys. Health insurance literacy increased from a baseline mean of 56.5 to a post-intervention mean of 67.1 (p<0.001). Surveys also showed an increase in confidence in talking to a healthcare provider about cost from a baseline mean of 6.1 to a post- intervention mean of 7.6 (p=.0094), an increase in confidence in saving for healthcare from a baseline mean of 5.8 to a post-intervention mean of 6.6 (p=.068), an increase in confidence in comparing prices from a baseline mean of 5.4 to a post-intervention mean of 6.9 (p=.005), and an increase in confidence in comparing quality from a baseline mean of 6.1 to a post-intervention mean of 7.6 (p=.0034). Post-intervention interviews showed that participants found the website easy to use and helpful for supporting use of cost- conscious strategies.

CONCLUSIONS: A novel behavioral intervention to support use of cost-conscious strategies was acceptable to HDHP enrollees with chronic conditions, feasible to deliver, and associated with an increase in health insurance literacy and confidence in using cost-conscious strategies. As more patients with chronic conditions enroll in HDHPs, this intervention should be tested in a randomized trial to evaluate its effects on cost-related access barriers, OOP costs, and other patient-centered outcomes.

HETEROGENEOUS ASSOCIATIONS OF E-CIGARETTE USAGE WITH HEALTH-SEEKING BEHAVIOR IN THE US: THE EXAMPLE OF INFLUENZA VACCINATION

Hannah Kerman¹; Kenneth J. Mukamal^2,3

¹Internal Medicine, Beth Israel Deaconess Medical Center, Boston, MA; ²Medicine, BIDMC, Brookline, MA; ³Harvard Medical School, Boston, MA. (Control ID #3872224)

BACKGROUND: Since entering the US market in 2007, E-cigarettes have grown steadily in popularity. However, concern remains that e-cigarette usage is heterogeneous in origin and outcome, with use both beneficially for smoking cessation and potentially deleteriously as a new habit or as a supplement to traditional tobacco products. We hypothesized that the association of e-cigarette usage with markers of engagement with the healthcare system, such as influenza vaccination, would differ by specific characteristics of users.

METHODS: The Behavioral Risk Factor Surveillance System (BRFSS) is a nationally-representative annual telephone survey conducted by the CDC. The 2021 survey queried e-cigarette usage and self-reported influenza vaccination in the preceding year. We determined the adjusted relationship of e-cigarette usage to influenza vaccination and tested for the consistency of this relationship across strata of age and current cigarette smoking. We used STATA 17.0 and generalized linear models to derive weighted relative risks that account for the BRFSS sampling design.

RESULTS: Among 406,664 participants with information on e-cigarette usage, 6.5% (95% confidence interval [CI], 6.3-6.7%) reported current use. The prevalence of influenza vaccination was 45.1% (44.7-45.4%). In analyses adjusted for age, sex, and race, e-cigarette usage was associated with an overall relative risk (RR) of 0.79 (95% CI: 0.76-0.83) for getting a flu vaccine. Age and current cigarette smoking both modified the association of e-cigarette usage with influenza vaccination (p<0.001 when mutually adjusted). RRs of influenza vaccination associated with e-cigarette usage were lower and similar among age groups of 18-24, 25-34, 35-44, and 45-54 (0.78, 0.67, 0.77, 0.78 respectively). In contrast, the magnitude of association of e-cigarette usage with vaccination rates was lower among respondents aged 55-64 (RR 0.90) and ≥65 (RR 0.91). Similarly, the RR of influenza vaccination associated with e-cigarette usage was 1.12 (1.07 - 1.17) among those concurrently smoking cigarettes, but 0.82 (0.77 - 0.90) among former cigarette smokers and 0.83 (0.77 - .88) among never cigarette smokers.

CONCLUSIONS: E-cigarette use is associated with a lower prevalence of self-reported receipt of influenza vaccination. However, we observed significant heterogeneity in this relationship related to both age and current smoking status. This relationship was attenuated among middle-aged and older adults and positive among current smokers, suggesting that e-cigarette usage may represent a marker of positive health behaviors among these groups. In contrast, e-cigarette usage was associated with lower prevalence of influenza vaccination among younger adults and former and never smokers, populations in which e-cigarette usage is not recommended. Guidelines for e-cigarettes should reflect the heterogeneity of their use with respect to health-seeking behaviors.

HYPOGLYCEMIA ASSESSMENT AND PREVENTION IN PRIMARY CARE VISITS

Scott J. Pilla¹; Kayla A. Meza^15,14; Judith A. Long^3,2; Howard Gordon^4,5; Jeffrey T. Bates^6,7; Donna L. Washington^8,9; Barbara G. Bokhour¹⁰; Anaïs Tuepker¹¹; Somnath Saha¹²; Mary Catherine Beach¹³; Nisa Maruthur¹⁴

¹General Internal Medicine, Johns Hopkins University, Baltimore, MD; ²Medicine, University of Pennsylvania, Perelman School of Medicine, Philadelphia, PA; ³Corporal Michael J. Crescenz VA Center for Health Equity Research and Promotion, Philadelphia, PA; ⁴Jesse Brown VA Chicago Healthcare System, Chicago, IL; ⁵University of Illinois Chicago, Chicago, IL; ⁶Michael E DeBakey VA Medical Center, Houston, TX; ⁷Baylor College of Medicine, Houston, TX; ⁸VA Greater Los Angeles Healthcare System, Los Angeles, CA; ⁹University of California Los Angeles, Los Angeles, CA; ¹⁰VA Bedford Healthcare System, Bedford, MA; ¹¹VA Portland Health Care, Portland, OR; ¹²Medicine, Johns Hopkins Medicine, Baltimore, MD; ¹³Medicine, Johns Hopkins University, Baltimore, MD; ¹⁴Medicine, Johns Hopkins University School of Medicine, Baltimore, MD; ¹⁵University of Colorado Anschutz Medical Campus School of Medicine, Aurora, CO. (Control ID #3870638)

BACKGROUND: Hypoglycemia is a major cause of morbidity and reduced quality of life in patients with diabetes, yet limited prior studies suggest it is rarely discussed in primary care visits. We conducted the first multicenter study examining how primary care providers (PCPs) assess for hypoglycemia in at-risk patients, provide anticipatory guidance, and modify medications for hypoglycemia prevention.

METHODS: This mixed methods study examined visits between patients (N=237) with diabetes using hypoglycemia-causing medications and their PCPs (N=50) at 4 geographically diverse Veterans Affairs medical centers. Visits were audio-recorded between 2017-2020, transcribed verbatim, and independently coded by 2 investigators to identify hypoglycemia dialogue that was analyzed using a directed content analysis approach. Medications and clinical data were extracted from the VA Corporate Data Warehouse. Predictors of deintensifying (decreasing/stopping) hypoglycemia-causing medications were assessed using logistic regression, adjusted for insulin use and accounting for clustering by PCP.

RESULTS: Patients’ mean age was 67 years, 89% were male, the mean A1c was 8.3%, and insulin and sulfonylureas were used by 66% and 42%, respectively. Hypoglycemia history was discussed in 77 of 237 visits (32%), assessed by the PCP in 47 visits (20%) and reported by the patient unprompted in 30 (13%). PCPs referred to hypoglycemia as “low blood sugar” while patients predominantly described symptoms or specific glucose values. Several patients expressed not understanding what the PCP meant by “low blood sugar” or reported concerns about normal glucose values being too low. Lower A1c level was associated with more frequent PCP assessment of hypoglycemia (p=0.03), but not with events. PCPs provided anticipatory guidance for hypoglycemia prevention in 50 visits (21%) consisting of holding diabetes medications while fasting and carrying glucose tabs; avoiding driving and glucagon emergency kits were not discussed. Hypoglycemia-causing medications were deintensified more often (p<0.01) where the patient reported hypoglycemia (14 of 51 visits, 27%) than where the patient reported no hypoglycemia or it was not assessed (4 of 182 visits, 2%). Deintensification was not associated with A1c, and only 6 of 15 patients (40%) who reported hypoglycemia with A1c <7.0% had medications deintensified.

CONCLUSIONS: PCPs assess for hypoglycemia in less than one third of primary care visits for at-risk patients, and use language that is not understood by some patients. Few PCPs provide hypoglycemia anticipatory guidance, which misses key concepts. Counter to diabetes guidelines, deintensifying hypoglycemia-causing medications often does not occur after hypoglycemia in patients with lower A1c levels. Routine assessment of hypoglycemia and provision of diabetes self-management education is needed to improve the safety of diabetes care.

IMPACT OF COVID-19 PANDEMIC ON STUDENT-RUN FREE CLINIC SERVICES FOR UNDERSERVED URBAN PATIENT POPULATION

Adlai Nelson¹; Sara L. Ma¹; Sara Kazyak¹; Nili Modi¹; Kathy Lu¹; Alex Wind¹; Mina Sitto¹; Jennifer Schmidt¹; Robert Sherwin²

¹School of Medicine, Wayne State University School of Medicine, Detroit, MI; ²Emergency Medicine, Detroit Medical Center, Detroit, MI. (Control ID #3874742)

BACKGROUND: The COVID-19 pandemic has significantly impacted healthcare delivery, especially in high density urban communities. This community clinic provides free medical care to patients in an urban Midwestern community. The clinic has offered longitudinal healthcare to many patients with chronic disease, such as diabetes and hypertension. Services include basic medical care services as well as limited prescriptions, vaccines, and over-the-counter medications. Adjunct services offered at the clinic by local partners include vision and hearing exams and women’s health screening. The objective of this study is to understand how the patient demographics and medical services offered at the clinic have been impacted by the COVID-19 pandemic.

METHODS: Walk-in student-run clinics were held weekly on Saturday mornings in an urban Midwestern city, with patients consisting primarily of underinsured and uninsured residents experiencing houselessness or housing instability. Number of returning and new patients, physicals, flu shots, blood glucose readings, HbA1c, lipid panels, and referrals were collected weekly by student coordinators and recorded in a spreadsheet by year (2019-2020 vs 2021-2022). Data was divided by year and quarter and compared using paired student two-tailed t-tests.

RESULTS: A total of 274 and 293 patients were seen between 2019-2020 and 2021-2022, respectively. Preliminary results demonstrate the number of patients significantly varied from December to February between these two time periods (p<0.05). Number of physicals and the number of blood glucose tests in 2021-2022 significantly differ from those of 2019-2020 (p<0.05). Number of Covid-19 vaccines significantly differed between these two time periods (p<0.01), while the number of flu shots did not.

CONCLUSIONS: Many of the patients at walk-in clinics rely on free, interdisciplinary services to obtain healthcare services, medications, and vaccines. Patients during the 2021-2022 year became more vigilant about their health, reflected by the increase in total physicals and increase in total patients (Dec-Feb). The clinic became better at pre-screening patients’ overall health, demonstrated by total blood glucose recorded, to diagnose patients earlier, provide necessary medications sooner and minimize the number of visits a patient needed to attend. Future projects could focus on factors that prevent patients’ from receiving and completing the COVID-19 vaccination series and attending clinics during peak flu/COVID seasons.

Exploring this would provide clinics insight in how to address patient’s needs and bolster patient compliance.

IMPACT OF INTRA-ARTICULAR CORTICOSTEROID INJECTION ON GLYCEMIC CONTROL: A POPULATION-BASED COHORT STUDY

Terin Sytsma¹; Laura Greenlund²; Karen Fischer³; Rozalina G. McCoy^1,4

¹Internal Medicine, Mayo Clinic, Rochester, MN; ²Internal Medicine, Mayo Clinic, Rochester, MN; ³Clinical Statistics, Mayo Foundation for Medical Education and Research, Rochester, MN; ⁴Mayo Foundation for Medical Education and Research, Rochester, MN. (Control ID #3875465)

BACKGROUND: Intra-articular corticosteroid (IACS) injections are a common pain management intervention, especially in complex, multimorbid patients. With nearly half of the United States adult population having diabetes or prediabetes, blood glucose control is a concern in a substantial proportion of patients receiving IACS. Previous case-series have described the short-term hyperglycemic effects of IACS. However, the impact of IACS on longer-term glycemic control is unknown. Furthermore, rates of dangerous hyperglycemic events following IACS are not well described. We investigated the hyperglycemic effects of IACS injections by evaluating the change in hemoglobin A1c (HbA1c) levels and examining rates of diabetic ketoacidosis (DKA) or hyperosmolar hyperglycemic syndrome (HHS) following large joint IACS.

METHODS: This retrospective cohort study examined change in HbA1c and rates of DKA/HHS among adults receiving large joint IACS in the US Upper Midwest from 2012-2018. For each individual, observed HbA1c change was compared with the predicted HbA1c change, which was calculated using linear regression. Change in actual HbA1c of ≥0.5% above predicted was deemed a “greater-than-expected rise”. Multiple logistic regression identified risk factors for greater-than-expected rise in HbA1c following IACS. Rate of DKA/HHS in the 30 days following IACS was also quantified.

RESULTS: 1169 patients were included (mean age 66.1 years-old [SD 12.06]; 52.8% female). Overall, the difference between actual and predicted HbA1c following IACS was 0.0% (SD 0.82). 184 patients (15.7%) had a greater-than-expected rise in HbA1c. This group had a higher proportion of patients with diabetes than the non-rise group (97.3% vs 91.8%, p = 0.009). Upon regression analysis, baseline HbA1c was the only factor associated with greater-than-expected rise in HbA1c following IACS with OR 4.80 (95% CI 2.83-8.14) for baseline HbA1c >8.0% compared to <7.0%. Patient age, sex, race, diabetes medications and number of comorbidities were not associated with a greater-than-expected rise in HbA1c. Only one patient had DKA attributable to IACS.

CONCLUSIONS: This is the first study evaluating effects of IACS on blood glucose control by following HbA1c after IACS in a population-based sample. The majority of patients did not have a substantial rise in HbA1c, suggesting that IACS does not substantially worsen diabetic control in most people. However, nearly 16% of patients did experienced a greater-than-expected rise in HbA1c following IACS. Baseline HbA1c was the most important predictor of greater-than-expected rise in HbA1c following IACS, suggesting that clinicians should counsel patients with sub-optimally controlled diabetes about risks of further hyperglycemia following IACS. Episodes of DKA/HHS following IACS were rare (<0.1%), thus concern for dangerous hyperglycemia should not delay IACS administration. Future, prospective studies are need to confirm these results and further delineate risk factors for elevated HbA1c following IACS.

IMPROVEMENT IN PRIMARY CARE PHYSICIAN (PCP) KNOWLEDGE ABOUT NONALCOHOLIC FATTY LIVER DISEASE (NAFLD) AFTER IMPLEMENTING A FIB-4 BASED REFERRAL STRATEGY

Janet N. Chu¹; Rena K. Fox²; Danielle Brandman⁵; Delia Falliers²; Kendall Islam⁴; Max L. Goldman³

¹Internal Medicine, University of California San Francisco, San Francisco, CA; ²Division of General Internal Medicine, University of California San Francisco, San Francisco, CA; ³Department of Medicine, University of California San Francisco, San Francisco, CA; ⁴School of Medicine, University of California San Francisco, San Francisco, CA; ⁵Center for Liver Disease, Weill Cornell Medicine, New York, NY. (Control ID #3863244)

BACKGROUND: PCPs care for many patients with NAFLD but lack confidence and consistency in its detection and management. FIB-4 is a non-invasive tool incorporating age, AST, ALT, and platelets that can be used for risk stratification of NAFLD severity, but few PCPs are aware of FIB-4 or its utility. We evaluated a FIB-4-based algorithm to help PCPs identify NAFLD patients appropriate for specialty referral and assessed PCPs’ experiences before and after this intervention.

METHODS: We conducted a three-part study at a large diverse health system: 1) a pre-intervention PCP needs assessment survey; 2) an intervention prompting PCPs’ attention to likely NAFLD patients with high FIB-4s; and 3) a post-intervention survey to determine the impact of the FIB-4 algorithm on PCPs’ practices and NAFLD knowledge one year later. During the intervention, PCPs received an electronic medical record (EMR) message for each patient with diabetes mellitus and either a low- or a high-risk FIB-4, including a recommendation to refer high-risk patients to hepatology. The message included educational content on NAFLD management by PCPs. We created two EMR tools for PCPs: an automated FIB-4 calculator with score interpretation guidance and a pre-populated NAFLD order set. Two-sample z-tests for proportions were used to compare differences between pre- and post-intervention survey responses.

RESULTS: Respondents included 88/115 PCPs to the pre- and 69/119 to the post-intervention surveys. Among post-survey respondents, 44 (64%) and 37 (54%) PCPs received a message for a low- or high-risk patient, respectively. Post-intervention, multiple knowledge areas improved including knowledge of the long- term complications of NAFLD (79% vs. 90%, p=0.01), fibrosis as a predictor of cirrhosis (76% vs. 86%, p=0.02), and the FIB-4 score components (25% vs 46%, p=0.003). PCPs increased use of FIB-4 for patients with elevated liver enzymes (16% vs. 59%, p<0.001). Yet, in diagnosing NAFLD, uncertainty remained high (42% vs. 44%, p=0.80) and post-intervention, only 13% were “very confident” and 52% “somewhat confident.” Pre-intervention, 67% reported a need for an EMR NAFLD order set with imaging, laboratory

and referral orders and 63% for an EMR FIB-4 calculator; post-survey, most were unaware of the new NAFLD order set (67%) or FIB-4 calculator (59%) yet stated they would use them if they knew it was available.

CONCLUSIONS: After implementing a FIB-4 risk stratification strategy, we found significant improvement in PCPs’ NAFLD knowledge and practices. PCPs still reported lack of confidence in diagnosing NAFLD but had increased awareness of FIB-4 and incorporated FIB-4 into their practice. Developing and promoting convenient FIB-4 tools and EMR order sets can help improve PCP confidence in diagnosing and managing NAFLD.

INCIDENCE OF NEW-ONSET DIABETES AFTER INFECTION WITH COVID-19 IN THE COVID-OUT RANDOMIZED TRIAL

Jacinda M. Nicklas¹; Thomas A. Murray²; Hrishikesh Belani³; Jared Huling⁴; David M. Liebovitz⁶; Ken Cohen⁷; Carolyn Bramante⁵

¹General Internal Medicine, University of Colorado School of Medicine, Aurora, CO; ²Biostatistics, University of Minnesota Twin Cities, Minneapolis, MN; ³Primary Care, UCLA Medical Center Olive View, Los Angeles, CA; ⁴Biostatistics, University of Minnesota School of Public Health, Minneapolis, MN; ⁵Medicine, University of Minnesota Twin Cities, Minneapolis, MN; ⁶Medicine, Northwestern University Feinberg School of Medicine, Chicago, IL; ⁷Optum Center for Research and Innovation, Minnetonka, MN. (Control ID #3875918)

BACKGROUND: There are mixed data on whether COVID-19 infection leads to increased incidence of type 2 diabetes. Recent studies have demonstrated an incidence of ~1.5% within 6 months of Covid-19 infection, as compared to a rate of 1.2% seen historically with other acute upper respiratory infections. The majority of data have so far been derived from retrospective datasets. We sought to examine the incidence of new onset type 2 diabetes in a prospective cohort enrolled in the multi-site phase 3 randomized quadruple- blinded placebo-controlled COVID-OUT clinical trial.

METHODS: Adults with overweight or obesity defined by race/ethnicity specific cutpoints and aged 30-85 were enrolled in the trial within three days of a positive SARS-CoV-2 test and fewer than seven days of symptoms. We randomized participants to one or two of three oral medications (metformin, ivermectin, fluvoxamine) or placebo using a two by three parallel treatment factorial design to efficiently share placebo controls. The intervention included immediate-release metformin with an increase in dose over the course of 6 days to 1500 mg daily for 14 days, ivermectin 430mcg/kg/day for 3 days, and fluvoxamine, 50mg on day one then 50mg twice daily through 14 days. Following their participation in the initial phase of the trial, we sent surveys via automated email or other per patient preference every 30 days through 300 days (10 months), beginning 60 days after randomization. Surveys included questions about any new medical conditions participants were diagnosed with, including diabetes mellitus. Trial registration: NCT04510194.

RESULTS: Among 1093 subjects in our sample, there were 25 new diagnoses of diabetes within 300 days of Covid-19 diagnosis, with similar incidence in each treatment condition: Metformin 2.4% vs. 2.2% control, Ivermectin: 2.9% vs. 1.9% control, Fluvoxamine 2.4% vs. 1.4% control. Participants with new onset diabetes had a higher median BMI than those who did not develop diabetes 34.3 (IQR 31.0 - 39.5) vs. 29.7 (IQR 26.8-33.9).

CONCLUSIONS: The incidence of new diabetes diagnoses was similar among outpatients treated with metformin, ivermectin, and fluvoxamine during their acute COVID-19 infection in this randomized controlled study, and higher than previously reported in retrospective cohorts studied over shorter time periods, perhaps related to the inclusion criteria restricted to participants with overweight/obesity. We will conduct multivariable models to identify predictors of diabetes incidence in this cohort.

INFLAMMATORY ARTHRITIS AFTER COVID-19 INFECTION: A CASE SERIES

Siddhant Yadav¹; Elizabeth A. Gilman¹; Nerissa Collins²; Michael Mueller²; Ryan Hurt²; Ravindra Ganesh¹

¹general internal medicine, Mayo Clinic Minnesota, Rochester, MN; ²General Internal Medicine, Mayo Foundation for Medical Education and Research, Rochester, MN. (Control ID #3873040)

BACKGROUND: Some patients who have recovered from acute infection with coronavirus 19 (COVID-19) develop persistent symptoms that have been termed post COVID syndrome (PCOS). PCOS may affect the musculoskeletal system, with arthralgias and myalgias being common. Preliminary evidence supports PCOS as an immune mediated condition predisposing and increasing risk for autoimmune diseases such as rheumatoid arthritis (RA) and reactive arthritis. In this case series, we describe patients who developed inflammatory arthritis (both reactive arthritis and RA) after COVID-19.

METHODS: We present 5 patients who developed joint pain several weeks after recovery from acute COVID-19 infection. These patients were seen in our Post COVID Clinic (PCOCC) and came from all regions of the United States.

RESULTS: All 5 patients were female with age of diagnosis of COVID-19 between 19-61 years (mean 37.8 years). All presented with joint pain as the primary complaint to the PCOCC. All patients were diagnosed with COVID between November 2020 and August 2021. The mean time difference between COVID diagnosis and onset of arthritic symptoms was 2.8 months (range 1-6 months). Abnormal joint imaging was present in all patients. Treatments varied and included NSAIDs, acetaminophen, corticosteroids, immunomodulators (golimumab), methotrexate, leflunomide, and hydroxychloroquine (Table 1).

CONCLUSIONS: COVID-19 is a potential cause of Inflammatory arthritis, with both RA and reactive arthritis demonstrated in our PCOS population. Approach to these patients with arthralgia and joint swelling requires a multidisciplinary approach which may eventually include Rheumatology.

INTEGRATING DEPRESSION AND HYPERTENSION CARE THROUGH COMMUNITY HEALTH WORKERS IN NAVRONGO, GHANA: THE COMBINE PILOT STUDY

David J. Heller¹; Engelbert A. Nonterah²; Benedict Weobong⁵; Irene Kuwolamo²; Khadija Jones¹; Allison Squires³; Carol Horowitz⁴; Raymond Aborigo²

¹Arnhold Institute for Global Health, Icahn School of Medicine at Mount Sinai, New York, NY; ²Navrongo Health Research Centre, Navrongo, Ghana; ³College of Nursing, New York University, New York, NY; ⁴General Internal Medicine, Icahn School of Medicine at Mount Sinai, New York, NY; ⁵University of Ghana, Legon, Greater Accra, Ghana. (Control ID #3876144)

BACKGROUND: Non-communicable chronic diseases now cause 71% of deaths worldwide, and the proportion is rising. However, the primary care infrastructure in many low- and middle-income countries focuses chiefly on treatment of acute infectious conditions, which have historically caused the majority of preventable morbidity and mortality. In Ghana, the nationwide Community-Based Health Planning and Services (CHPS) initiative has decreased infectious and perinatal mortality by sending nurses and health volunteers to rural homes and villages to provide vaccinations, health education, and other primary care where doctors are scarce. However CHPS does not yet provide non-communicable disease care.

METHODS: We trained eight nurses and eight health volunteers at four CHPS coverage zones to provide door to door screening for hypertension and depression, the leading global causes of death and disability respectively. Persons screening positive (a blood pressure between 140 and 179 mm Hg or 80-99 mm Hg systolic or diastolic for hypertension, or a physician’s health questionnaire [PHQ-9 ]s core of 10-19 for depression) received an eight-week intervention combining seven home counseling visits from a health volunteer with four clinic visits from a CHPS nurse offering medication. Persons with higher disease scores were referred to specialty care. Counseling comprised motivational interviewing techniques for medication adherence for hypertension, and behavioral activation techniques for depression, as well as coaching on changes in diet, exercise, and alcohol and tobacco use for all patients.

RESULTS: Results: As of November 2022, we screened 574 persons, and enrolled 37 persons with hypertension and 23 with depression. Seven persons declined the intervention. Of the 60 enrolled, one was referred to specialty care for suicidality; three were lost to follow-up; one was referred due to treatment failure, one developed liver disease unrelated to the intervention and expired, and the last migrated out of the study area. Among the remaining 56, 44 have completed the intervention to date, of which 100% achieved the endpoint of a blood pressure under 140 mm Hg systolic or a PHQ-9 score under 10. Final analysis will occur in February 2023.

CONCLUSIONS: Our findings in this pilot study suggest that CHPS nurses and volunteers can effectively screen for, diagnose, and treat these two noncommunicable diseases in the setting of routine care. Subsequent qualitative evaluation will identify strategies to improve program efficiency and efficacy, and a powered randomized controlled trial will follow to formally evaluate clinical effectiveness.

INTENSIFICATION AND DEINTENSIFICATION RECOMMENDATIONS IN CLINICAL PRACTICE GUIDELINES FOR MANAGEMENT OF CHRONIC CONDITIONS IN ADULTS

Audrey D. Zhang, Susan N. Hastings

Medicine, Duke University School of Medicine, Durham, NC. (Control ID #3875978)

BACKGROUND: There is growing interest in low-value care, defined as care in which risk of harm outweighs potential benefit. However, popular approaches such as Choosing Wisely have largely focused on highlighting specific practices constituting low-value care, rather than normalizing the identification and discontinuation of low-value practices within a continuum of care. We aimed to describe the extent to which current clinical practice guidelines (CPGs) include recommendations for both intensification and deintensification as part of guidelines for the routine care of adults, focusing on pharmacologic management of chronic conditions.

METHODS: We identified CPGs for 9 prevalent chronic conditions previously recognized to be managed in primary care (Boyd 2005) using GuidelineCentral. We included the most recent and comprehensive CPG for each chronic condition issued by national and international organizations. Within each CPG, we abstracted each recommendation pertaining to pharmacologic management of the chronic condition, identified by discrete headings or table rows with an assigned strength of recommendation and/or level of evidence. We characterized each recommendation by clinical indication, drug/drug class, and whether it was focused on initiation, withholding, intensification, or deintensification of medications. We further characterized intensification/deintensification recommendations based on whether they identified criteria (e.g. target thresholds), strategies (e.g. dose change plan), or monitoring plans (e.g. follow-up schedule) for titration. We summarized recommendations aggregated by drug or drug class-indication pairs, and compared their characteristics using chi-square tests with a p-value threshold of 0.05.

RESULTS: We identified 10 relevant CPGs for chronic conditions (two separate CPGs were identified for osteoporosis in women and men) with recommendations aggregated into 185 drug-indication pairs. Of these, 142 (76.8%) described medication initiation and use and 43 (23.2%) described medication withholding for lack of benefit or harm. Among 142 recommendations for medication use, 29 (20.4%) included any guidance on anticipated duration of use. Recommendations were more likely to include intensification than deintensification criteria (43 [30.3%] vs 16 [11.3%], p<0.001), strategies (36 [25.4%] vs 7 [4.9%], p<0.001), or monitoring plans (23 [16.2%] vs 7 [4.9%], p=0.002). Among 59 indications containing recommendations addressing low-value care, most (n=43, 72.9%) addressed withholding initiation of a medication rather than deintensification or discontinuation of an existing medication (n=16, 27.1%).

CONCLUSIONS: Existing CPGs frequently offer recommendations for initiation and intensification of medications for chronic diseases without specifying recommendations for anticipated duration of use or associated deintensification. This presents an opportunity for further elaboration to enhance high-value, patient-centered care.

INVESTIGATING THE CURRENT AND FUTURE ROLE OF PRIMARY CARE PROVIDERS IN MANAGEMENT OF POST-ACUTE SEQUELAE OF COVID-19

Katherine C. Wu⁴; Aimee E. Willett¹; Michelle L. Pershing³; Kim M. Jordan²

¹Graduate Medical Education, Internal Medicine, OhioHealth, Columbus, OH; ²Internal Medicine, OhioHealth, Columbus, OH; ³Research Institute, OhioHealth, Columbus, OH; ⁴Department of Psychiatry, University of Virginia, Charlottesville, VA. (Control ID #3876164)

BACKGROUND: Post-acute sequelae of COVID-19 (PASC) is an evolving constellationof symptoms, most notably including fatigue, malaise, cognitive dysfunction, dyspnea, and musculoskeletal pain. PASC persists for at least 12 weeks following acute COVID-19 infection. As of December 2022, 256 clinics have been established to provide multidiscipinary care for an estimated 7.7 - 23 million people suffering from PASC. To fulfill evolving needs for PASC care, primary care providers (PCPs) may have to expand their knowledge of PASC. The project aims to understand how PCPs work in collaboration with PASC clinics to ensure continuity of care, to engage in accrual of knowledge about PASC, and to increase access to PASC care.

METHODS: This project is a cross-sectional descriptive study using qualitative survey methods. A list of post-COVID-19 clinics in the United States was compiled using Survivor Corps and Becker's Hospital Review online resources. An 18-item investigator developed survey was designed on RedCap and electronically shared with clinics. Questions in Likert and multiple-choice format inquired about referral pathways to post-COVID-19 clinics and how clinics ensured dissemination of treatment information to PCPs. Survey results were collected over a 6-week period. This project recieved non-human subject research determination by the OhioHealth Riverside Methodist Hospital IRB.

RESULTS: Of the 113 PASC clinics in the United States with publically available contact information, 39 clinics (34.5%) returned completed surveys. The responding clinics were primarily community and academic hospitals with some outpatient therapy sites. PCP involvement in the reporting clinics varied: 48.7% reported PCPs were not involved at all, 41.0% indicated that PCPs are clinic physician-leads, and 2.9% indicated that PCPs can serve as consultants or in other roles in the clinic. Although 83.8% of clinics stated they agree with the aim to transfer care back to the PCP as soon as possible, 30.8% of clinics stated that that there is no formalized process to transfer care back to a PCP. Regarding instruction to PCPs on PASC care, 25.6% of clinics said that they offered workshops and educational seminars on PASC to PCPs.

CONCLUSIONS: As the COVID-19 pandemic and knowledge of PASC continues to evolve, it is anticipated that PCPs will play an increasingly important role in long term PASC care. Though a majority of survey repondents report the goal is to transfer care back to the PCP, not all PASC clinics have made active efforts to involve PCPs within patient care or have a standardized process to transfer care back to a PCP. A minority of clinics offer educational outreach to PCPs. This report highlights the need for increased post- COVID-19 clinic integration with and education of PCPs as we work to continue to provide long term care to a growing population of PASC patients.

IS ONE PILL BETTER THAN FOUR? A SYSTEMATIC REVIEW AND META-ANALYSIS OF RANDOMIZED CONTROLLED TRIALS COMPARING THE POLYPILL TO STANDARD OF CARE FOR SECONDARY CARDIOVASCULAR PREVENTION

Rimsha Ali, Taaha Mirza, Jessica Cunningham

Internal Medicine, UPMC Central PA, Harrisburg, PA. (Control ID #3870964)

BACKGROUND: Atherosclerotic cardiovascular disease (ASCVD) is known to be the leading cause of mortality and disability worldwide, yet secondary prevention rates remain low primarily due to suboptimal medication compliance. The idea of a fixed-dose combination of an anti-hypertensive, a cholesterol-lowering agent and an antiplatelet treatment into one pill (also known as the polypill) has been proposed as a strategy to potentially decrease this risk by reducing medication non-adherence. The aim of this systematic review and meta-analysis is to assess the safety and efficacy of the fixed dose combination regimen compared to the standard of care.

METHODS: A search was performed on PubMed, Cochrane, and Web of Science using the MeSH terms for "(Polypill) AND (Secondary cardiovascular prevention)” from the conception of data to 10/19/2022. A total of 514 articles were screened by two independent reviewers out of which 20 full length articles were evaluated for eligibility.

RESULTS: We identified six randomized controlled trials with a total population of 13,139 with 6567 in the treatment group and 6562 in the standard of care group. Fifty four percent were males, and the mean age of the patients was 66.8 years. The median follow-up period was one year. The main outcome of interest was the effect of the fixed dose combination on medication non-adherence. The random effects model meta-analysis for non-adherence showed a Mantel-Haenszel (MH) odds ratio of 0.40 (confidence interval (CI) 0.25-0.64; p value <0.0001) favoring the polypill. High heterogeneity was noted in this with an I² of 93% as the methods to assess compliance differed among the trials. Secondary outcomes assessed included a change in systolic blood pressure, decrease in low density lipoprotein, and mortality. The random effects model meta-analysis on the standard difference of means of systolic blood pressure among the two groups showed a minor decrease of 0.083 mmHg with a standard error of +/- 0.019 (p value < 0.0001), favoring the polypill. Low heterogeneity was noted with an I² of 0.000. Similarly, the random effects model meta-analysis showed a statistically significant decrease in the standard difference of means of mean for LDL of -0.088 mg/dL with a standard error of +/-0.024 (p value < 0.0001) with an I² of 0.000. Two trials discussed cardiovascular mortality, with random effects meta-analysis favoring the polypill group an MH odds ratio of 0.60 (CI 0.46-0.77, p value < 0.001).

CONCLUSIONS: Polypill use plays an important role in decreasing medication non-adherence. There is a minor statistical benefit of the polypill in decreasing systolic blood pressure, LDL, and cardiovascular mortality but a clinical benefit may not be appreciated. More randomized controlled trials specifically focused on secondary cardiovascular prevention with longer follow-up periods are needed to assess these outcomes.

LONG COVID AND SYMPTOM SEVERITY

Raymond Van Cleve¹; Amanda Purnell²; Robert Petrin³; Alan Roshwalb³; Kelly Bell³; Megan Shaheen³. ¹Office of Health Innovation and Learning, Veterans Health Administration, Washington, DC; ²Veterans Health Administration, Washington, DC; ³Ipsos, New York, NY. (Control ID #3876475)

BACKGROUND: Social determinants of health are an emerging field of research and that has given the Veterans’ Health Administration a new understanding of how to best care for and promote the health of their patients. Acute COVID-19 infection, morbidity, and mortality is highly associated with many social determinants of health, including but not limited to race, ethnicity, gender, income, social support structures, and access to health care. The social determinants of health associated with acute COVID-19 is aligned with other disease patterns, people who experience more stress, belong to disadvantaged demographic groups, and have lower levels of wealth are more likely to become infected with COVID-19 and are more likely to experience severe illness and mortality from COVID-19. This is true for the general US population as well as the US Veteran population (Ferguson et al, 2021)

METHODS: This study used KnowledgePanel®, a probability-based online panel, to survey 3340 Veterans. This survey asked questions regarding demographic factors, social determinants of health, their health care use, health history, and their experience with COVID-19. Specifically, this survey asked which symptoms related to COVID-19 each person experienced and the severity of those symptoms.

RESULTS: Given our preliminary analysis of US Veterans with Long COVID, none of the social determinants of health associated with Acute COVID-19 are statistically significantly associated with contracting Long COVID. Not only is the risk of developing Long COVID not associated with traditional social determinants of health, the severity of Long COVID is also not associated with any traditional social determinant of health. Race, marital status, education, income, and employment status are not in any way predictive of Long COVID or severity of Long COVID.

CONCLUSIONS: This is surprising, but it is congruent with prior research on Long COVID, that it seems to be very difficult to predict and is very different on a population level from other diseases. This presentation shares these findings and possible new observations that can be explored to better understand the course and predictors of Long COVID.

MANY PATIENTS TAKE MEDICATIONS FOR CHRONIC DISEASES WITHOUT KNOWING THEIR DIAGNOSIS

Alexander Chaitoff¹; Nancy Haff⁴; Julie Lauffenburger²; Niteesh K. Choudhry³

¹Medicine, Brigham and Women's Hospital, Boston, MA; ²Division of Pharmacoepidemiology and Pharmacoeconomics, Brigham and Women's Hospital, Harvard Medical School, Boston, MA; ³Division of Pharmacoepidemiology and Pharmacoeconomics, Harvard Medical School, Boston, MA; ⁴Medicine, Brigham and Women's Hospital Department of Medicine, Boston, MA. (Control ID #3875484)

BACKGROUND: Patients' understanding of a given medical condition is often a prerequisite for participating in shared-decision making (SDM) conversations and other aspects of care. Arguably the most fundamental piece of information a patient can have is whether they carry the diagnosis for a given disease. The aim of this study was to assess the prevalence of discrepancies between patient self-reported chronic conditions and their prescribed medications.

METHODS: The sample consisted of adults who took part in the National Health and Nutrition Examination Survey (NHANES) from January 2017 to March 2020. NHANES is a cross-sectional survey that uses a complex, multistage, clustered probability method to sample non-institutionalized civilians to provide nationally representative estimates. Respondents self-reported (yes/no) whether they had been told they had hypertension, hyperlipidemia, or diabetes, respectively. Respondents also self-reported (yes/no) whether they were currently taking any medications for each of these diseases, respectively. Actual medication use was confirmed by pill bottle examination. Frequencies and counts were used to compare discordance between self-reported disease status, self-reported medication status, and actual medication use status. Chi-square and t-tests were used to assess for characteristics associated with discordance between stated disease status and actual medication use status. All percentages reported in the results are weighted while all counts are unweighted unless otherwise specified.

RESULTS: Of the 9217 respondents, 8.2% (95%CI 7.6%-8.8%, n=888), corresponding to a weighted count of 19,696,464 individuals (95%CI 17,509,460-21,883,469), reported never receiving a diagnosis of diabetes, hypertension, or hyperlipidemia but were found to be taking at least one medication to treat these diseases. Of respondents who stated they had one of these chronic diseases, 5.4% (95%CI 4.3%-6.5%, n=285) stated they were currently taking medications for the respective condition but did not have a corresponding medication on pill-bottle examination, while 2.8% (95%CI 2.2%-3.4%, n=110) stated they were not currently taking medications for the respective condition but did have a corresponding medication on pill-bottle examination. Age, sex, race/ethnicity, education, and total number of prescriptions were not associated with discordance between respondents’ self-reported and actual medication statuses (p>0.05). However, those who had discordance between their self-reported and actual medication status did have lower average family-income- to-poverty-ratios (2.8 versus 3.2, p=0.03).

CONCLUSIONS: Many people are taking medications for chronic diseases that they report not knowing they have been diagnosed with. These discordances are not clearly related to sociodemographic factors. The prevalence of this issue and the barrier it presents for SDM conversations necessitates implementation science strategies to better educate patients.

OUTCOMES OF PATIENTS WITH MITRAL REGURGITATION AND CONCOMITANT AORTIC AND TRICUSPID VALVE DISEASE AFTER TRANSCATHETER EDGE-TO-EDGE MITRAL VALVE REPAIR

Kristina D. Menchaca¹; Nemanja Draguljevic²; Catherine A. Ostos¹; Virginia Velez Quinones¹; Guo Cheng³; Marcos Nores¹; Robert Chait¹

¹Internal Medicine, University of Miami School of Medicine, Miami, FL; ²Univerzitet u Beogradu, Beograd, Beograd, Serbia; ³University of California System, Oakland, CA. (Control ID #3875415)

BACKGROUND: Patients with severe mitral valve regurgitation (MVR) and coincidental aortic or tricuspid valve disease represent a management challenge. Traditionally, if an interdisciplinary heart team decides to perform open-heart surgery to replace the mitral valve, aortic stenosis or regurgitation is usually corrected simultaneously. However, patients at high surgical or prohibitive risks are usually referred to a less invasive alternative, such as transcatheter edge-to-edge mitral valve repair (TEER) for severe mitral valve regurgitation. We decided to observe the outcomes of the patients who underwent TEER for severe mitral valve regurgitation associated with aortic or tricuspid valve disease.

METHODS: In this retrospective cohort study, 170 patients participated that underwent TEER. Patients only with severe MVR 73/170 (42.9%) were compared to patients who had associated either moderate to severe aortic stenosis or regurgitation or tricuspid insufficiency 97/170 (57.1%). The patients had similar basic characteristics in terms of gender, prior diseases such as stroke, peripheral arterial disease, diabetes, heart failure, chronic lung disease, KCCQ12 score, MR severity, and ejection fraction. We assessed the impact of coincidental aortic or tricuspid valve disease on post-TEER outcomes such as MR reduction, total-in-hospital stay, and mortality.

RESULTS: Post-TEER, no difference was found in reducing the severity of MR (p=0.91.), total-in-hospital stay (p=0.77), and survival between patients with only MVR and patients with MVR associated with either aortic stenosis or regurgitation or tricuspid insufficiency. Both groups exhibited a reduction in the severity of MR, average total in-hospital stay of 6.33~ 6.73 (mean 6.53), and survival of 100% compared to another group with good outcomes.

CONCLUSIONS: There was no difference in feasibility and short-term outcomes of TEER in patients only with MVR compared to patients with MVR and associated aortic or tricuspid valve disease. In the era of minimally invasive procedures such as TEER, when performed in patients with concomitant and very often complicated valve pathology, we need large-scale trials to follow up on these patients' short- and long-term outcomes.

PATIENT EXPERIENCES NAVIGATING CARE COORDINATION FOR LONG COVID: LESSONS FOR HEALTHCARE SYSTEMS AND PROVIDERS

Sarah MacEwan¹; Saurabh Rahurkar²; Willi Tarver³; Cortney Forward⁴; Jennifer Eramo⁴; Alice Gaughan⁴; Laura J. Rush⁴; Erin McConnell⁵; Andrew Schamess⁶; Ann S. McAlearney⁷

¹General Internal Medicine, The Ohio State University, Columbus, OH; ²Biomedical Informatics, The Ohio State University, Columbus, OH; ³Cancer Prevention and Control, The Ohio State University, Columbus, OH; ⁴CATALYST, The Ohio State University, Columbus, OH; ⁵General Internal Medicine, The Ohio State University Wexner Medical Center, Columbus, OH; ⁶Division of General Internal Medicine, The Ohio State University Wexner Medical Center, Columbus, OH; ⁷Family and Community Medicine, The Ohio State University, Columbus, OH. (Control ID #3874495)

BACKGROUND: Long COVID is emerging as a new chronic condition for individuals who continue to experience negative health effects after the resolution of an acute COVID-19 infection. Little is known about best practices to evaluate, diagnose, and treat long COVID which presents challenges to patients as they seek care to address their symptoms. This study served to better understand the perspectives of patients as they navigate care coordination for long COVID.

METHODS: One-on-one interviews were conducted with adult patients seeking care at a post-COVID recovery clinic. Patients for which long COVID symptoms were continuing to have an impact on their lives at 3 months or more after an acute COVID infection were eligible to participate. Interviews were conducted using a semi-structured interview guide that covered topics including long COVID symptoms and treatments, support from healthcare providers, and what participants would want providers to know about being a patient with long COVID. Thematic analysis of interview transcripts was used to characterize perspectives relative to the following components of care coordination: access to care, evaluation/diagnosis, care planning, and follow-up.

RESULTS: Interviews were conducted with 21 patients ranging in age from 19-68 years. The majority (76%) were female. Nearly one half of patients (48%) were first infected with COVID in 2020; approximately a quarter were first infected in 2021 (28%) or 2022 (24%). Patients described differing experiences accessing care for long COVID. Some noted that providers listened to their concerns and validated their symptoms which facilitated access to subsequent care. Others reported experiences in which providers did not believe their symptoms which they felt hindered their access to care. Many patients described experiences of evaluation and diagnosis related to long COVID as complicated processes involving coordination among primary and specialty providers. Patients reported confusion about communicating with numerous providers and not knowing which overlapping symptoms to discuss with each specialist. Some patients reported frustration when diagnoses did not affirm a connection with long COVID. Overwhelmingly, patients acknowledged that their providers were still learning how to treat long COVID, and they expressed an appreciation for providers who were willing to try new approaches to treatment. Finally, patients described concerns related to their follow-up including questions about if/when their long COVID symptoms would improve and if long COVID would affect their life expectancy or other aspects of aging, such as cognitive decline.

CONCLUSIONS: Our findings demonstrate the challenges of patients with long COVID as they navigate healthcare for this new chronic condition. Shedding light on these patient perspectives serves to inform improvements in the provision of compassionate and effective care, and especially in care coordination, for this population.

PATIENT-REPORTED BARRIERS TO EXERCISE AND ACCEPTABILITY OF A REMOTELY DELIVERED HOME CARDIAC EXERCISE REHABILITATION FOR HEART FAILURE

Yulia Khodneva, Andrea Cherrington, Elizabeth A. Jackson

Medicine, The University of Alabama at Birmingham School of Medicine, Birmingham, AL. (Control ID #3875826)

BACKGROUND: Adults living in the Deep South are disproportionally affected by heart failure (HF) mortality — with HF annual deaths reported as 203 per 100,000 persons compared to 168 per 100,000 persons in the Northeast of the US. This significant regional difference is related to higher poverty rates, poor adherence to important lifestyle modifications, such as physical activity, and higher levels of comorbidities, including hypertension, obesity and diabetes, in adults with HF in the South. Cardiac rehabilitation (CR) is an existing intervention for patients with HF that focuses on exercise training, dietary modifications and stress management; it is an effective but grossly underutilized intervention. Only 10% of eligible patients with HF are referred to CR anually. This study investigated the roots of this under-utilization by engaging HF patients into a patient-oriented participatory research.

METHODS: We aimed to identify patient-reported barriers to physical activity and to investigate the HF patients’ level of acceptability of home-based CR program, delivered remotely. We conducted a qualitative study of patients with HF via in-depth semi-structured interviews, guided by the framework of the “Theory of Planned Behavior”.

RESULTS: 22 patients with HF, stage C, aged 27-85, were recruited from two outpatient HF clinics in Birmingham, AL; of them 58% were women, 42% men, 73.7% African American, 21.1% White, 6.2 % Hispanic; 58% had HF with preserved ejection fraction. Patients with advanced HF, on home inotropic infusion and/or on Left Ventricular Assist Device (LVAD) were excluded. Interview were coded using two independent coders and Nvivo software. Major and minor themes were identified. Patient-reported barriers to physical activity included severity of heart failure symptoms, chronic pain, obesity, depressive symptoms, stress, feeling of grief after receiving HF diagnosis, fear of heart attack, low motivation, lack of knowledge about the type of exercise, appropriate for HF, lack of guidance from physicians and lack of access to specialized programs and qualified trainers. Participants requested more supervision during the exercise and were skeptical about programs, deliver via video-based platforms.

CONCLUSIONS: In the diverse sample of adults with HF the level of physical activity was low, patients reported numerous barriers to exercise and were skeptical about remotely delivered CR. This results will inform CR intervention development for HF; and a hybrid (combination of in-center and home-based) may be a more acceptable mode of delivery.

PERITONEAL DIALYSIS MANAGEMENT AUGMENTED WITH POINT OF CARE LUNG ULTRASONOGRAPHY:

A PILOT IMPLEMENTATION PROJECT

Robert J. Williams, Carolina Ortiz-Lopez, Isaac Teitelbaum, Michelle Fleshner

Internal Medicine, University of Colorado Anschutz Medical Campus, Aurora, CO. (Control ID #3877132)

BACKGROUND: Point of care lung ultrasound (lung POCUS) has been increasingly utilized as a tool to augment traditional physical exams. More specifically, it has been utilized to perform volume assessments in patients with end stage renal disease (ESRD) on hemodialysis (HD) by identifying sonographic signs of pulmonary congestion. It is unknown, however, if lung POCUS can guide treatment in the peritoneal dialysis (PD) patient population. We aimed to perform a pilot implementation study to assess the feasibility and utility of integrating lung POCUS into a single outpatient nephrology clinic (UCH) to help manage a cohort of PD patients.

METHODS: A cohort of eighteen PD patients were followed over four consecutive monthly nephrology appointments. Three monthly appointments included a standardized 12-zone lung POCUS exam performed by credentialed Internal Medicine (IM) hospital medicine providers. The presence or absence of lung congestion identified by lung POCUS (presence of bilateral B-lines and/or pleural effusion(s)) was communicated to the patients’ treating nephrologist following their nephrologist’s standard examination. The patients treating nephrologist could decide whether to incorporate the lung POCUS into final day-of-visit management decisions. Patient weight, blood pressure, physical exam versus POCUS exam findings, and management changes were recorded over the study period. Informal qualitative interviews were performed with two treating nephrologists to assess their perception of feasibility and utility of lung POCUS in their PD clinic.

RESULTS: Forty-five of an anticipated fifty-four (83.3%) lung POCUS exams were successfully completed. Of all offered POCUS exams, only one was declined by a participant (2.1%) during their visit and eight exams (14.8%) were not completed primarily due to patient scheduling issues. Five participants (11.1%) had volume overload identified by physical exam (pretibial edema, elevated jugular venous pulse, or crackles on lung auscultation) whereas pulmonary congestion was noted on 11 (24.4%) participants’ lung POCUS. Lung POCUS findings were discordant with a traditional physical exam 26.6% (N = 12) of the time. Fourteen patient encounters were associated with a clinical management change, of which 72% (N= 8) occurred when lung POCUS and traditional volume exams were discordant . Informal qualitative interviews with treating nephrologists revealed feasibility of lung POCUS but uncertainty as to the utility of this addition, particularly in relatively stable outpatients without respiratory complaints.

CONCLUSIONS: It is feasible to integrate lung POCUS into the care of outpatient peritoneal dialysis patients. Lung POCUS may provide additional information regarding volume status of PD patients, however, its influence on treatment decisions, particularly among stable outpatients, is not clear. Further research is needed to better understand the utility of lung POCUS in managing PD patients.

POST-ACUTE SEQUELAE OF SARS COV-2 INFECTION (PASC) IN VETERANS SEEN AT A LARGE, INNER-CITY VA MEDICAL CENTER

Howard Gordon^1,2; Israel Rubinstein^1,2

¹Jesse Brown VA Chicago Healthcare System, Chicago, IL; ²University of Illinois Chicago College of Medicine, Chicago, IL. (Control ID #3875865)

BACKGROUND: After contracting acute SARS-CoV-2 illness, some patients experience a plethora of physical and mental health manifestations that last four or more weeks, the Post-Acute Sequelae of SARS

CoV-2 infection (PASC). We sought to determine the burden of PASC among Veterans with prior PCR- positive COVID-19 infection seen at a large, inner-city VAMC.

METHODS: We mailed surveys to Veterans at a large inner-city VAMC, who had a positive RT-PCR test for SARS-CoV-2 and had survived at least 30-days after diagnosis (N = 1,849). Respondents indicated symptoms they experienced during their acute COVID-19 illness and if they were still experiencing symptoms. Demographic and clinical characteristics were obtained through medical record linkage. Frequency of initial COVID-19 symptoms and long-lasting symptoms were reported. Chi-square tests were used to compare differences between COVID-19 initial and on-going symptoms by race and age. Written responses to an open-ended question about experiences with COVID-19 were analyzed with qualitative content analysis.

RESULTS: A total of 211 surveys were completed and returned between February and July 2022. Survey respondents had mean age 65.7 years (SD 12) and were 93.8% male (N = 197), 54% Black (N = 114), 13.2% Latinx (N = 27), 71.6% had a high school education or higher (N = 151), 69.1% (N = 143) lived with at least one other person, and 53.5% (N = 93) did not report trouble affording basic needs such as food and utilities. Survey respondents experienced a mean of four (SD 4.1) PASC symptoms. The most frequent on-going symptoms reported were fatigue (41.9%), trouble sleeping or staying asleep (40.5%), mental fog (36.7%), shortness of breath (32.2%), and body aches (28.4%). There were no statistically significant differences between initial symptoms, on-going symptoms, and race. When asked about on-going symptoms, those 65 years and older were more likely to report trouble sleeping (p = 0.029) and mental fog (p = 0.01). These older Veterans were twice or more likely to seek care for their PASC symptoms compared to younger patients (p = 0.02). Older Veterans experienced more symptoms when initially diagnosed with COVID-19 in comparison to younger Veterans (p = 0.002). We collected 95 (45%) responses to an open-ended question about experiences with COVID-19 and 30 Veterans provided accounts of severe illness that included descriptions of perceived near death experiences and prolonged hospital stays. Also, 12 Veteran respondents described difficulty with PASC, including ongoing respiratory, GI, neurological and cognitive symptoms.

CONCLUSIONS: Though these findings are limited by the low response rate, they indicate a substantial burden of PASC in this inner-city Veteran population. On average, survey respondents continued to experience 4 symptoms long after acute COVID-19 illness. We posit that assessment of PASC burden by age, race, sex, and other clinical characteristics could inform the need for dedicated PASC clinics in Veterans Health Administration.

PREDICTORS OF DEVELOPING LONG COVID FROM PROSPECTIVELY COLLECTED DATA FROM A PHASE 3 CLINICAL TRIALS.

Hrishikesh Belani²; Seema Belani³; Jacinda M. Nicklas⁴; Sarah Fenno⁵; Spencer Erickson⁶; Carolyn Bramante¹

¹Medicine, University of Minnesota Twin Cities, Minneapolis, MN; ²Medicine, Los Angeles County Department of Health Services, Los Angeles, CA; ³College of Allopathic Medicine, Nova Southeastern University College of Allopathic Medicine, Davie, FL; ⁴General Internal Medicine, University of Colorado School of Medicine, Aurora, CO; ⁵University of Minnesota Medical School, University of Minnesota Medical School Twin Cities, Minneapolis, MN; ⁶General Internal Medicine, University of Minnesota Medical School Twin Cities, Minneapolis, MN. (Control ID #3877381)

BACKGROUND: The COVID-OUT trial was a phase 3, randomized, quadruple blinded clinical trial of early outpatient treatment of Covid-19 using generic medications (Trial registration: NCT04510194). One unique aspect of the trial is that it continued 10-month follow-up, the longest follow-up from any phase 3 outpatient Covid-19 treatment trial at the time of this submission.

METHODS: For 10 months after randomization, participants were sent monthly serveys asking questions including: had they been diagnosed with Long Covid; had they received any new diseases; started medications; had they been re-infected with Covid-19, and if so how was this diagnosed (home test or a lab), and was their subsequent infection more or less severe than their initial infection; had they received booster vaccinations. Patient-reported diganoses i.e. Long Covid; new onset diabetes, were confirmed in the health record. This is an observational analysis of risk factors for developing Long Covid, controlling for study drugs.

RESULTS: Overall 1,125 participants consented for longterm follow-up and completed at least one survey about Long Covid. Two percent identified as Native American; 3.7% as Asian; 7.4% as Black/African American; 82.8% as white; and 12.7% as Hispanic/Latino. The median BMI was 29.8 kg/m² (IQR 27 to 34); 51% had a BMI >30kg/m². Overall, 8.4% reported a medical provider diagnosed them with Long Covid. As seen in the table, among those who reported being reinfected with Covid-19, 13% developed Long Covid. Of those with Long Covid, 74.5% were female, versus 54% of those with no Long Covid; 11.7% were Black, versus 7.0% without Long Covid; 44% were vaccinated before enrollment versus 56% without Long Covid. Of those with Long Covid, 83.7% did not seek care at the ED; 10.9% had ED visits; and 5.4% were hospitalized during their initial Covid infection. Of those with Long Covid, 28 (29.8%) had been re-infected with Covid-19; and 54% had obesity, compared to 48% of those without a Long Covid diagnosis.

CONCLUSIONS: In this cohort of adults who participated in a randomized clinical trial, overall 8.4% developed long Covid. Women, those with obesity, those who had been reinfected with SARS-CoV-2, and those who identify as Black appear more likely to report a medical provider diagnosis of Long Covid.

PREDISPOSING, ENABLING, AND RETAINING FACTORS FOR PRIMARY CARE STAFF AS COACHES IN AN ELECTRONIC HEALTH RECORD (EHR)-BASED WEIGHT MANAGEMENT STUDY: DATA FROM THE MAINTAIN PRIME STUDY (PROMOTING REAL-WORLD IMPLEMENTATION)

Maribel Cedillo²; Jesell Zepeda³; Emily Zheutlin⁴; Polina Kukhareva⁵; Bernadette Kiraly⁸; michael c. flynn⁶; Jorie Butler⁵; Rachel Hess¹⁰; Kensaku Kawamoto⁷; Paul Estabrooks⁹; Molly Conroy¹

¹General Internal Medicine, University of Utah, Salt Lake , UT; ²General Internal Medicine, University of Utah, Salt Lake, UT; ³Department of General Internal Medicine, University of Utah Health, Salt Lake City, UT; ⁴Internal Medicine, University of Utah Health, Salt Lake City, UT; ⁵Department of Biomedical Informatics, University of Utah, Salt Lake City, UT; ⁶Community Clinics, University of Utah, West Valley CIty, UT; ⁷Biomedical Informatics, University of Utah, Sandy, UT; ⁸Family and Preventive Medicine, University of Utah Health, Salt Lake City, UT; ⁹Health and Kinesiology, University of Utah Health, Salt Lake City, UT; ¹⁰Population Sciences, University of Utah, Salt Lake City, UT. (Control ID #3876432)

BACKGROUND: MAINTAIN PRIME (MP) is an adaptation of an effective weight management program designed for implementation by primary care staff. Key to the success of MP is the use of coaching to facilitate sustained weight loss. We sought to determine the utility of using a theory-based planning tool focused on predisposing, enabling, and retaining (PER) factors to guide coach training strategies.

METHODS: Medical assistants, nurses, and other staff were recruited as coaches from participating University of Utah primary care clinics. The PER tool guided our training plan. Predisposing: At an environmental level, motivation was achieved by clinic participation and leadership engagement. Leadership support for MP was crucial for staff motivation in understanding MP value. Staff required clinic leaders’ approval to participate in training, to which all staff was invited. Next, we aimed to attract staff with intrinsic motivation for lifestyle coaching. We designed a stepwise self-selection process, including registration, confirmation of interest, online self-paced training, and electronic health record (EHR) hands-on training. Enabling: Training provided skills, gave access to evidence-based resources (including EHR smart phrases), and removed barriers such as lack of confidence, writing skills, and time allocation. Reinforcing: Reminders and support from registered dietitians were planned to promote retention.

RESULTS: Predisposing: 65 of 187 staff from various clinical roles (35%) at 11 participating clinics expressed interest in training. 50 potential coaches viewed an informational video; 46 (92%) confirmed

interest. Intrinsic Motivation. 31 of 46 (67%) who confirmed interest completed training; 28 completed training and started coaching, 3 completed training but did not coach, and 12 potential coaches started but did not complete training. Enabling coaches. Coaches who completed post-training evaluations (97%) felt they had learned enough to coach patients (5.9, 7=you feel ready) and were likely to recommend the training to others (8.0, 10=extremely likely). All (100%) reported achieving learning objectives. Most (93%) felt the course had practical value; 97% reported the course and instructor met their needs and 93% agreed with instruction method, rating both online and EHR hands-on training components highly (5.2 and 6.1, respectively; 7=max). Reinforcing. Turnover of clinical staff is common. 13 (43%) coaches have left the clinic site, and completed an exit survey reporting coaching seemed implementable, possible, doable, and easy (4.2, 4.2, 4.2, 3.7, respectively; 5=completely agree). The highest rated coaching role components were helping and connecting with patients. The tool that most facilitated coaching was EHR smart phrases. Opportunities for improvement related to EHR functions included task due dates, patient reminders, and EHR documentation.

CONCLUSIONS: The PER strategies allowed us to implement a well-received training program found effective by primary care coaches.

PREVALENCE AND DISPARITIES IN AWARENESS AND KNOWLEDGE OF MEDICATION OVERUSE HEADACHE AMONG PATIENTS WITH MIGRAINE

Stacy Bailey¹; Wei Huang¹; Andrea Zuleta¹; Yvonne Curran³; rodolfo zuleta¹; Melissa Herman⁴; Steven M. Kymes²; Allison Pack¹

¹General Internal Medicine, Northwestern University, Evanston, IL; ²Lundbeck LLC, Deerfield, IL; ³Neurology, Northwestern University Feinberg School of Medicine, Chicago, IL; ⁴H Lundbeck A/S, Valby, Hovedstaden, Denmark. (Control ID #3875824)

BACKGROUND: Medication overuse headache (MOH) occurs when patients with a pre-existing headache disorder overuse medication in an attempt to relieve headache pain or symptoms. MOH results in additional headaches and, for many patients, is a root cause of the transition from episodic to chronic migraine. Patient education is necessary to inform patients of the risks associated with MOH and to ensure that prescription and over-the-counter (OTC) pain relievers are being used safely and effectively. Yet few studies have comprehensively examined patient knowledge of MOH, its risk factors, and signs of progression.

METHODS: English-speaking adults with a history of migraine and a clinic visit within the past month were identified via an electronic health record (EHR) query from an academic health center in Chicago, IL.

Potential participants were contacted by research coordinators, who introduced the study, engaged interested patients in the informed consent process, and enrolled participants. Participants completed one survey that measured MOH awareness and knowledge, patient-provider communication, and sociodemographic characteristics.

RESULTS: A total of 200 adults participated. Participants were predominately female (82.9%) and middle- aged, 57.4% were White, 31.8% made <$50,000 annually, and 18.2% had limited health literacy. On average, participants had been diagnosed with migraine for over a decade; 64% had severe migraine-related disability. More than a third (35.9%) reported taking an OTC pain reliever on ≥15 days in the past month, putting them at risk of MOH. A total of 38.6% of participants reported that they had never heard the term ‘medication overuse headache.’ In bivariate analyses, participants who were Black or Hispanic, those with limited health literacy, and patients with less severe disability were less likely to have heard of the term. Almost 1 in 5 participants (19.0%) did not know that ‘medications taken to treat headache can also cause headaches.’ Participants who were older, less educated, and with limited health literacy were significantly less likely to be aware of this concept. Participants who reported better patient-provider communication were more likely to have greater MOH knowledge and to know how often they could take OTC medicine to avoid MOH.

CONCLUSIONS: Findings indicate that awareness and knowledge of MOH is suboptimal, with clear disparities by patient sociodemographic factors. Results also indicate about one-third of migraine patients are at risk of MOH due to high frequency OTC use. Additional patient education and improved patient-provider communication about MOH may help prevent the condition.

PREVALENCE OF PRESCRIPTION MEDICATION USE THAT CAN CAUSE OR EXACERBATE HEART FAILURE AMONG ADULTS WITH HEART FAILURE

Alexander Chaitoff¹; Joshua D. Niforatos²; Alexander Zheutlin³

¹Medicine, Brigham and Women's Hospital, Boston, MA; ²Johns Hopkins Medicine, Baltimore, MD; ³University of Utah Health, Salt Lake City, UT. (Control ID #3876279)

BACKGROUND: Professional societies are increasingly recognizing the role of iatrogenesis in disease outcomes. The 2022 American Heart Association (AHA)/American College of Cardiology (ACC)/Heart Failure Society of American (HFSA) heart failure (HF) guidelines identify medications that can cause or exacerbate HF and should be avoided when possible. In this study, we determined the prevalence of using these medications among adults with self-reported HF.

METHODS: The sample included US adults (> 20 years) with self-reported HF participating in the National Health and Nutrition Examination Survey (NHANES) between 2011 and March 2020. NHANES samples non-institutionalized civilians to provide nationally representative estimates. Respondents' use of medications the AHA/ACC/HFSA identified as potentially causing/exacerbating HF was assessed by pill-bottle review. We describe the prevalence of taking these medications overall, by sex (male/female), by race/ethnicity (non- Hispanic White, non-Hispanic Black, Hispanic, non-Hispanic Asian, Other/Multi-race), and stratified by level of evidence for medication harm as reported by the guidelines (A vs. B/C). We used multivariable logistic regression to generate odds ratios (aOR) of being prescribed one of these medications by gender and race/ethnicity adjusting for age, education status, and income-to-poverty ratio, insurance status, self-reported comorbidities (arthritis/diabetes/hypertension/nocturia), total number of medications, and survey cycle. Proportions are presented as weighted estimates while counts are unweighted unless otherwise stated.

RESULTS: Among adults with HF (n=687), 26.1% (95% CI 21.3%-30.9%, n=173) were prescribed a medication that can cause/exacerbate HF, which represents approximately 1.42 million adults. 4.1% of respondents were on a medication with A level evidence of potential to cause/exacerbate HF, while 24.4% were on a medication with B/C level evidence of potential to cause/exacerbate HF. 32.1% (95% CI 25.4-38.8%, n=113) and 20.4% (95% CI 13.8%-27.0%, n=60) of males and females, respectively, were prescribed a medication that can cause/exacerbate HF. Male participants were more likely to be prescribed a medication that can cause/exacerbate HF (aOR 2.24, 95% CI 1.40-3.60). In multivariable analysis there were no statistically significant differences in medication use that can cause/exacerbate heart failure by race (p>0.05 for all). History of hypertension (aOR 2.27 95%CI 1.09-4.70), nocturia (aOR 1.23, 95%CI 1.07-1.41), and number of medications (aOR 1.23, 95%CI 1.15-1.31) were the only model covariates that reached statistical significance.

CONCLUSIONS: 26.4% of adults with self-reported HF were prescribed medications that the AHA/ACC/HFSA recommends avoiding. Balancing the competing demands of HF- and non-HF medications is imperative to maximizing the quantity and quality of life for adults with HF, and increased attention to polypharmacy likely has a role to play in reducing HF morbidity.

PROCEDURAL/SURGICAL RHYTHM CONTROL FOR PATIENTS WITH ATRIAL FIBRILLATION: A META-ANALYSIS

Bryce Montane¹; Shiyang Zhang¹; Jonathan D. Wolfe¹; Daniel H. Cooper¹; Michelle Doering¹; Brian F. Gage². ¹Washington University in St Louis, St Louis, MO; ²Medicine, Washington University in St Louis School of Medicine, St Louis, MO. (Control ID #3875471)

BACKGROUND: Atrial fibrillation quadruples the risk of ischemic strokes. Whether rhythm control via catheter ablation or surgical rhythm reduces this risk is unknown. Our objective was to understand long term ischemic stroke risk for different treatments of atrial fibrillation.

METHODS: We combined prior trials to quantify the reduction on ischemic stroke from rhythm control via catheter ablation or surgery (e.g. the Maze procedure). We included trials that met the following inclusion criteria: >1 month of follow up, adult patients with atrial fibrillation, English language, and randomization to either catheter ablation or surgical rhythm control vs. conservative therapy. The primary outcome was the relative risk of ischemic stroke after a 30-day blanking period (when no strokes were included). We combined trials using the Cochran-Mantel-Haenszel method to quantify the relative risk (RR) of stroke. We used the Haldane correction for trials that observed 0 strokes in either arm.

RESULTS: The 22 trials of catheter ablation randomized a total of 5,659 patients and had a RR (95% CI) of 0.64 (0.43-0.95). The 20 trials of surgical rhythm control randomized a total of 2,083 patients and had a relative risk (95% CI) of 0.64 (0.41-0.998). Together, the 42 trials had a RR (95% CI) for ischemic stroke of 0.64 (0.48-0.87).

CONCLUSIONS: After the 30-day blanking period, catheter ablation or surgical rhythm control reduced the risk of ischemic stroke in trial participants with atrial fibrillation. We hypothesize that over a longer period, maintenance of sinus rhythm associated with procedural/surgical rhythm therapy reduces the risk of ischemic stroke for patients suffering from atrial fibrillation.

READINESS TO TAPER: PSYCHOSOCIAL FACTORS ASSOCIATED WITH MOTIVATION TO TAPER OPIOIDS AMONG INDIVIDUALS WITH CHRONIC PAIN

Danielle Wesolowicz^1,2; Sara Edmond^1,2; Brent A. Moore^3,1; Manik Chhabra⁵; Liana Fraenkel^4,2; William Becker^1,2

¹Research, VA Pain Research Informatics Multi-morbidities and Education Center, West Haven, CT; ²Yale School of Medicine, New Haven, CT; ³Psychiatry, Yale School of Medicine, New Haven, CT; ⁴Berkshire Medical Center, Pittsfield, MA; ⁵VA Center for Health Equity Research and Promotion, Philadelphia, PA. (Control ID #3872943)

BACKGROUND: Guidelines encourage primary care providers to taper or discontinue long-term opioid therapy for chronic pain when harms outweigh benefits; however, tapering may come with challenges including withdrawal symptoms, distress, and fears of increased pain. Patient-centered approaches to tapering are necessary to manage such risks and enhance likelihood of success. Given the difficulties associated with tapering, a patient’s motivation to taper may be an important variable to target when considering approaches to support patients interested in tapering. The purpose of this analysis was to explore psychosocial predictors of motivation among veterans with chronic pain voluntarily undergoing an opioid taper.

METHODS: We performed this cross-sectional study among thirty veterans with chronic pain in VA care (80% men, mean age= 65, 33% White, 60% Black, 7% other) enrolled in a pilot randomized controlled trial evaluating the feasibility and acceptability of a web-based, interactive program to support opioid tapering. Prior to randomization, participants completed baseline questionnaires, including assessment of motivation to change opioid regimens (Change Questionnaire; CQ). Based on CQ scores, we divided participants into categories reflecting low, moderate, and high motivation to taper. We conducted ordinal logistic regression to examine associations between independent variables (demographics, experience of opioid-related side effects, functioning, pain, anxiety, and depression) and the dependent variable (motivation to taper category).

RESULTS: Results from ordinal logistic regression indicated patients were more likely to have high motivation to taper if they reported higher pain intensity (OR=10.93, 95% CI=1.66-72.10), greater anxiety (OR=1.30, 95% CI=1.30-20.18), lower pain interference (OR=.31, 95% CI=.11-.88), were younger (OR=.57; 95% CI=.37-.87), and reported higher physical functioning (OR=.16, 95% CI=.03-.07). Gender, race, opioid side effects, and depression were not associated with motivation level.

CONCLUSIONS: Among patients voluntarily tapering opioids, having higher pain, greater anxiety, and having lower pain interference were independently associated with increased odds of being more motivated to taper. Difficulty engaging in physical activity and activities of daily living may serve as a barrier to tapering among patients considering tapering opioids, and could be an important target for efforts to enhance motivation. Future studies should further explore the association between motivation and tapering to inform potential interventions to enhance tapering outcomes.

RELATIONSHIP BETWEEN HEMATOLOGIC INFLAMMATORY MARKERS AND GLYCATED HEMOGLOBIN IN PATIENTS WITH TYPE 2 DM ATTENDING A TERTIARY TEACHING HOSPITAL IN NIGERIA

Lemchukwu Amaeshi¹; Olufunto Kalejaiye²; Oluwarotimi Olopade²; Michael Kehinde²

¹Internal Medicine, Montefiore Health System, Bronx, NY; ²Lagos University Teaching Hospital, Surulere, Lagos, Nigeria. (Control ID #3876309)

BACKGROUND: The pathophysiologic hallmark of Type 2 DM is insulin resistance, and sub-clinical inflammation is involved in its pathogenesis. Inflammatory markers such as neutrophil-lymphocyte ratio (NLR), platelet-lymphocyte ratio (PLR), and mean platelet volume (MPV) are elevated in Type 2 DM. Studies have shown that glycated hemoglobin (HbA1c), a standard measure of glycemic control, correlates with these markers. In low-resource countries, where the burden of Type 2DM is increasing, and many cannot afford HbA1c as a measure of glycemic control, a less expensive alternative is necessary. Complete blood count (CBC), which has the indices from which these hematologic inflammatory markers are derived, are easily available and affordable even for the poor rural population. This study aimed to assess the role of NLR, PLR, and MPV as surrogates of HbA1c in predicting glycemic control in patients with Type 2 DM.

METHODS: This cross-sectional study comprised 109 adults with Type 2 DM and 109 age-and sex-matched controls. Convenience sampling was used to recruit the participants. Adult patients (>18 years) with diabetes were recruited from the diabetic clinic of a large teaching hospital in Lagos, Nigeria. The controls were healthy adults without diabetes. The NLR, PLR, and MPV were derived from the CBC of the participants. Those with Type 2 DM were subdivided in terms of their glycemic control, where optimal glycemic control comprised those with Hb A1c <7% and suboptimal glycemic control had Hb A1c > 7%. Data were analyzed using the Statistical Package for Social Sciences software. Categorical and continuous variables were compared using the chi-square test and student t-test, respectively. The strength of the association between Hb A1c and the inflammatory markers were assessed using Spearman’s rank correlation coefficient and scatter plot charts. A P-value of <0.05 was considered statistically significant.

RESULTS: The mean age of the participants was 57 + 11 years. Ninety-five percent of the patients with diabetes were on glucose-lowering medications. Patients with Type 2 DM had significantly higher NLR compared with controls [(6.2 + 1.8) vs. (4.9 + 1.2) x 10⁹ cells/L, P<0.001]. PLR and MPV were not statistically higher among Type 2DM patients compared with controls. In patients with Type 2DM, there was no statistically significant difference in NLR, PLR, and MPV between those with good glycemic control and those with suboptimal glycemic control. In addition, there was no correlation between these markers and Hb A1c.

CONCLUSIONS: Neutrophil-lymphocyte-ratio is elevated in Type 2 DM, but there is no relationship between NLR, PLR, MPV, and HbA1c. The medications that patients with diabetes take, may have anti- inflammatory properties that reduce the levels of these inflammatory markers. Therefore, these markers may not be helpful in predicting glycemic control in the management of patients with diabetes. A larger study may show an association between Hb A1c and these markers.

RISK FACTORS FOR INPATIENT MORTALITY IN PATIENTS WITH END STAGE KIDNEY DISEASE ON MAINTENANCE DIALYSIS, DURING AN ADMISSION FOR COVID-19 USING A NATIONALLY REPRESENTATIVE SAMPLE

Sae X. Morita, Akshay Agrawal, Bessy S. Flores Chang, Carlos E. Arias Morales

Medicine, City University of New York, New York, NY. (Control ID #3876961)

BACKGROUND: Several studies reported that the COVID-19 pandemic has disproportionately affected socioeconomically vulnerable populations. No study evaluated risk factors for inpatient mortality in patients with end stage kidney disease (ESKD) on maintenance dialysis admitted with the diagnosis of Covid-19, using a US nationwide sample.

METHODS: We extracted data from the National Inpatient Sample year 2020, the largest nationally representative inpatient-care database in the US. We created a cohort of all adults with ESKD on maintenance dialysis hospitalized for Covid-19 by ICD-10 codes. The outcome was inpatient mortality. We excluded transplant patients. We compared baseline characteristics and used multiple logistic regression to determine risk factors for inpatient mortality in this cohort.

RESULTS: A total of 789,855 admissions with a diagnosis of Covid-19 were detected in patients with ESKD on maintenance dialysis. The mean age was 62.1±0.2 years and Females accounted for 46%. Inpatient mortality was 5.9%. The table showed the results of the multivariable logistic model. Older patients, Males, with higher Charlson Comorbidity Index scores were associated with a higher risk of inpatient mortality. White patients had a higher risk compared to Black patients and among ethnicities/races, Native Americans had the highest risk, although not statistically significant. Those with commercial payments had a higher risk than those with Medicare and Medicaid payments. Admission to a larger, urban, teaching, publicly owned hospital and those in the Northeast had higher odds of inpatient mortality.

CONCLUSIONS: Our study showed Males, commercial payment type, hospital settings such as regions and teaching status were risk factors for inpatient mortality in patients with ESKD on maintenance dialysis.

Further studies are needed to evaluate whether differences in clinical practice underly some of these risks.

RISKY USE BEHAVIORS IN PATIENTS CERTIFIED FOR MEDICAL CANNABIS FOR CHRONIC PAIN

Yuval Zolotov¹; Deepika Slawek¹; Chenshu Zhang¹; Jonathan Ross¹; Chinazo Cunningham¹; Joanna L. Starrels¹; Julia Arnsten¹; Nancy Sohler²

¹Division of General Internal Medicine, Albert Einstein College of Medicine and Montefiore Medical Center, Bronx, NY; ²City University of New York, New York, NY. (Control ID #3875103)

BACKGROUND: Although use of medical cannabis is increasing, data on risky use behaviors among patients certified for medical cannabis is limited. We explored several risky use behaviors in a cohort of patients who were newly certified for medical cannabis for pain in New York State.

METHODS: We analyzed data from the Medical Marijuana and Opioids study, which enrolled patients with chronic or severe pain who reported opioid use and were newly certified for medical cannabis. We included data from 170 participants who completed nine months of follow-up. At baseline and every three months, we assessed: (1) DSM-IV Cannabis Abuse and Cannabis Dependence, measured by The Mini-International Neuropsychiatric Interview; (2) medical cannabis diversion, measured by a modified version of the Massachusetts General Hospital Medication Questionnaire; (3) medical cannabis misuse potential, determined using two items from the Current Opioid Misuse Measure. Unregulated cannabis use was measured by self-report in web surveys every two weeks. Participants were categorized with a risky use behavior if it was reported at any time during follow-up. Days of medical cannabis use, measured in the web surveys, was the average number of days of medical cannabis use during a two-week period. We calculated frequencies and used logistic regressions to determine the effect of the frequency of medical cannabis use on each behavior, adjusting for demographic characteristics.

RESULTS: The median age was 54 years and 92 participants (54.1%) identified as female, 69 (40.6%) as non-Hispanic White, 50 (29.4%) as non-Hispanic Black, and 40 (23.5%) as Hispanic. Participants reported medical cannabis use on an average of 7.3 days every two weeks. Fourteen participants (8.2%) met the criteria for cannabis abuse or dependence, 6 (3.5%) reported medical cannabis diversion, 95 (59.0%) reported potential for medical cannabis misuse, and 68 (40.0%) reported unregulated cannabis use. Adjusted for age, insurance status, and pain severity, frequency of medical cannabis use was associated with lower odds of unregulated cannabis use (OR 0.91, 95% CI 0.85, 0.98). No other associations reached statistical significance.

CONCLUSIONS: In this longitudinal study of patients newly certified for medical cannabis with chronic pain, we observed low frequencies of cannabis abuse or dependence and medical cannabis diversion over nine months. More participants reported potential for medical cannabis misuse and unregulated cannabis use, and those with more frequent medical cannabis use were less likely to report unregulated cannabis use. These results indicate that the availability of medical cannabis alone may not prevent risky use behaviors. A limitation of our study is that the study measures were developed for studies of people who use unregulated cannabis or prescription opioids and may not be valid for medical cannabis studies.

SELF-REPORTED FRAILTY AND HEALTHCARE UTILIZATION IN COMMUNITY-DWELLING MIDDLE-AGED AND OLDER ADULTS IN THE UNITED STATES

Aaron Yao¹; Shengyu Zhou²; Joyce Cheng³; Dae Hyun Kim⁴

¹VillageMD, Chicago, IL; ²Center for Management and Policy Research, Shandong University, Jinan, Shandong, China; ³School of Medicine, Johns Hopkins University School of Medicine, Baltimore, MD; ⁴Harvard Medical School, Boston, MA. (Control ID #3872013)

BACKGROUND: The aging of the population has increased the prevalence of frailty substantially, especially among the costliest patients; however, the current healthcare response to frailty is mainly reactive and acute care-based. And previous studies on frailty have mainly been conducted using nonrepresentative samples in people aged 65 years and above. With a nationally representative sample, we describe the prevalence of frailty in the United States and evaluate the associations between frailty, demographic characteristics, and healthcare utilization in community-dwelling middle-aged and older adults. And these results are discussed to inform policymaking and quality improvements in primary care settings.

METHODS: This study included 2019 National Health Interview Survey demographic and healthcare utilization data for about 20000 participants. Frailty was measured using the revised FRAIL scale (Fatigue, Resistance, Ambulation, Illnesses, and Low BMI) and five survey questions in the National Health Interview Survey questionnaire. Individuals were classified into different health status categories: robust, pre-frail, or frail. About healthcare utilization, emergency room usage was determined by asking participants, "During the past 12 months, how many times have you gone to a hospital emergency room about your health?" Similarly, overnight hospitalization, delay of care due to cost, and omission of medical care due to cost were also measured using self-reported data. Rao-Scott chi-squared tests and logistic regression analyses were used to analyze relationships between age, frailty, family income, educational attainment, and healthcare utilization.

RESULTS: The weighted prevalence of frailty was 39% in those aged 85 or older, 25% in those aged 75-84 years, 15% in those aged 65-74 years, and 9% in those aged 45-64 years. Frailty prevalence was two times higher in older adults and eleven times higher in middle-aged adults earning less than the federal poverty line (FPL) compared to those earning ≥400% FPL. Frailty was associated with increased emergency room visits and hospitalizations, and delay and omission of medical care due to cost. Lower family income and educational attainment exacerbated the associations between frailty and delay and omission of medical care. These trends appeared stronger in middle aged adults compared to older adults.

CONCLUSIONS: Frailty is prevalent in both middle-aged and older adults; however, delays and omissions of medical care were more prevalent in middle-aged adults. Primary care providers can play an important role in identifying frailty early and addressing both over-use and under-use of health care services.

SPINAL CORD INJURY AND PRESCRIBED OPIOIDS FOR PAIN: A SCOPING REVIEW

Jo Ann Shoup¹; JoEllen Welter^2,3; Ingrid A. Binswanger^1,4; Florian Hess²; Alexander Dullenkopf³; Jennifer co*ker⁵; Jeffrey Berliner^5,6

¹Institute for Health Research, Kaiser Permanente Colorado Institute for Health Research, Aurora, CO; ²Department of Orthopedic Surgery and Traumatology, Spital Thurgau AG, Frauenfeld, Thurgau, Switzerland; ³Institute for Anesthesia and Intensive Care Medicine, Spital Thurgau AG, Frauenfeld, Thurgau, Switzerland; ⁴Division of General Internal Medicine, University of Colorado, Denver, CO; ⁵Craig Hospital Research Department, Craig Hospital, Englewood, CO; ⁶CNS Medical Group PC, Englewood, CO. (Control ID #3874669)

BACKGROUND: Spinal cord injury (SCI) is a life-altering neurological condition affecting physical and psychosocial functioning. Individuals with traumatic and nontraumatic SCI experience very high rates of pain. An estimated 60% to 80% of people with SCI report high rates of chronic pain, with one-third experiencing intense pain. Effective treatment options for SCI pain management beyond pharmaceuticals are limited, resulting in elevated exposure to prescribed opioids. A scoping review was conducted to identify and synthesize literature on traumatic and nontraumatic SCI and prescription opioid use for pain. We focused on the following research questions: (i) what type of evidence on this topic is available in the peer-reviewed published literature, (ii) what research gaps exist, and (iii) what recommendations and priorities for future research and clinical practice have been or should be proposed.

METHODS: We searched six electronic bibliographic databases (PubMed (MEDLINE), Ovid (MEDLINE), EMBASE, Cochrane Library, CINAHL, PsychNET) for articles published from 2014 through 2021. Terms for “spinal cord injury” and “prescription opioid use” were used. Included articles were in English and peer reviewed. Data were extracted using an electronic database by two independent reviewers. Opioid use risk factors for chronic SCI were identified. An analysis was performed to identify gaps in knowledge on SCI and prescription opioids.

RESULTS: Of the 347 publications screened for eligibility, 16 articles included in the scoping review. A majority were conducted in the United States (n=9). Most articles lacked information on income (87.5%), ethnicity (87.5%), and race (75%). In studies reporting prescription opioid use, the range was from 35% to 60%. Descriptively, polypharmacy of benzodiazepine, sedative or hypnotic prescriptions co-occurred with opioid prescriptions in up to 38% of participants. Risk factors for prescription opioid use in those with chronic SCI included middle age, lower income, osteoarthritis diagnosis, prior opioid use, and lower-level spinal injury. The three primary gaps in the literature were: lack of reporting of diversity such as race, ethnicity, and age in study populations and settings; absence of polypharmacy as a variable in statistical modeling in prescription opioid risk; lack of knowledge of opioid-related overdose and death in the SCI population; and limited high quality research designs and methods.

CONCLUSIONS: In a population with high rates of pain and high exposure to prescription opioids and other pharmaceutical treatments, knowledge is needed to understand needs and potential effectiveness of treating SCI-related pain. Future research should address gaps in the literature such as reporting demographics of the research population, inclusion of polypharmacy in prescription opioid risk modeling, reporting opioid-related overdose and death in those with SCI and implementing clinical trials on pain management and harm reduction.

SYMPTOMS DESCRIPTION AND SEVERITY OF LONG COVID

Kelly Bell¹; Raymond Van Cleve²; Robert Petrin¹; Alan Roshwalb¹; Megan Shaheen¹; Amanda Purnell²

Learning, Veterans Health Administration, Washington, DC. (Control ID #3876548)

BACKGROUND: Post Acute Sequelae of COVID-19, also known as Long Haul COVID-19 or simply Long COVID, has proven to be a significant threat to public health. One of the current challenges with Long COVID is that we do not know exactly how to classify the disease beyond COVID-19 related symptoms that lasted more than four weeks after the initial infection. This paper presents results from a specialty sample survey of US Veterans, which provides insight into COVID-19 symptom incidence rates for twenty-two acute COVID and long COVID symptoms, as well as levels of overall symptom severity for the Veteran population. This is a descriptive study to examine the nature of Veterans’ experience with Acute COVID and Long COVID.

METHODS: This study used KnowledgePanel®, a probability-based online panel, to survey 3340 Veterans. This survey asked questions regarding demographic factors, social determinants of health, their health care use, health history, and their experience with COVID-19. Specifically, this survey asked which symptoms related to COVID-19 each person experienced and the severity of those symptoms.

RESULTS: Of the 3340 Veterans who had COVID-19, 658 Veterans had tested positive for COVID-19 and 303 experienced symptoms for longer than four weeks. 245 Veterans had slight or moderate symptoms in the long period, while 68 veterans had severe or debilitating symptoms in the long period. The most common symptoms experienced in the long period were fatigue, difficulty staying asleep, body aches, general pain, and coughing. People who had debilitating or severe symptoms in long period did not necessarily have debilitating or severe symptoms in acute period.

CONCLUSIONS: The symptoms experienced ranged from psychological, to respiratory, to broader physiological. Long COVID symptoms span a wide range of medical areas, and some symptoms could become move from mild to severe. People who test positive for COVID have a low but significant chance of developing severe symptoms many weeks after initially testing positive.

THE 12-MONTH COST-EFFECTIVENESS OF TELEPHONE-DELIVERED "BLENDED" COLLABORATIVE CARE FOR TREATING HEART FAILURE AND COMORBID DEPRESSION Bruce L. Rollman¹; John F. Dickerson²; Scott D. Rothenberger¹; Amy Anderson¹; Kwonho Jeong³; Kaleab Z. Abebe¹; Maureen O'Keeffe-Rosetti²; Lucy Fulton²; Bea H. Belnap¹; Frances L. Lynch²

¹Medicine, University of Pittsburgh, Pittsburgh, PA; ²Kaiser Permanente Center for Health Research Northwest Region, Portland, OR; ³Biostatistics, University of Pittsburgh, Pittsburgh, PA. (Control ID #3868941)

BACKGROUND: Depression is co-morbid in 20-25% of patients with heart failure (HF), associated with poorer clinical outcomes, and often goes unrecognized and untreated. We previously reported telephone- delivered collaborative care (CC) for treating HF and depression together improves 12-month health-related quality of life and mood symptoms more than primary care physicians’ (PCPs’) usual care (UC). We now report the cost-effectiveness of our intervention so as to guide health systems on whether to adopt our “blended” CC strategy.

METHODS: From 3/14-10/17 we enrolled 629 hospitalized patients with systolic HF (ejection fraction ≤45%) who screened positive for depression on the 2-item Patient Health Questionnaire (PHQ-2); had at least a moderate level of mood symptoms two weeks after hospitalization (PHQ-9 ≥10); and randomized them to either 12-months of nurse-provided “blended” CC, CC for HF-alone, or to their PCPs’ UC. Later, we obtained 12-months continuous enrollment insurance claims data from Medicare and the two largest health insurers in our region on 503 (80%). We then applied 2016 Medicare prices to merge claims datasets,

approximated our interventions’ costs to estimate incremental costs from the payor’s perspective, used generalized linear models with gamma distribution and log link to correct for skewness in costs, and conducted a sensitivity analysis removing the 1% most costly patients. We also converted baseline, 3-, 6-, and adjusted life year (QALY) gained at 12 months.

RESULTS: At baseline, the 503 patients with continuous 12-month claims data were similar by sociodemographic and clinical characteristics (58% male, 65.3 mean age, 77.5% White, PHQ-9 14.0 (SD: 3.6)) to the 126 excluded from our analyses due to incomplete claims data. At 12-months following randomization, “blended” CC, CC for HF-alone, and UC all had similar mean total claims costs of $285,499, $306,727, and $322,359 respectively, after adjusting for gender, hospital recruitment type (university, community, underserved), and total claims for the 12 month period before randomization, or removing the top 1% (five) most costly patients in a sensitivity analysis (12-month costs >$2.7M). Notably, rehospitalizations costs accounted for most expenditures ($337,203 “blended” ($310,403 excluding top 1%), $392,084 ($351,552), HF-only, and $412,209 ($374,031) UC, respectively). Blended CC yielded incremental cost effectiveness ratios (ICER) per QALY of -$798,276 (95% CI: -$31.8M to +$4.9M) versus UC, and -$1.9M (-55.5M to +5.0M) versus HF-alone. A bootstrapped cost-effectiveness plane also demonstrated a 61% probability of blended CC "dominating" UC, and an 80% probability of blended CC “dominating” HF- alone (more QALYs at lower cost).

CONCLUSIONS: Telephone-delivered blended CC for treating HF and depression is a quality-improving and cost-effective treatment that meets generally accepted criteria for high-value care.

THE ASSOCIATIONS OF VITAMIN D AND PRO-INFLAMMATORY CYTOKINES IN WEST TEXAS PATIENTS WITH RHEUMATOID ARTHRITIS: A PILOT STUDY

Kristina M. Cross¹; Kushal Gandhi¹; Asley Sanchez¹; John Garza²; Srikanth Mukkera¹; Swapna Kolli¹; Hema Kondakindi¹; Barath Rangaswamy¹

¹School of Medicine, Texas Tech University Health Sciences Center School of Medicine, Lubbock, TX; ²Mathematics, The University of Texas Permian Basin, Midland, TX. (Control ID #3872077)

BACKGROUND: Rheumatoid arthritis (RA) is a systemic autoimmune multifactorial complex disease primarily presenting as pain and inflammation in joints affecting about 1.5 million Americans. Disease progression and pathogenesis in RA is directly related to pro-inflammatory cytokine levels, and the systemic increase in pro-inflammatory cytokines is believed to be correlated to immunomodulators. Vitamin D (VD) is a known immunomodulator in innate and adaptive immunity. This study seeks to provide the novel quantification of VD measurements and pro-inflammatory cytokine levels in West Texas patients with RA.

METHODS: Adult participants with BMI <30 and taking less than 1000 IU/day of VD were separated into a control group (n = 40) and a RA group (n = 40). Participants had blood samples collected, completed a Rapid 3 assessment, and submitted a demographics survey. RA patients were excluded if taking prednisone or methotrexate. Serum samples were analyzed for 25-hydroxy VD level using multiplex electrochemiluminescence assay and 10 cytokines were assessed: Interferon gamma (IFN-γ), Interleukin (IL)-1β, IL-2, IL-4, IL-6, IL-8, IL-10, IL-12p70, IL-13, and tumor necrosis factor alpha (TNF-α). Patients were stratified into sufficient (≥ 30 ng/mL), insufficient (20 ng/mL ≤ VD < 30 ng/mL), or deficient (< 20 ng/mL) VD levels.

RESULTS: The association of RA and the distribution of cytokines was observed to be very large (η² = 0.31) and highly significant with modified permutational MANOVA resulting in p < 0.001. Additionally, nine of the ten cytokines had statistically significant higher levels in the RA group with large effect sizes observed for TNF-α, IL-10, and IL-6, moderate effect sizes observed for INF-γ, IL-12, and IL-8, and small effect sizes observed for IL-4, IL-1β, and IL-2. After stratifying patients as sufficient, insufficient, or deficient VD level, the association of VD level on cytokine distributions was observed to be small (η² = 0.07) and statistically insignificant (p = 0.882). However, using Spearman correlation, small but statistically significant positive associations with VD level in IL-10 (ρ_s = +0.338, p = 0.002), IL-1β (ρ_s = +0.287, p = 0.010), and TNF-α (ρ_s = +0.222, p = 0.048) were observed.

CONCLUSIONS: As expected, the RA group had higher proinflammatory cytokines most likely due to the primary mechanism of RA disease progression being increased systemic inflammation. Although much research has shown that VD is immunomodulatory in inflammatory diseases, our results showed that VD levels across groups has a comparatively small effect on cytokine levels in RA patients with only IL-10, IL-1β, and TNF-α having small positive associations with VD level. Based on these results, the routine practice of VD use in RA may not be an effective method of immunomodulation.

THE EFFECTS OF AIR QUALITY ON ASTHMA HOSPITALIZATION RATES IN CALIFORNIA COUNTIES

Sebastian Cano-Besquet, Mei Liang, Karina Smolyar, Jeff Chen, Colton Rosendahl, Anmol Shrestha, Ashley Choi, Nicholas Eremita, Vinh Nguyen, Hani Atamna, Shabana Masood, Joseph Dhahbi

School of Medicine, California University of Science and Medicine, Colton, CA. (Control ID #3853728)

BACKGROUND: Asthma is a chronic lung disease that affects nearly 25 million people in the United States. Common symptoms include coughing, chest tightness, wheezing, and shortness of breath due to oxidative injury to the airways. Outdoor air pollution can exacerbate asthma conditions. The U.S. Environmental Protection Agency (EPA)’s air quality index (AQI) accounts for five major air pollutants including ground-level ozone.

Through our study, we explore whether air quality index and ground-level ozone can each predict asthma hospitalization rates in counties within California. We will consider the interconnections among these variables as well as the broader health impact.

METHODS: Data sources:

1. Open access data from the U.S. EPA and associated air quality data for California counties from 2014-2018: https://aqs.epa.gov/aqsweb/airdata/download_files.html#Meta

2. Open access data from California Health and Human Services (CHHS), Asthma Hospitalization Rates for California (CA) counties from 2014-2018: https://data.chhs.ca.gov/dataset/asthma-hospitalization-rates-by- county

Statistical procedures in brief: Asthma Hospitalization Rates across the years 2014-2018 were averaged for each CA county, and for CA statewide. Median AQI and Days Ozone for 2014-2018 were separately averaged for each CA county, except for counties with missing data. Simple linear regression was performed on (1) Asthma Hospitalization Rates versus Median AQIs and (2) Asthma Hospitalization Rates versus Days Ozone. p-values were calculated using a t-distribution and a two-sided alternative hypothesis.

RESULTS: Using simple linear regression, the study revealed a moderate correlation (r = 0.399, adjusted R-squared = 0.16, p-value = 0.002) between median AQI and asthma-related hospitalization rates in CA counties with nonzero asthma hospitalizations. Our linear models predict that with every 18.18 unit increase in median AQI, there is 1 additional hospitalization per 10,000 people. The study also found a moderate relationship (r = 0.434, adjusted R-squared = 0.19, p-value = 0.001) between the days of ozone exceedances per year and asthma-related hospitalization rates. For every 100 additional days per year of ozone exceedance, there is 1 additional hospitalization per 10,000 people.

CONCLUSIONS: Our study shows moderate correlations between asthma hospitalization rates and (1) median AQI and (2) days of ozone exceedance per year. However, future investigations might include patients' age, social economic status, and medication list. Other confounding variables such as geography, agriculture, and industries in each California county can also affect asthma hospitalizations.

Understanding the connections between environmental conditions and asthma symptoms will enable physicians and public health officials to prepare for asthma exacerbations and hospitalizations before they occur. Furthermore, proper health education and action among community members can allow medical personnel to divert resources and improve patient outcomes.

THE LONG-TERM RELATIONSHIP BETWEEN α-KLOTHO WITH ALL-CAUSE AND CAUSE- SPECIFIC MORTALITY

Nargiza Kurbanova, Cynthia Avila, Waleed Kassabo, Ayeman Basith, Hannah Chung, Simrah Siddiqui, Rehan Qayyum

Internal Medicine, Eastern Virginia Medical School, Norfolk, VA. (Control ID #3876452)

BACKGROUND: The relationship between α-Klotho (αK) and all-cause and cause-specific mortality is controversial. We investigated the association of serum αK levels with all-cause and cause-specific mortality in the 2007-16 National Health and Nutrition Examination Survey.

METHODS: Serum αK levels were measured for participants aged 40-79 years. Associations between mortality and αK levels as a continuous variable and in quintiles were examined using Kaplan-Meier (KM) survival and Cox proportional hazard models. To examine the non-linear relationship, we included two variables in models: a continuous αK variable and its squared αK variable. Adjusted models included age, gender, race, hypertension, diabetes, smoking, alcohol use, BMI, serum cholesterol, estimated GFR, educational status, and poverty threshold to income ratio.

RESULTS: Of the 13747 participants, 1569(11%) died, 7092(52%) were female, and 5,918(43%) were Caucasian. Mean (SD) age was 58(11) years, BMI 29.7(6.7) Kg/m2, and αK was 0.85(0.31) ng/mL. Individuals in the lowest quintile had the highest mortality (Figure). The adjusted Cox proportional hazards model found a U-shaped relationship; initially all-cause mortality decreased with increasing αK level, but the rate of decrease slowed until mortality risk started rising (continuous αK HR=0.58, 95%CI=0.38, 0.88; squared-αK HR=1.24, 95%CI=1.10, 1.40). A similar U-shaped relationship was noted between αK and cancer-mortality (continuous αK HR=0.46, 95%CI=0.20, 1.07; squared αK HR=1.31, 95%CI=1.07, 1.61). However, no relationship was noted with cardiovascular- or other-cause mortality. Examining αK as quintiles, we found no difference between the 1st vs. 5th quintiles while the middle three quintiles were significant, confirming a U-shaped relationship (2nd vs. 1st=0.76, 95%CI=0.59, 0.97, 3rd vs. 1st=0.77, 95%CI= 0.62, 95, and 4th vs. 1st=0.72, 95%CI= 0.56, 0.93). No association was found between αK quintiles and cause- specific mortality.

CONCLUSIONS: In this large diverse cohort representative of the US population, we report a U-shaped relationship between αK and all-cause and cancer mortality. Further research elucidating the underlying biological mechanism of the observed relationships is needed.

THE OFFICE-GUIDELINES APPLIED TO PRACTICE OFFICE-GAP PROGRAM: INTERVENTIONS TO IMPROVE DIABETIC PATIENTS' MEDICATION ADHERENCE

Adesuwa B. Olomu¹; Joseph C. Gardiner²; Aaron Nichols¹; Nephertiti Efeovbokhan³; Richard Pellizzari⁵; Raqhuel D. Williams¹; Karen Kelly-Blake¹; William Hart-Davidson⁴

¹College of Human Medicine, Department of Medicine, Michigan State University, East Lansing, MI; ²Epidemiology and Biostatistics, Michigan State University, East Lansing, MI; ³Cardiology, NEA Baptist Memorial Hospital, Jonesboro, AR; ⁴College of Arts & Letters, Michigan State University, East Lansing, MI; ⁵College of Human Medicine, Michigan State University, East Lansing, MI. (Control ID #3876383)

BACKGROUND: Timely and continuous use of prescribed medications is key to effective chronic disease management. Yet as many as half of 187 million patients in the United States do not take medications as prescribed. 21-42% of diabetic (DM) patients are not adherent to their medications. Medication nonadherence in DM patients leads to poor DM and BP control with increased risk for hospitalization, disability, and death. Our objectives were 1) determine the impact of combined patient activation (using our Office Guidelines Applied to Practice {Office-GAP} Program) and evidence-based text messages (Way to Health program) interventions compared to text messages only on improving medication adherence in patients with diabetes mellitus type 2, in Federally Qualified Healthcare Centers (FQHCs); and 2) evaluate the differences by gender of combined Patient Activation + Text messaging versus Text messaging intervention only in improving diabetic patient medication adherence in FQHCs.

METHODS: Patients at 12 FQHCs in Michigan were randomized to two groups: Intervention (Arm 1): Office-GAP + Texting vs Texting only (Arm 2). Office-GAP is a patient activation intervention to improve communication and patient-provider partnerships through brief training in shared decision-making and use of a guideline-based checklist. Way to Health program is a two-way text messaging platform developed at the University of Pennsylvania for health communication interventions. Adherence to one prescribed primary medication was assessed with eCAP electronic monitoring. We report adherence rates for 78 patients who completed 3 months in an ongoing cluster randomized controlled trial. Adherence rate per patient is calculated as a proportion of distinct eCap open dates within the time window of 92 consecutive days. Generalized linear models with Beta regression on the continuous proportion was used to estimate group and gender effects.

RESULTS: Seventy-nine patients completed their 3-month assessment, Arm 1 (n=53, 71.7% female), Arm 2 (n=26, 76.9% female). A total of 5044 eCAP openings were within the time window, 70.6% were in Arm 1 and 29.4% in Arm 2. From Beta regression, the estimated adherence rate (as percent) was 68.4% (95% CI: 61.4-74.8) in Arm 1, and 54.8% (95% CI: 44.5-64.7) in Arm 2. The difference between arms is significant (p=.027). There was no significant difference in adherence by gender: Female 65.2% (95% CI: 58.3-71.5), male 66.0% (95% CI: 55.1-75.5).

CONCLUSIONS: Adherence rate was significantly higher in patients randomized to combined Patient Activation + Text messaging compared to patients receiving only text messaging. Patient activation and mHealth interventions hold promise as a feasible support strategy to improve patient medication adherence in FQHCs.

THE PRESSURE DOESN’T LET UP: A VIRTUALLY-DELIVERED HYPERTENSION PROGRAM DURING COVID

Karen Goldstein^2,1; Allison Lewinski^2,3; Abigail Shapiro²; Tiera Lanford²; Courtney White-Clark²; Madison Burns²; Julie Schexnayder⁴; Leah Zullig^2,5; Jennifer Gierisch^2,5; Hayden Bosworth^2,5

¹Division of General Internal Medicine, Duke University School of Medicine, Durham, NC; ²Durham VA Medical Center, Durham, NC; ³n/a, Duke University School of Nursing, Durham, NC; ⁴The University of Alabama at Birmingham School of Nursing, Birmingham, AL; ⁵Population Health Sciences, Duke University, Durham, NC. (Control ID #3875188)

BACKGROUND: Even prior to COVID-19 hypertension control has been declining, especially among historically marginalized populations. During the pandemic, clinic shutdowns and social distancing further worsened hypertension control. Virtually-delivered chronic disease management programs provided critical support during the pandemic and offer a promising approach to supplement the efforts of busy primary care teams to improve the health of diverse patient populations with hypertension. Our multi-site, quality improvement program, Team-supported, EHR-leveraged, Active Management (TEAM), rolled out over a year prior to COVID-19 and has continued steadily through the evolution of the pandemic. We report the impact of TEAM on hypertension control among a diverse, largely rural Veteran population surrounding the COVID-19 pandemic.

METHODS: TEAM is a population health-based, virtually-delivered hypertension program designed to provide supplemental support to patients with uncontrolled hypertension. Eligible patients were identified via the electronic health record or referred by their primary care provider. Patients were sent a letter with personalized information about their blood pressure (BP) followed with a phone call from the population health manager. The call focused on patient identification of individualized goals and specific action steps around hypertension management via motivational interviewing and ordering of blood pressure cuffs if needed. One to 2 follow-up calls (with or without video-enabled connection) assessed progress towards Veteran’s goals and self-monitoring of home BP measurement. Patient primary care teams received information about patient goals, potential regimen changes, and woman-tailored CVD risk information. We evaluated the impact of our program with a mixed methods approach using quantitative EHR data and participant and staff interviews.

RESULTS: Since 2019, TEAM has been implemented at 16 geographically diverse rural sites and enrolled 917 patients. Of the total enrolled, 119 were enrolled prior to COVID-19. In total, 27.4% of patients are women, 45.6% are Black and 50.7% White. Seventy-six percent live in rural areas. Mean age is 64.9 years (s.d. 9.3) and mean BP at program entry pre-COVID was 150.5/85.6mmHg and post-COVID was 154.3/84.6mmHg. Mean reduction in BP from program entry to completion was 5.5/3mmHg. Mean change in BP for women was 6/2.3mmHg and for Black patients was 4/2.5mmHg. Themes from patient qualitative interviews included appreciation for outreach from the population manager, convenience of remote counseling during COVID-19 restrictions, and improvement in self-management behaviors

CONCLUSIONS: Virtually-delivered hypertension control programs are an effective way to improve blood pressure control in a diverse population of patients and are feasible during pandemic-related conditions despite worse blood pressure control. Next steps are to ensure sustainability during evolving health care and social-environmental stressors.

THREE-MONTH OUTCOMES FOR PATIENTS TREATED IN AN ACADEMIC MEDICAL CENTER LONG COVID CLINIC FOR COGNITIVE SYMPTOMS

Andrew Schamess¹; Kathleen Woschkolup¹⁰; Laura J. Rush⁶; Erin McConnell⁵; Aaron Friedberg³; Taru Saigal³; alexa meara¹¹; Susan Bowman Burpee⁸; Stacy Stanwick⁷; Jennifer Eramo^2,6; Dana Hodges⁹; Vandana Ohri⁷; Jodi Grandominico¹; Ann S. McAlearney⁴

¹Division of General Internal Medicine, The Ohio State University Wexner Medical Center, Columbus, OH; ²Family Medicine, The Ohio State University Wexner Medical Center, Columbus, OH; ³Internal Medicine, The Ohio State University, Bexley, OH; ⁴Family and Community Medicine, The Ohio State University, Columbus, OH; ⁵General Internal Medicine, The Ohio State University Wexner Medical Center, Columbus, OH; ⁶CATALYST, The Ohio State University, Columbus, OH; ⁷General Internal Medicine, The Ohio State University, Columbus, OH; ⁸Physical Medicine and Rehabilitation, The Ohio State University Wexner Medical Center, Columbus, OH; ⁹Neurology , Dianne Collins Neuroscience Institute, Greenville, SC; ¹⁰Neurology, Bon Secours Mercy Health, Cincinnati, OH;¹¹internal medicine, The Ohio State University, Columbus, OH. (Control ID #3873611)

BACKGROUND: Post-Acute Sequelae of COVID-19 (PASC or "Long COVID") is a new chronic multisystem condition resulting from the COVID-19 epidemic and affecting as many as 200 million persons worldwide. Among the symptoms are cognitive slowing and mental fatigue, commonly known as “brain fog.” Specialized clinics have emerged at many academic medical centers to care for PASC patients. There is paucity of evidence about the treatments offered in these clinics or their impacts on patient outcomes. Our PASC clinic is staffed by general internists collaborating with subspecialists and rehabilitation professionals. We report on the use of behavioral, rehabilitative and pharmacologic treatments in the first 70 patients seen for neurocognitive symptoms (NCS) in our clinic, and on change in patient quality of life and functional status (QoL/FS) in the 3 months following initial clinic visit.

METHODS: Electronic medical records were searched for specified ICD-10 codes to identify patients with PASC and NCS who had completed the PROMIS29 QoL/FS inventory at the initial clinic visit (t1) and again 3 months later (t2). Data extracted by manual review and entered into a REDCap database included demographics, medication history, treatment recommendations, and PROMIS29 scores at t1 and t2. The primary outcome was change in any PROMIS29 T-score. Exact 95% confidence intervals (CIs) were calculated by binomial proportion. Comparisons were made using Fisher’s exact test. The study was approved by the University Institutional Review Board.

RESULTS: Of 70 patients identified for this study, 70% were female, 25% were non-white, and 27% had public insurance. Patients received a variety of treatments including workplace accommodations (25%), cognitive rest (21%), cognitive rehabilitation (30%), rehabilitation psychology (25%), and pharmacotherapy (25%). The frequencies and CIs of clinically significant improvement in QoL/FS (at least 5-point increase in PROMIS29 T-score) in each domain were: physical function: 19% (11-30), anxiety: 41% (4-63), depression: 39% (28-51), fatigue: 44% (40-64), sleep disturbance: 42% (30-55), ability to participate in social roles and activities: 16% (13-33). In this small sample, most subgroup comparisons were not statistically significant. However, 50% of patients treated with N-methyl-D-aspartate (NMDA) inhibitors (primarily amantadine) showed improvement in their ability to participate in social roles and activities compared with 18% of those not treated with NMDA inhibitors (2.7 times greater rate of improvement, p=0.033).

CONCLUSIONS: While a substantial number of patients showed clinically significant improvement in quality of life in several PROMIS29 domains after 3 months of treatment, many did not improve. More work is needed to establish effective, evidence-based treatments for PASC patients with NCS. NMDA inhibitors are a promising class of drug for these symptoms and warrant further study.

TRENDS IN THE BURDEN OF RISK FACTORS FOR CHRONIC DISEASE AMONG US ADULTS

Benjamin G. Mittman^1,2; Michael B. Rothberg¹; Phuc H. Le¹

¹Center for Value-Based Care Research, Cleveland Clinic, Cleveland, OH; ²Department of Population and Quantitative Health Sciences, Case Western Reserve University School of Medicine, Cleveland, OH. (Control ID #3874712)

BACKGROUND: Half of US adults suffer from chronic disease. Unhealthy alcohol use, obesity, low physical activity, poor diet, and current smoking are major preventable risk factors. An understanding of the national burden of these risk factors is essential for guiding continued surveillance and intervention. We examined temporal trends in the prevalence of risk factors for chronic disease in US adults.

METHODS: We completed a retrospective analysis of the continuous 1999-2018 National Health and Nutrition Examination Survey, including adults aged 18-79 years old. Five risk factors were defined according to appropriate national guidelines. We estimated the proportion of US adults meeting each criterion and the proportion of people with 0-5 risk factors. We used logistic regression to examine temporal trends and Poisson regression to determine socioeconomic and demographic characteristics associated with greater number of risk factors. We used R 4.2.1 and accounted for survey weights.

RESULTS: The final sample included 54,947 participants. From 1999-2018, the prevalence of current smoking and unhealthy alcohol use decreased significantly, whereas that of obesity increased by 50%. The prevalence of low physical activity decreased significantly from 2007-2018 while that of poor diet decreased from 2005-2012 and then increased again (Figure). Between 2007 and 2018, 20% of people had 0 risk factors, approximately two-thirds had 1-2, and 15% had ≥3. Compared to those without any risk factors, people with ≥1 risk factor were older (mean age 45.5 years vs. 43.3 years) and more likely to be female (51.9% vs. 49.1%) and Black (12.7% vs. 7.9%). In the multivariable model, lower income-poverty ratio and poor self-reported general health were associated with higher rates of risk factors, whereas higher education and private insurance coverage were associated with lower rates.

CONCLUSIONS: From 1999-2018, unhealthy alcohol use, low physical activity, and current smoking became less prevalent, whereas obesity and poor diet (more recently) increased in prevalence. Women, minorities, and people with low socioeconomic status had a higher burden of risk factors for chronic disease. Surveillance and intervention efforts should address the increasing rates of obesity and poor diet and focus on more vulnerable populations.

UNDERDIAGNOSES OF OBESITY/OVERWEIGHT IN ACADEMIC OUTPATIENT SETTING

Anu Satheeshkumar¹; Barath Rangaswamy²; Hina Tariz¹; John Garza³; Kejal Shah¹; Rahul Atodaria¹

¹Internal Medicine, Texas tech university of health sciences - Permian basin, Odessa, TX; ²internal Medicine, Texas Tech University Health Sciences Center, Lubbock, TX; ³Mathematics, The University of Texas Permian Basin, Midland, TX. (Control ID #3877321)

BACKGROUND: Obesity is a growing epidemic in USA. Despite high prevalence, underdiagnosis of obesity/overweight remains a significant problem. Body mass index (BMI) is a reliable predictor of obesity and overweight, but is under acknowledged by physicians. The primary purpose of this study is to examine whether obese or overweight patients receive an obesity or overweight diagnosis and weight-related counseling from their physician at a West Texas academic outpatient Internal Medicine Clinic.

METHODS: This study is a single center retrospective cross sectional analysis of 299 patients with BMI > 25 who visited Internal Medicine Clinic from October 1st 2020 to October 31st 2021. We collected body mass index, diagnosis of obesity and the diagnosis of overweight in the problem list of the office visit note documented by physicians. We evaluated if patients with BMI > 30 had a documented diagnosis of obesity and if patients with BMI from 25 to 30 had a documented diagnosis of overweight. Inclusion Criteria: Both male and female patients of age: 20 - 89 years on October 1, 2020. Exclusion criteria included pregnant patients, amputee, patients with diagnosis of hyperthyroidism, hypothyroidism, heart failure, radiation or chemotherapy treatment, or metastatic cancer prior to Oct 1, 2020.

RESULTS: Among 299 patients, 204 patients were qualified for the diagnosis of Obesity due to their BMI> 30. However only 48.52 % (99 patients) had a diagnosis of Obesity in the problem list. 95 patients were qualified for the diagnosis of Overweight due to their BMI > 25 but less than 30. However, only 3.15 %( 3 patients) of the patients had a diagnosis of overweight in their problem list. Majority of patients with obesity and overweight are not diagnosed in our outpatient clinic setting. Effect sizes (Cramer V) and p values (chi- squared test) for the difference of sample proportions for overweight, obese, and morbidly obese are summarized in table. The data provide strong evidence that overweight and obese patients are under diagnosed and under counseled regarding diet and exercise.

CONCLUSIONS: There is a great need to educate the physicians for early screening and identification of obesity and overweight. If we are not making the diagnosis in the problem list, then we are not acknowledging the disease process and are not subsequently treating this epidemic. Routine screening and accurate diagnosis are among the first steps leading to proper treatment. Underdiagnosing of obesity is an important barrier to treatment initiation

UNDERSTANDING CORE VALUES OF PATIENTS WITH A HISTORY OF OBESITY: CONSIDERATIONS FOR PROVIDERS WHEN TALKING ABOUT GREEN PHYSICAL ACTIVITY

Michelle E. Carfa*gno1; Katherine Tinco3; Andrea Aquines3,2; Coral Aquino3; Allison Squires4,3; Melanie Jay3,2; Stephanie L. Orstad3

¹Medicine, Stony Brook University Renaissance School of Medicine, Stony Brook, NY; ²VA NY Harbor Healthcare System, New York, NY; ³Division of General Internal Medicine and Clinical Innovation, New York University Grossman School of Medicine, New York, NY; ⁴Nursing, New York University, New York, NY. (Control ID #3870259)

BACKGROUND: Physical activity (PA) has long been used to help patients with a history of obesity lose weight, but patients are often unable to maintain that weight loss. PA that takes place in natural environments such as parks, or green PA, presents an opportunity for a cost-effective and scalable intervention design for lower-income communities, while providing a myriad of health benefits. Some physicians recommend and prescribe green PA through nature prescription programs. However, more research is needed to enhance patient-provider conversations about green PA, such as through the consideration of patient values. Using data from a qualitative pilot study, the purpose of this sub-analysis was to explore the interplay of patient values, specifically within the 8 Dimensions of Wellness, to inform future patient-provider interactions about green PA.

METHODS: We recruited 30 individuals who reside in the same urban community, have a history of obesity (BMI >25 prior to weight loss), and have lost at least 3% of their body weight in the last 24 months. Patients were recruited through targeted mailings, physician referrals, in-person recruitment at a partnering community health center, health fairs, and snowball sampling. Semi-structured interviews were conducted in English and Spanish in a “walk-along” format in the local park, audio recorded and professionally translated and transcribed to English. 15 participants who were between the ages of 21-53, majority Latina, and had an average BMI prior to weight loss of 37.8 were included in the sub-analysis. Using a pragmatic lens, analysis was conducted using a combination of inductive values coding and deductive coding based on the 8 Dimensions of Wellness Model. Analytic code memos were synthesized throughout the coding process to uncover how participants valued each of the dimensions of wellness.

RESULTS: Analysis uncovered patient values around four Dimensions of Wellness: Physical, Emotional, Environmental, and Social. The importance of promoting physical and emotional wellness for loved ones (e.g., children) arose in addition to individual physical and emotional wellness. Participants place significant importance on connecting with and helping their local community, with some citing group green PA as a means to help motivate each other to exercise. Patients enjoy the benefits of fresh air, open space, and greenery when spending time/exercising outdoors in parks but note that park and community safety is a principal consideration when being outdoors. The other four Dimensions of Wellness, Financial, Intellectual, Occupational, and Spiritual, were not as prominent in discussion.

CONCLUSIONS: We identified multiple dimensions of wellness that individuals value in general and in the context of green PA. Although these findings are community-specific, future research can focus on surveying a broader population to see if these findings may be generalizable to patient populations like this one to further inform conversations about green PA.

UNDERSTANDING THE CHALLENGES AND COPING STRATEGIES OF PERSONS LIVING WITH NON-COMMUNICABLE DISEASES DURING CLIMATE-INDUCED DISASTERS IN THE US VIRGIN ISLANDS: A MIXED METHODS APPROACH

Saria M. Hassan¹; Stephen Perez²; Michelle Teresa Wiciak¹; LaVerne Ragster³; Tess Richards⁵; Maxine Nunez³; Marcella Nunez-Smith⁴

¹Internal Medicine, Emory University, Atlanta, GA; ²Emory University School of Public Health, Atlanta, GA; ³University of the Virgin Islands, Saint Thomas, Virgin Islands (U.S.); ⁴General Internal Medicine, Yale University, New Haven, CT; ⁵St. Thomas East End Medical Center, St Thomas, Virgin Islands (U.S.). (Control ID #3876864)

BACKGROUND: At least 30% of mortality after severe weather events are due to poorly controlled non- communicable diseases (NCDs). With the advent of climate change the US territories including the US Virgin Islands (USVI) will experience more climate-induced disasters that are bound to worsen territorial health disparities. We sought to understand the experience of persons living with NCDs (PLNCDs) in the setting of climate-induced disasters in the USVI to inform future interventions.

METHODS: We conducted a concurrent mixed methods approach to document the impact of severe weather events on PLNCDs who are patients at a Federally Qualified Health Center in the USVI. Survey and interviews were conducted during the 2022 hurricane season (June-October 2022). Quantitative: we used convenience sampling to conduct a survey of PLNCDs to determine the impact of recent severe weather events on their mental health and document the challenges they experienced in the aftermath of the disaster. Descriptive and summary statistics were produced using SPSS. Qualitative: Purposive sampling was used to recruit 13 respondents with varied age, sex, and chronic disease for in-depth interviews to understand effective coping strategies for managing their NCD. Thematic content analysis was conducted with 3 coders to identify emergent themes. Integration: A joint display is used to integrate the quantitative and qualitative data.

RESULTS: Out of 87 survey respondents , 85% were female with an average age of 52.14 (sd=16.18) and

13% had no health insurance, 23% had 3 or more NCDs, with 30% having high blood pressure and 12% having diabetes. Twenty percent satisfied criteria for mild anxiety, with 8% having moderate to severe anxiety. Eighteen percent had mild depression with 8.6% moderate to severe depression. Despite experiencing a severe hurricane, 39% were not very prepared for another disaster. Thirty percent of respondents had experienced challenges managing their chronic disease during a past disaster. Emergent themes from the qualitative work included: NCDs were often not the biggest priority in the aftermath of a disaster, mental health poses an immense challenge due to the stress and trauma after a storm, there is a need for resources to strengthen the capacity of patients to self-manage their disease, and the importance of comprehensive and reliable information dissemination on NCDs in preparation for severe weather events

CONCLUSIONS: Our work has documented the challenges of PLNCDs in disasters in a vulnerable population. The work has identified areas for intervention at the individual, institutional, and policy level that can improve outcomes for the most vulnerable in the aftermath of a disaster. We will work with our FQHC partners to develop and test interventions at the individual and organizational level in preparation for the next hurricane season.

‘I AM RELUCTANT TO CONTINUE, YET KNOW SHE COULD GO INTO WITHDRAWAL’: A QUALITATIVE ANALYSIS OF PROVIDER REQUESTS FOR OPIOID MANAGEMENT ECONSULTS

Laila Khalid¹; Sharon Rikin²; Dana Watnick⁵; Tiffany Lu³; Ana Valle²; Joanna L. Starrels⁴

¹General Internal Medicine, Montefiore Medical Center, Bronx, NY; ²Internal Medicine, Montefiore Medical Center, Brooklyn, NY; ³Medicine, Montefiore Medical Center, Bronx, NY; ⁴Medicine, Albert Einstein College of Medicine, Montefiore Medical Center, Bronx, NY; ⁵Albert Einstein College of Medicine, Bronx, NY. (Control ID #3876839)

BACKGROUND: Primary care providers lack training in opioid management for chronic pain and access to pain specialists is limited. Electronic consultation (eConsult) offers a platform for providers to request specialist input about individual patients’ care. To inform dissemination of opioid management eConsults at other institutions, we sought to understand provider motivations for requesting an eConsult and to identify which provider questions could be addressed through eConsult without requiring direct specialist assessments.

METHODS: We conducted an inductive thematic analysis of the text of providers’ eConsult requests for opioid management from July 1, 2019 to September 30, 2020. Two researchers independently identified recurrent themes, compared findings and developed an initial codebook. In team discussions (LK, SR, DW, JS), code matrices were used to discuss emergent themes, and to explore themes by eConsult outcome.

RESULTS: Forty eConsults were requested by 29 unique providers. Respondents practiced general internal medicine (42%), family medicine (27%), hematology (20%) gynecology/obstetrics (8%), and geriatrics (3%). We identified four themes. (1) providers were motivated to reduce harm, but approaches varied from seeking reassurance to wanting to hand off opioid management entirely. (2) Providers needed help assessing and diagnosing opioid misuse and use disorder, including under-diagnosing despite describing the elements of misuse or, conversely, diagnosing and treating OUD with little evidence. (3) Providers had difficulty articulating questions, evidenced by absence of a specific question or overly broad requests, e.g. needing help with pain management. (4) Providers had misconceptions about common opioid management practices including short vs. long acting opioids, tapering drug testing urinalysis. Reducing harm and misconceptions were prominent among requests that were deemed to require direct assessments as compared to those that did not require direct assessments by specialists.

CONCLUSIONS: Providers use opioid eConsults to seek support and reassurance for balancing pain management with worries of OUD. However, providers’ lack of ability to assess for opioid use disorder (OUD) or misuse, difficulty articulating questions and miconceptions about opioid management may reduce the efficiency of the econsult model. Future econsult programs should train providers in not only diagnosing opioid misuse but also to articulate questions differently to maximize scarce specialist resources in a timely manner.

Scientific Abstract - Clinical Informatics and Health Information Technology

A DECISION SUPPORT TOOL’S IMPACT ON INTERPRETATION OF AND RESPONSE TO URINE DRUG TESTS FOR PATIENTS PRESCRIBED OPIOID ANALGESICS: A RANDOMIZED VIGNETTE-BASED NATIONAL SURVEY

Marc LaRochelle¹; Radhika Puppala¹; Sarah Kosakowski¹; Shapei Yan¹; Ziming Xuan²; Ricardo Cruz¹; Karen E. Lasser¹; Mari-Lynn Drainoni³; Erin E. Krebs⁴; Daniel P. Alford¹; Jeffrey H. Samet¹; Jane Liebschutz⁵

¹General Internal Medicine, Boston Medical Center, Boston , MA; ²Boston University School of Public Health, Boston, MA; ³Medicine, Boston University School of Medicine, Boston, MA; ⁴Medicine, Minneapolis VA Health Care System, Minneapolis, MN; ⁵Medicine, University of Pittsburgh, Pittsburgh, PA. (Control ID #3874218)

BACKGROUND: Urine drug tests (UDT) are a recommended monitoring tool for patients prescribed opioids. However, UDT are difficult to interpret, and it is unclear how they impact clinical decision-making.

METHODS: We conducted a randomized vignette-based survey. We recruited physicians who completed a national educational safer opioid prescribing program. We randomized participants to one of three UDT conditions: Group 1 (standard report); Group 2 (standard report plus interpretation); and Group 3 (standard report plus interpretation and clinical guidance). We presented 4 vignettes that ranged from low to high concern of opioid misuse. Participants rated their concern for opioid misuse on a 7-point scale (1 = not concerned, 7 = highly concerned), and identified changes to opioid treatment plan in three domains: monitoring strategy, opioid dosage, and referral or treatment for opioid use disorder. We used ANOVA to test differences in concern for opioid misuse and chi-squared tests to identify treatment plan differences by group.

RESULTS: We invited the first 140 physicians meeting enrollment criteria to participate, and 121 (86%) completed the survey. Of the respondents, 56 (46%) identified as female, 67 (55%) completed residency within 5 years or were still residents, and 36 (30%) completed residency more than 10 years ago. The median (IQR) concern for misuse for the four vignettes was 2 (1,4), 4 (3,5), 5 (4,6), and 6 (5,7). Concern for misuse varied by Group in only the vignette with highest concern for misuse. Compared to Group 1, Group 2 chose different treatment plan changes in 5 of 12 decisions, each consistent with clinical guidance (p<0.05; Table). We did not identify differences in treatment plans between Groups 3 and 2.

CONCLUSIONS: In this randomized vignette survey, physicians provided UDT interpretation modified opioid treatment plans to be more consistent with clinical guidance. However, providing clinical guidance in addition to UDT interpretation did not influence clinical decision-making.

A MACHINE LEARNING APPROACH TO PREDICT ENGAGEMENT FOR A DIGITAL DIABETES PREVENTION PROGRAM

Danissa V. Rodriguez¹; Ji Chen¹; Ratnalekha V. Viswanadham²; Katharine Lawrence³; Devin Mann³

¹Population Health , NYU Langone Health, New York, NY; ²Pediatrics, NYU Langone Health, New York, NY; ³Population Health, NYU Langone Health, New York, NY. (Control ID #3874291)

BACKGROUND: Digital diabetes prevention programs (dDPPs) through mobile applications have been used as “digital prescriptions” within healthcare delivery systems. Despite their effectiveness, dDPPs suffer from high attrition rates and non-completion. Previous research focuses on the development of PAMS, a personalized automatic messaging system that leverages SMS and data integration into clinical workflows to help enhance patient-provider communication and increase engagement with a dDPP. Although preliminary data for PAMS show positive results, we propose ML-PAMS, a machine learning (ML) approach to predict individual engagement for a more personalized content experience to develop an optimized messaging system. This abstract describes our methods for building this ML model and our preliminary results.

METHODS: We partnered with a dDPP app to obtain daily engagement data from 2384 users (704,242 engagement records) external to the original PAMS study who did not receive personalized messaging. We aim to predict next-day engagement in four individual activities (steps, exercise, meal logs, and weigh-ins) and general activity (engagement in any activity) based on previous short-term and long-term activity in the application. We use the gradient-boosted forest model. The data was split into a 70% training set, a 15% validation set, and a 15% test set. the models were trained with η= 0.1 for 1000 rounds with early stopping. Area under ROC (AUROC) and Area under Precision-Recall curve (AUPRC) were used to measure model performance. Shapley values were calculated to reflect feature importance of the models.

RESULTS: The model accurately predicted future engagement and obtained similar results when predicting any of the four activities, and each individual activity (Figure 1.A). Using Shapley values (Figure 1.B), we found that the short-term frequency of various activities was the most informative feature for all five models. However, it could mean that the model is vulnerable to short-term disruption of user behavioral patterns. Additional social factors have the potential to better capture short-term pattern changes in advance and mitigate this issue.

CONCLUSIONS: ML approaches that allow engagement prediction represent an acceptable mechanism to improve messaging interventions to support patient adherence to digital prescriptions. Next, understanding user engagement phenotypes based on their predicted engagement; we can then develop a messaging intervention based on our findings.

AN EVALUATION OF DIAGNOSTIC ERRORS IN PRIMARY CARE TELEMEDICINE VISITS FOR ABDOMINAL PAIN

Jorge A. Rodriguez, David W. Bates, Anuj Dalal

General Internal Medicine, Brigham and Women's Hospital, Boston, MA. (Control ID #3876921)

BACKGROUND: The risk of diagnostic error (DE) is potentially heightened in telemedicine. Telemedicine allows synchronous video and audio-only visits. While telemedicine can improve care access, it also introduces challenges to a timely and accurate diagnosis. The goal of our study was to explore and characterize DEs after telemedicine visits for abdominal pain.

METHODS: We evaluated an enriched cohort of adult patients presenting to primary clinics at an academic medical setting in Boston, MA from March 2020 to February 2022. The cohort included patients who were likely to have DE based on presentation to the emergency department or inpatient setting for “can’t miss” abdominal pain diagnoses (e.g., appendicitis) within the 30 days after an index telemedicine visit. We performed a chart review using validated tools: Safer Dx instrument to assess the likelihood diagnostic error, and the DEER Taxonomy to identify diagnostic process failures. Two clinicians reviewed the charts, then met to adjudicate outcomes. All discrepancies were reconciled by a third clinician. We used a t-test to compare the mean number of diagnostic process failures in cases with DEs compared to those without.

RESULTS: The enriched cohort include 40 cases (29 audio-only, 11 video visits). Patients had a mean age of 60, were 69.4% female, and 66.6% White, 16.7% Black, 5.6% Hispanic, and 11.1% were of Other/Unknown race/ethnicity. We found a DE in 8 cases (20 %) (Table). Of these cases, 3 were harmful/potentially harmful. The missed diagnoses for these cases were: ischemic colitis and pancolitis. Diagnostic process failures were more likely in cases with vs without a DE. (6.3 vs. 1.3, p < 0.001). The most common significant diagnostic failure points in cases with a DE were: failure/delay in considering correct diagnosis (5/8, 62.5%), suboptimal weighing of a piece of history (3/8, 37.5%), failure to perform physical exam (3/8, 37.5%%), suboptimal weighing/prioritizing of diagnoses (3/8, 37.5%), and too much weight given to lower probability diagnosis (3/8 , 37.5%).

CONCLUSIONS: We demonstrate the challenges of evaluating abdominal pain using telemedicine, a difficult condition for this modality. While the lack of physical exam is cited as a limitation of telemedicine, our findings highlight additional diagnostic challenges (e.g., history-taking). As telemedicine use increases, we will need to develop best practice for diagnostic safety.

A NOVEL RISK PREDICTION ALGORITHM FOR ELDERLY PATIENTS AT HIGH RISK OF MEDICATION-RELATED ADVERSE EVENTS

Michelle Frits²; Diane L. Seger³; Christine Iannaccone¹; Aqsa Mugal¹; Frank Chang³; Lynn A. Volk⁴; Stuart R. Lipsitz⁵; Lisa Rotenstein⁶

¹General Internal Medicine, Brigham and Women's Hospital, Boston, MA; ²General Internal Medicine, Brigham and Women's Hospital, Somerville, MA; ³IS- Clinical and Quality Analysis , Partners Healthcare System, Wellesley, MA; ⁴Clinical and Quality Analysis, Partners HealthCare, Wellesley, MA; ⁵Division of General Internal Medicine, Brigham and Women's Hospital, Boston, MA; ⁶Medicine, Brigham and Women's Hospital, Boston, MA. (Control ID #3875950)

BACKGROUND: Adverse drug events contribute to 10-30% of hospitalizations in the elderly. Leveraging medication related information in risk stratification algorithms can facilitate identification of the patients at highest risk for preventable utilization and medication regimen modification. We evaluated the ability of the FeelBetter machine learning algorithm to identify elderly, multimorbid patients cared for by Brigham and Women’s Hospital (BWH) who were at highest risk of emergency department (ED) visits and hospitalizations.

METHODS: The FeelBetter algorithm uses patient medications, demographics, social history, diagnoses, lab results, procedures, allergies, ED and hospitalization patterns, and costs to stratify patients’ risk of ED visits and hospitalizations in the subsequent 1, 3, and 6 months. The algorithm was trained using data from BWH patients ages ≥65 with at least 1 chronic disease and ≥3 active medications as of 7/1/2016.

After algorithm training, patients were assigned 1 of 20 risk-levels based on risk percentile of either ED visits or hospitalizations as of 7/1/2019. These levels were then divided into risk quintiles (80%–100%, 60%–<80%, 40%–<60%, 20%–<40%, 0%–<20%).

We summarized demographics, mean problem list diagnoses, and mean number of medications for patients in each risk quintile and quantified their average ED visits and hospitalizations at 1-, 3-, and 6-months post algorithm prediction. We used an interval variable representing the 20 risk levels in two Poisson regression models to quantify increases in ED visits or hospitalizations uniquely predicted by the FeelBetter risk-levels over other covariates used to predict healthcare utilization (age, sex, medication count, and problem count).

RESULTS: The 108,817 patients had a mean (SD) age of 77.1 (6.6) years, a mean of 15.3 (9.3) problem list diagnoses and were taking an average of 9.0 (6.4) medications. Mean (SD) utilization at 6 months was .10 (.44) for ED visits and .09 (.40) for hospitalizations. Each risk quintile had distinct patterns of ED visits and hospitalizations, with patients in the 80-100% quintile having the highest mean ED and hospitalization rates at 1, 3, and 6 months.

In the Poisson model including the above covariates predicting 6-month utilization, each 1 of 20 increase in the patient’s FeelBetter risk level predicted a 14.2% increase in ED visits. In the hospitalization model, each 1 of 20 increase in the risk-level was associated with an 18.4% increase in hospitalizations. The top 5% of patients identified by FeelBetter were predicted to average 3.7 times the ED visits of the overall cohort and 4.1 times the admissions.

CONCLUSIONS: A novel machine learning algorithm that incorporates information about elderly, multimorbid patients’ medication regimens effectively stratifies patients’ risk of ED use and hospitalizations. Use of this model can help value-focused healthcare systems focus medication optimization efforts and facilitate avoidance of unnecessary healthcare utilization.

A QUALITATIVE ASSESSMENT OF A DIGITAL HEALTH NAVIGATOR PROGRAM IN A PRIMARY CARE CLINIC

Jorge A. Rodriguez¹; Angela Rui¹; Ronen Rozenblum¹; Lipika Samal²

¹General Internal Medicine, Brigham and Women's Hospital, Boston, MA; ²Division of General Internal Medicine, Brigham and Women's Hospital, Boston, MA. (Control ID #3877094)

BACKGROUND: The 21^st Century Cures Act and the rise of telemedicine led to a focus on patient portals. However, portal use disparities persist and are in part driven by limited digital literacy. To address digital disparities, we implemented an integrated digital health navigator (DHN) program to support portal use by patients. The DHN is a dedicated clinic member who enrolls/trains patients on using the portal. Our objective was to assess the perceived value and experience of program organizers and staff member end users of the program.

METHODS: The DHN program was a piloted from March to July 2021 at a primary care clinic in Boston, MA. We conducted semi-structured interviews to allow for exploration of participant experiences. The interview guide was developed based on literature review and expert opinion. Two researchers independently conducted a qualitative analysis using the grounded theory inductive approach. In cases of disagreement, the other researchers assisted in reaching consensus. The researchers coded the transcripts to identify, categorize and sort key concepts. Codes were then grouped into emergent themes and relationships after iterative review and discussion.

RESULTS: We interviewed 13 staff members, 4 who organized the program (i.e., 1 director, 1 practice manager, and 2 administrative assistants) and 9 who were staff end users of the program (i.e., 3 physician assistants, 1 pharmacist, 1 medical assistant, 1 resident, and 1 primary care physician). Interviews with the organizers highlighted that the DHN program: 1) required integration into existing workflows and changes to the clinic space to allow the DHN to interact with patients and 2) helped the organization understand patient’s digital needs, and 3) supported patients that needed extra support to use the portal (Table). Interviews with the staff end users revealed that the DHN 1) provided a dedicated support for portal training, 2) integrated well into the workflow of a primary care visit, especially due to the multiple ways to contact the DHN, 3) facilitated the integration of technology into care (i.e., sending glucose readings using the portal), and 4) aligned with the health equity efforts of the clinic.

CONCLUSIONS: A DHN program can be integrated into a primary care clinic workflow. Digital health navigation offers a promising model for healthcare organizations to take a proactive approach to digital health equity in the primary care setting.

ARE FEWER CASES OF DIABETES MELLITUS DIAGNOSED IN THE MONTHS AFTER SARS- COV-2 INFECTION? A POPULATION LEVEL VIEW IN THE EHR-BASED RECOVER PROGRAM.

Neha V. Reddy¹; Steven G. Johnson³; MICHAEL D. EVANS⁴; Jeremy Harper⁵; Carolyn Bramante²

¹General Internal Medicine, University of Minnesota Medical School Twin Cities, Minneapolis, MN; ²Medicine, University of Minnesota Twin Cities, Minneapolis, MN; ³Institute for Health Informatics, University of Minnesota Twin Cities, Minneapolis, MN; ⁴Clinical and Translational Science Institute, University of Minnesota Twin Cities, Minneapolis, MN; ⁵Owl Health Works, Indianapolis, IN. (Control ID #3876852)

BACKGROUND: Long-term sequelae of severe acute respiratory coronavirus-2 (SARS-CoV-2) infection may include an increased incidence of diabetes. Our objective was to describe the temporal relationship between new diagnoses of type 2 diabetes mellitus and SARS-CoV-2 infection in a nationally representative database.

METHODS: We conducted a cross-sectional analysis to portray the temporal relationship between new- onset type 2 diabetes and SARS-Co-V-2 infection, among persons diagnosed with diabetes after September 2019.

RESULTS: There appears to be a sharp increase in diabetes diagnoses in the 30 days immediately surrounding SARS-CoV-2 infection, followed by a decrease in new diagnoses in the post-acute period, up to 360 days after infection (Figure 1).

CONCLUSIONS: These results underscore the need for further investigation, as understanding the timing of new diabetes onset after COVID-19 has implications regarding potential etiology and screening and treatment strategies.

BARRIERS AND FACILITATORS OF DIGITAL HEALTH SERVICES: QUALITATIVE FINDINGS FROM PATIENTS IN FEDERALLY QUALIFIED HEALTH CENTERS

Allison Pack¹; Stacy Bailey¹; Paula Rusca²; Josephine Llaneza²; rodolfo zuleta¹; Wei Huang¹; Guisselle Wismer³

¹General Internal Medicine, Northwestern University Feinberg School of Medicine, Chicago, IL; ²Erie Family Health Center, Chicago, IL; ³General Internal Medicine, Northwestern University Feinberg School of Medicine, Chicago, IL. (Control ID #3876202)

BACKGROUND: As a result of the COVID-19 pandemic, digital health services (i.e., patient portals and video telehealth visits) have become ubiquitous. To examine potential disparities in the use of these services, we previously conducted a survey of 500 English or Spanish-speaking adults. Survey participants had attended ≥1 visits (in person or via telehealth) for themselves or their child over a 12-month period within one large federally qualified health center network in and around Chicago, IL. Using a sequential, mixed- methods design, we then conducted qualitative interviews to solicit explanations for survey findings.

METHODS: Interview participants were eligible based on prior survey completion, and purposive sampling ensured a mix of ‘users’ and ‘non-users’ of digital health services. A bilingual research coordinator (RC) telephoned participants and conducted audio-recorded interviews using a semi-structured interview guide. Questions explored barriers and facilitators to use of the patient portal and video telehealth visits. Analyses were conducted using a modified version of the Rapid Identification of Themes from Audio Recordings (RITA) procedures. Themes and illustrative quotes, transcribed verbatim, and translated into English when necessary, were identified.

RESULTS: A total of 24 participants were enrolled (12 users and 12 non-users of digital health services). Enrolled participants were diverse by language (English and Spanish), age, sex, and caregiver status. Barriers to portal uptake included a lack of knowledge, with some non-users reporting limited awareness of the portal itself, limited understanding of how to use it, and/or how the portal can be personally beneficial. Barriers to sustained use of the portal, reported among users, included challenges with the password reset feature and difficulties linking multiple children to a single account. Participants in both groups viewed in-person support, supplemented with written materials, as a valuable facilitator of portal uptake. Flexible scheduling of appointments and viewing of lab results were perceived as compelling features. For video telehealth visits, barriers to uptake were similar to those reported for the portal. Several non-users were unfamiliar with the option, while some participants in both groups expressed a general preference for in-person visits. On the other hand, users found video telehealth visits easy to access and join, and participants in both groups noted the benefit of increased flexibility with scheduling and transportation.

CONCLUSIONS: Results revealed key recommendations for enhancing the uptake and sustained use of the patient portal and video telehealth visits. Consideration should be given to enhancing awareness of digital health services, with educational materials or in-person support highlighting when and how such services can be personally beneficial. Enhancing the user-experience of portal security and profile features may facilitate sustained use.

BUILDING PATIENT-PROVIDER RELATIONSHIPS AND PROMOTING SELF-MANAGEMENT: A QUALITATIVE STUDY OF SECURE MESSAGING IN THE ECLIPPSE COHORT

Hannah Wilson³; Dean Schillinger¹; Mary E. Reed²; Nicholas Duran⁴; Wagahta Semere¹

¹General Internal Medicine at Zuckerberg San Francisco General Hospital, University of California San Francisco, San Francisco, CA; ²Division of Research, Kaiser Permanente, Oakland, CA; ³University of Central Florida College of Medicine, Orlando, FL; ⁴Arizona State University, Tempe, AZ. (Control ID #3876412)

BACKGROUND: Secure messaging (SM) via online patient portals allows asynchronous two-way communication between patients and physicians, which can be especially beneficial for patients with chronic diseases who have substantial inter-visit care needs. Recent estimates suggest nearly 1/3 of patients with type 2 diabetes engaged in patient portal use. Among patients with diabetes, prior studies have reported higher rates of patient satisfaction and improved self-management associated with use of SM. Yet, little is known about how this communication platform is used by patients with type 2 diabetes and comorbidities, and the role that secure messaging plays in their longitudinal care.

METHODS: This is a sub-study of the Employing Computational Linguistics to Improve Patient-Provider Secure Emails (ECLIPPSE) study. We conducted an exploratory, in-depth qualitative analysis of all SMs involving a random sample of patients and their PCPs from the ECLIPPSE cohort who sent ≥ 1 English- language SM between July 1, 2006, and Dec. 31, 2015. Applying reflexive thematic analysis, two researchers independently analyzed and coded all patient SM threads (defined as SMs sent and received under the same subject heading) regarding management of diabetes and comorbidities, moving from specific observations to describe broader themes. Through regular meetings and iterative input, the research team established consensus to develop a codebook applied to the entire patient sample of SMs/threads.

RESULTS: The sample included 25 participants with mean age of 58 yrs (SD 7.9); 11 were women, 7 identified as Asian, 5 as Black, and 2 as Hispanic. We examined all SMs patients authored over the study period. Among all patients, we analyzed a total of 1,354 SMs embedded within 455 threads, and identified five overarching major themes describing the role of SM communication in management of patients’ diabetes and other comorbidities: (1) Clarifying and updating care plans between visits, (2) Developing partnered strategies to address emergent health concerns, (3) Building rapport and trust, (4) Promoting patient self-management and health empowerment, and (5) Coordinating care between primary care and specialty providers.

CONCLUSIONS: Patients with diabetes use SM communication with their PCPs to support their health care over time, allowing patients to engage in self-management, deal with emergent health concerns, and invest in their relationship with their PCP. Future research should aim to further investigate the impact of secure messaging engagement on patient’s diabetes care and outcomes.

CAN SOCIAL MEDIA SENTIMENT ANALYSIS (SA) PREDICT POPULATION SENTIMENT TOWARD THE COVID-19 VACCINE?

RAJMOHAN RAMMOHAN¹; Atul Sinha²; Melvin Joy¹; Dilman K. Natt¹; Tulika Saggar¹; Charlene Curtis- Thomas¹; Susan Bunting¹; Prachi Anand¹; Sai Greeshma Magam²

¹INTERNAL MEDICINE, NASSA UNIVERSITY MEDICAL CENTER, East Meadow, NY; ²Internal Medicine, Nassau University Medical Center, East Meadow, NY. (Control ID #3875170)

BACKGROUND: Covid 19 vaccination is one of the debatable topics in today's societies. The COVID pandemic fueled one of the fastest vaccine developments in history. However, one major drawback we face is the skepticism and public opinion regarding the new vaccines. In the study, we analyze the public sentiments towards the Covid vaccine during the pandemic and its current state. The social media platform has a rich data source that can analyze population sentiment during a crisis. Positive sentiments are likely attributed to data showing less infectivity rate and hospitalizations. The negative sentiments could be due to vaccine side effects, lack of knowledge, specific cultural backgrounds, and social media misinformation.

METHODS: We used the Twitter application programming interface (API), Hypertext preprocessor (PAP), and Response capture interface (RAI) to collect the response across Twitter, YouTube, Reddit, and Facebook. Lexicon-based sentiment analysis of the tweets was done using the VADER sentiment analysis tool. Descriptive statistics were used to summarize absolute numbers, frequencies, means, and standard deviations (SD). Independent t-tests were used to compare the average positive and negative sentiment. Finally, correlation analysis was used to examine the relationship between the sentiment across the three years. The data were analyzed with SPSS

RESULTS: We recorded 2,001,450 responses from December 2020 - March 2020, 1,920,872 responses between September 2021-November 2021, and 1,987,590 responses between September 2022-October 2022. Positive sentiment toward the Covid vaccine has declined significantly from 2020 to 2021 and 2022 (68% vs.55% vs. 31%, P=0.021 95% CI). Negative sentiment toward Covid vaccine has increased significantly between 2020 to 2021 and 2022 (18% vs. 25%vs. 47%, p<0.01 95%CI). No change in Neutral sentiment (15% vs. 20% vs. 23% p=0.67).

CONCLUSIONS: Sentiment analysis using a response from social media platforms is helping to understand the general population's ideology. In our study, a time-based assessment of Social media response has shown a reduction over time in the frequency of Positive sentiment toward a Covid vaccine. Furthermore, the polarization in the Social media users, based on sentiment polarity, showed that the users were well connected, highlighting that such issues bond our society rather than segregate it.

CONTRIBUTORS TO VARIATION IN EHR TIME AMONG PRIMARY CARE PHYSICIANS

Lisa Rotenstein¹; A Jay Holmgren²; Daniel M. Horn³; Russell Phillips⁴; Stuart Lipsitz¹; Richard S. Gitomer⁵; David W. Bates¹

¹Medicine, Brigham and Women's Hospital, Boston, MA; ²Medicine, University of California San Francisco School of Medicine, San Francisco, CA; ³Medicine, Massachusetts General Hospital, Boston, MA; ⁴Center for Primary Care, Harvard Medical School, Boston, MA; ⁵Internal Medicine, Brigham Health, Boston, MA. (Control ID #3873898)

BACKGROUND: Primary care physicians (PCPs) spend the most time on the EHR of any specialty, and increased EHR time, particularly after-hours, has been associated with physician burnout. It is unclear how PCP, panel, schedule, messaging, clinic, and teamwork factors influence EHR time.

METHODS: Our sample consisted of PCPs who practiced at Brigham and Women’s Hospital and Massachusetts General Hospital in 2021. Using Epic Signal data, we calculated physician-level daily averages of total EHR time, time outside of scheduled hours, and time spent on the electronic inbox, all normalized to 1.0 clinical full time equivalent (FTE). We also calculated PCP-level averages of percent of orders with team contribution (signed orders pended by a different team member) and the number of appointments and messages completed in 2021.

We descriptively analyzed PCPs’ demographic, panel, schedule, and message volume characteristics and EHR time during 2021. We additionally described clinic staffing ratios and supports. We then built multivariable linear regression models with standard errors clustered by clinic to determine the association of PCP, panel, schedule, message volume, clinic, and teamwork characteristics with total daily EHR time, EHR time outside scheduled hours, and daily time spent on the electronic inbox, all normalized to 1.0 clinical FTE.

RESULTS: We evaluated 307 PCPs (60% female). PCPs’ mean (SD) clinical FTE was 0.5 (0.3). PCPs had completed residency a mean (SD) of 20.9 (11.3) years prior. Mean (SD) panel size was 883.9 (510.7), mean panel risk score was 1.9 (0.5), and a mean (SD) of 58.9 (22.6%) of PCPs’ panels were female. PCPs had completed a mean (SD) of 1154 (605) visits and received a mean (SD) of 12,475 (5,622) messages during 2021.

Per clinical FTE, PCPs spent a mean (SD) of 308.5 (135.5) total daily minutes on the EHR, 173.5 (153.3) minutes daily outside scheduled hours on the EHR, and 67.8 (34.5) minutes daily on the electronic inbox, with significant variation by clinic.

Female PCP gender and greater total message volume per clinical FTE were associated with significantly greater daily time in all categories (p<0.05). In contrast, PCPs with greater team contribution to orders and above median medical assistant to provider ratios in their practices had significantly lower daily total EHR time, time outside scheduled hours, and electronic inbox time (p<0.05).

CONCLUSIONS: We assessed EHR use patterns of more than 300 PCPs and found significant associations between modifiable primary care practice factors and time spent on the EHR in total, outside of scheduled hours, and on the electronic inbox. These results suggest the potential value of increasing team contributions to orders and medical assistant staffing, as well as reducing the burden of messages from all sources, to optimize PCPs’ EHR time.

DEVELOPING A MODEL TO PREDICT INPATIENT HYPOGLYCEMIA

Sarah R. Stern¹; Richa Bundy¹; Lauren Witek¹; Adam Moses¹; Cynthia Burns¹; Matthew A. Gorris¹; christopher kelly¹; Ajay Dharod²

¹Internal Medicine, Wake Forest Baptist Medical Center, Winston-Salem, NC; ²Internal Medicine, Wake Forest University School of Medicine, Winston-Salem, NC. (Control ID #3870102)

BACKGROUND: Inpatient hypoglycemia is associated with increased length of stay and mortality.¹ Prior work has shown predictive models which include logistic regression models including the variables, weight, creatinine clearance, basal insulin dose and whether the patient is receiving mealtime and sliding scale insulin, can predict which patients are at high risk of developing hypoglycemia during their hospitalizations^2,3,4Application of this equation, when paired with a system in which charge nurses were trained to assess the hypoglycemia score and recommend changes to the prescribing teams’ insulin regimens, was able to reduce inpatient hypoglycemic episodes.⁵ We wanted to apply a similar model without altering nurse workflow by allowing providers to view the hypoglycemia risk score in the EHR.

METHODS: A logistic regression model was trained on inpatient hospitalizations of diabetic patients receiving insulin at Wake Forest Baptist Hospital from January 2020 to December 2021. The model was developed to predict a hypoglycemic event (glucose < 70 mg/dL), within 24 hours of a borderline-low glucose measurement (70-90 mg/dL). Predictors were defined a priori based on previous predictive models and input from our clinical team. We split 60% of the dataset for model training and 40% for testing. We are currently building the final risk score into our EHR to display as a print group on the Glucose Management Tab, which displays a patient’s blood sugar trends and insulin administration.

RESULTS: The full dataset included 8,545 inpatient admissions, of which 1,307 (15%) had an outcome of hypoglycemia within 24 hours of a borderline- low glucose result. The most predictive model parameters included maximum insulin doses, a prior episode of severe hypoglycemia, and a diagnosis of type 1 diabetes. The model area under the curve (AUC) was 0.69 on the validation dataset.

CONCLUSIONS: We trained a logistic regression model to predict which patients are at high risk of hypoglycemia and found the model was moderately predictive. We are currently prospectively validating the model after building it into our EHR. We then plan to pilot the risk score on several inpatient medicine services, with associated clinical decision support recommendations.

EFFECTS OF BEHAVIORAL INTERVENTIONS ON OVERUSE IN OLDER ADULTS SIX MONTHS AFTER STOPPING INTERVENTIONS: EXTENDED FOLLOW UP OF THE BEHAVIORAL ECONOMIC APPLICATIONS TO GERIATRICS LEVERAGING ELECTRONIC HEALTH RECORDS (BEAGLE) RANDOMIZED CONTROLLED TRIAL

Stephen Persell¹; Tiffany Brown²; Jason N. Doctor³; Craig R. Fox⁴; Noah J. Goldstein⁵; Ji Young Lee⁶; Jeffrey A. Linder⁷; Daniella Meeker⁸; Yaw A. Peprah⁹; Lucia C. Petito¹⁰; Theresa A. Rowe¹¹

¹Medicine, Northwestern University, Evanston, IL; ²Department of Medicine; Division of General Internal Medicine and Geriatrics, Northwestern University Feinberg School of Medicine, Chicago, IL; ³University of Southern California, Los Angeles, CA; ⁴Anderson School of Management, UCLA, Los Angeles, CA; ⁵UCLA, Los Angeles, CA; ⁶School of Medicine, Northwestern University, Chicago, IL; ⁷Division of General Internal Medicine and Geriatrics, Northwestern University Feinberg School of Medicine, Chicago, IL; ⁸Preventive Medicine, Pediatrics, University of Southern California Keck School of Medicine, Los Angeles, CA;⁹General Internal Medicine, Northwestern Universiry, Chicago, IL; ¹⁰Preventive Medicine , Northwestern University, Chicago, IL; ¹¹Department of Internal Medicine and Geriatrics, Northwestern University, Chicago, IL. (Control ID #3875128)

BACKGROUND: Overtesting and overtreatment of older adults may lead to harm. Redirecting attention to possible harms of overuse and social and reputational concerns using behavioral-science-informed decision support may be effective ways to change clinician behavior.

METHODS: We performed a pragmatic cluster-randomized controlled trial randomizing clinics to a behavioral-science-informed clinical decision support intervention designed to reduce overtesting and overtreatment in older adults based on Choosing Wisely or standard of care. The intervention targeted 3 co- primary overuse outcomes: 1) prostate-specific antigen (PSA) screening in men 76y and older; 2) urine testing for non-specific reasons in women 65y and older and 3) overtreatment of diabetes in patients 75y and older (HbA1C<7% and treated with insulin or sulfonylurea). These alerts drew attention to possible harms of overuse; 2 alerts further required written justification in the encounter document or else the statement “No justification was given” was added, appealing to social and reputational concerns. The study was conducted in 60 primary care practices with 371 clinicians in northeastern Illinois. The intervention period lasted 18 months starting September 1, 2020 and we assessed outcomes for an additional 6-months post-intervention cessation. To adjust for secular trends, we estimated a piecewise hierarchical mixed effects logistic regression model with a knot at month 0 (start of intervention) with treatment indicators and clinician-level random effects.

RESULTS: At baseline, annual rates of overuse were 28.1% (2,490 / 8,865) for prostate cancer screening, 30.7% (3,391 / 11,051) for urinary testing for non-specific reasons, and 20.8% (1,155 / 5,553) for diabetes overtreatment. Overuse of all three clinical targets was significantly reduced during the study period by the decision support interventions: for prostate cancer screening (adjusted OR at 18 months compared to control, 0.61; 95% CI, 0.51 to 0.74), urine testing for non-specific reasons (aOR, 0.69; 95% CI, 0.59 to 0.80), and diabetes overtreatment (aOR, 0.75; 95% CI, 0.59 to 0.96). The differences persisted after incorporating an additional 6 months of post-intervention data: prostate cancer screening, aOR 0.69 (95% CI, 0.55 to 0.87); urine testing, aOR 0.63 (95% CI, 0.52 to 0.77); and diabetes overtreatement, aOR 0.71 (95% CI, 0.52 to 0.97).

CONCLUSIONS: Clinical decision support tools informed by behavioral science that were designed to increase clinicians’ attention to possible harms and to draw on social and reputational concerns to reduce overuse and overtreatment in older patients moderately reduced overuse of diagnostic testing, a finding that persisted 6 months post-intervention.

EPIC SECURE CHAT PRACTICES AND EXPERIENCES AT A LARGE TERTIARY CARE CENTER

Frans J. Beerkens¹; Amanda Owens¹; Nikola Kocovic¹; Aleesha Shaik¹; Aviv Alter¹; Amit Blumfield¹; Brandon Needelman¹; Brian Hsia¹; Vinh-Tung Nguyen¹; Beth G. Raucher¹; David C. Thomas¹; Jonathan L. Halperin²

¹Medicine, Icahn School of Medicine at Mount Sinai, New York, NY; ²Medicine, Cardiology, Mount Sinai Medical Center, New York, NY. (Control ID #3872191)

BACKGROUND: Miscommunication in healthcare is a source of medical error, increases provider stress, and interrupts patient care. Epic Secure Chat (ESC) is an instant messaging system integrated within the electronic medical record that allows for immediate communication between members of the healthcare team. We investigated the burden and experiences with ESC among care team providers at a large urban academic hospital.

METHODS: An online survey was sent to multidisciplinary care team members who regularly use ESC, including physicians, nurses, pharmacists, social workers, case managers, nutritionists, speech language pathologists, radiologists, and physical and occupational therapists. The primary outcome was satisfaction with current ESC use, measured with a Likert scale 1-5 with 5 indicating greatest satisfaction. Secondary endpoints included the perceived burden of ESC measured by Likert scale, estimated number of messages sent and received daily, frequency and characteristics of appropriate and inappropriate messages, and perceived professionalism.

RESULTS: Responses from 155 care team members are included in this analysis. Of surveyed users, almost all (n = 126) were satisfied (scale 4/5 or 5/5) with ESC communication (Figure 1). The majority (n = 44) found that ESC posed an average burden (scale 3/5) with an estimated send/receive message volume 20-50 per day (n = 56). The daily number of messages care team members estimated exchanging varied by discipline, with pharmacists and nutritionists estimating <20 messages compared with nurses and physicians estimating 50-100 messages daily. Despite high satisfaction with ESC, many (n = 60) reported occasionally receiving inappropriate ESC messages, and some (n = 15) noted they often or always receive inappropriate messages. Common themes characterizing inappropriate communication included critical or urgent messages, unprofessional tone, and high frequency messaging.

CONCLUSIONS: Satisfaction with ESC was high but perceived burden varied among care team members. Many reported at least occasional inappropriate messaging. These results highlight a need to educate users on appropriate ESC practices.

EVALUATING ACCEPTABILITY AND FEASIBILITY OF COMMSENSE, A NOVEL WEARABLE TECHNOLOGY TO IMPROVE CLINICIAN-PATIENT COMMUNICATION

Virginia LeBaron¹; Tabor E. Flickinger²; David Ling²; James Edwards¹; David Lee²; Anant Tewari²; Anna Marie Kutcher¹; Zhiyuan Wang³; Hassan Nusayer³; Laura Barnes³

¹University of Virginia School of Nursing, Charlottesville, VA; ²University of Virginia School of Medicine, Charlottesville, VA; ³University of Virginia School of Engineering and Applied Science, Charlottesville, VA. (Control ID #3875818)

BACKGROUND: Evaluating the quality of clinician-patient communication during real clinical encounters is challenging. Technology offers novel opportunities to provide clinicians with actionable feedback to measure, track, and improve their communication skills, but assessing end-user perspectives is critical to ensure successful implementation. Our team has developed and piloted CommSense technology to address this need.

METHODS: This prospective, descriptive study evaluated the acceptability and feasibility of CommSense, a novel wearable technology designed to record and extract key metrics of communication performance from

real-time conversations. Nursing and medical students and practicing clinicians were recruited from an academic medical center. Participants wore a smartwatch with the CommSense application and read 2-4 scripts with a ‘patient’ (member of our study team). Scripts focused on conveying difficult news, discussing prognosis, and managing pain and contained examples of positive (e.g., active listening) or negative (e.g., interruptions, medical jargon) communication skills. Conversations were recorded by CommSense, de- identified and off-loaded for analysis. Participants completed an electronic exit survey (Qualtrics) to assess their experience with CommSense and opinions about its potential application in clinical practice; survey results were exported to SPSS and descriptive statistics calculated.

RESULTS: Forty (n=40) participants (5 clinicians; 19 nursing students; 16 medical students) pilot tested CommSense, the majority aged 18-24 (n=29, 72.5%); female (n=32, 80%), and White (n=24, 57.1%). Most participants reported having ‘a little’ experience communicating with seriously ill patients (n=33, 55%) and formal communication training (n=21, 52.5%). Over 90% of participants (n=38, 95%)‘strongly agreed’ or ‘agreed’ that CommSense could improve compassionate communication between patients and clinicians and help healthcare organizations deliver high quality care (n=39, 97.5%). 87.5% (n=35) ‘strongly agreed’ or ‘agreed’ they were comfortable with CommSense recording clinical conversations; 92.5% (n=37) ‘strongly agreed’ or ‘agreed’ they would be willing to use CommSense in the future; 47.5% (n=19) ‘strongly agreed’ or ‘agreed’ to concerns regarding privacy/confidentiality. 100% (n=40) ‘strongly agreed’ or ‘agreed’ they were interested in seeing information from CommSense about their communication performance

CONCLUSIONS: Overall, our pilot demonstrated strong feasibility and acceptability of CommSense. Participants positively evaluated its potential to help improve patient-clinician communication and healthcare delivery. Although participants expressed willingness to use CommSense and openness to feedback from it, there were concerns regarding privacy and confidentiality; attention to data protection will be critical in future studies of CommSense with real patients. It will also be important to test CommSense with a more diverse sample and with more practicing clinicians.

EVALUATION OF PROVIDER CLINICAL DECISION SUPPORT SYSTEM ADOPTION RATES BY PATIENT RACE

Ratnalekha V. Viswanadham¹; Simon Jones²; Kellie Owens²; Safiya Richardson^3,4

¹Pediatrics, New York University Grossman School of Medicine, New York, NY; ²Population Health, New York University Grossman School of Medicine, New York, NY; ³Deparment of Population Health, NYU Langone Health, New York, NY; ⁴Medicine, New York University Grossman School of Medicine, New York, NY. (Control ID #3876433)

BACKGROUND: Best practice alerts (BPAs) are a type of clinical decision support (CDS) within electronic health records (EHRs). Meta-analyses of the effects of CDS like BPAs have shown that providers who use decision support are more likely to order appropriate treatments and provide preventive care services. However, it is unknown whether there are differences in CDS adoption based on patient race. Differences in CDS adoption could contribute to health inequities. This is a critical gap in the literature for two reasons. First, differences in CDS and BPA adoption could contribute to health inequities. Second, understanding heterogeneity in using BPAs can help determine how CDS can be effectively designed to improve patient care. This study aimed to determine whether differences in BPA utilization by patient race exist and, if so, the underlying reasons for these differences.

METHODS: We retrospectively analyzed over 3 million BPA firings in the EHR with the corresponding patient, provider, encounter, and BPA information during outpatient primary care encounters at a large academic health system from July 1, 2018, to June 30, 2019, and from July 1, 2021, to June 30, 2022. We used linear mixed models to assess the probability of BPA adoption based on patient race, adjusting for random effects at the patient, provider, BPA, and encounter levels.

RESULTS: In the unadjusted analyses, we found that the percentage of BPAs responded to during a visit varied by patient race. Patients of non-white races had fewer BPA responses during an outpatient visit (-7% for African American/Black patients, - 8.06% for Asian patients, and -0.86% for patients of other or multiple races) compared to White patients (intercept, 16.3% (SE = 0.00025, p < 0.0001). The significant differences in BPA responses remained consistent when controlling for patient, provider, and encounter-level random effects. However, when controlling for BPA-level random effects (i.e., controlling for any possible characteristic of a BPA, like its content, how it appears in the EHR workflow, and the actions a provider can take after a BPA), patients of non-White races had more BPA responses during an outpatient visit (2.58% for African American/Black patients, 2.75% for Asian patients, and 5.15% for patients of other or multiple races) compared to White patients (intercept, 18.3% (SE = 0.0344), p < 0.01).

CONCLUSIONS: BPA design variables, like the types of BPAs, their placement in the EHR workflow, and the content disseminated in them, contribute significantly to providers using clinical support less for patients of color. Future directions will include machine learning and other statistical learning-based methods to estimate what set of BPA-related factors best predicts the probability of a BPA response based on patient race and assess how these factors may impact racial health inequities.

EXPANDING AND ENHANCING MY DIABETES CARE: A USER-CENTERED DESIGN STUDY

Elsa Rodriguez¹; Shilo Anders¹; Carrie Reale¹; Amber Hackstadt¹; Zhihong Yu¹; Lyndsay Nelson¹; Lindsay Mayberry¹; S. Trent Rosenbloom¹; Audriana Nigg¹; Adam Wright²; Tom Elasy¹; Lipika Samal³; William Martinez¹

¹Vanderbilt University Medical Center, Nashville, TN; ²Biomedical Informatics, Vanderbilt University Medical Center, Nashville, TN; ³Division of General Internal Medicine, Brigham and Women's Hospital, Boston, MA. (Control ID #3868884)

BACKGROUND: My Diabetes Care (MDC) is a SMART on FHIR patient portal app designed to help patients better understand their diabetes health data while promoting and supporting self-management. Our prior research suggested MDC may improve patients’ understanding of diabetes health measures (eg, HbA1c) and increase diabetes self-efficacy. We aimed to expand the diabetes health data displayed in MDC beyond HbA1c, blood pressure, and cholesterol, to include urine microalbumin and body weight, and enhance access to Spanish-speaking patients.

METHODS: We applied user-centered Design Sprint methodology to design a prototype of a revised MDC user interface to include urine microalbumin and body weight. We conducted two iterative rounds of prototype usability testing. We used purposive sampling to ensure representation of participants aged ≥65 and with limited health literacy. Participants were asked to review the prototyple and perform standardized tasks (eg, interpret values) using ‘think-aloud’ procedures to understand their user experience. Participants’ statements were qualitatively analyzed to inform revisions. Participants rated overall satisfaction and ease of use from 1-worst to 5-best and completed the 10-item System Usability Scale (SUS) scored 0-worst to 100- best. To inform a planned Spanish version of MDC, we conducted a community engagement studio (CES) with Spanish-speaking patients.

RESULTS: Prototype usability testing participants (n=12) had a mean HbA1c of 7.9; 6 (50%) were female, 3 (25%) were age ≥65, and 4 (33%) had limited health literacy. Round 1 participants' (n=6) statements indicated confusion on how to enter and edit goal weight and whether home or clinic weight was displayed. The mean overall ease of use and satisfaction were 4.8 and 4.2, respectively. Participants appreciated the green, yellow, and red color coding and labeling of health data values (eg, HgbA1c) to indicate values in the goal, caution, and high ranges, respectively. Among round 2 participants (n=6), the mean ease of use remained high (4.7), and overall satisfaction rose to 4.7. SUS scores were high in both rounds, consistent with excellent usability (mean SUS was 92 and 95, respectively). CES participants (n=3) felt the direct translation of the app title (“mi Cuidado a la diabetes”) was awkward. They shared that “Cuidado” is not very positive and may be interpreted as “caution.” Participants also recommended including resources to assist patients in discussing diabetes with their family and friends to engage their support in self-care (eg, diet)

CONCLUSIONS: Participants were able to use MDC with ease and highly rated its usability. Revisions led to an improved user interface for body weight. These findings suggest that the fully programmed version of MDC will likely meet the needs of target users. CES findings highlighted the need to translate app language with an awareness of cultural nuance. Further research is underway to test the usability of fully programmed, English and Spanish, versions of MDC.

EXPERIENCES USING AT-HOME VIRTUAL REALITY FOR CHRONIC PAIN: A QUALITATIVE STUDY

Genevieve Bryant¹; Hector R. Perez²; Lesly Sanchez Alvarez³

¹General Internal Medicine, Albert Einstein College of Medicine, Bronx, NY; ²Division of General Internal Medicine, Albert Einstein College of Medicine, Bronx, NY; ³Internal Medicine, Albert Einstein College of Medicine, Bronx, NY. (Control ID #3876608)

BACKGROUND: Virtual reality (VR) is a promising nonpharmacological option for chronic pain management. However, little is known about the quality and depth of patients’ experiences with at-home VR for chronic pain. As part of an ongoing pilot clinical trial, we sought to elucidate patients’ experiences using an at-home VR device for pain management in a qualitative study.

METHODS: We conducted a qualitative survey of participants in the Bronx, NY who used a VR device for chronic pain at-home for one week. Patients were eligible if they were English-speaking adults with at least moderate chronic pain. Patients were excluded if they had contraindications to using the device, including epilepsy. Participants were recruited from electronic medical record databases and from prior pain clinical trials. The study utilized a device marketed for chronic pain and used in prior clinical trials, the RelieVRx (AppliedVR, Van Nuys, CA). The RelieVRx uses immersive mindfulness and cognitive behavioral interactive visuals and VR activities, including breathing exercises with visual cues, to reduce pain. After using the device for a week, we conducted a qualitative interview focused on how patients used the device, their thoughts and feelings during use, and how they thought VR helped or hindered their pain. Interviews were coded and analyzed by two authors using thematic analysis.

RESULTS: Of 5 people that completed the at-home study, 4 participants agreed to a qualitative interview. Median age was 55, 3 were female, 1 was non-Hispanic Black, and 3 were Hispanic. The median pain level was 4.8/10. All participants had never used VR prior to the study. In interviews, participants reported an overall positive experience. All participants described how the immersion provided by the device facilitated relaxation (“It’s like I’m a part of [nature] as I do those exercises, because I breathe along with the flow and the movement of the picture inside the VR.”) Three participants also elaborated on how relaxation had beneficial effects on mood and sleep (“It’s teaching you how to breathe, to relax your muscles, your nerves. I did feel good when I was doing it.”). Lastly, two participants reported how the device led to positive effects on pain and relaxation even when not using the device (“Any time my body experienced pain I can go right to that state of mind and start to visualize that I am wearing the machine.”). In contrast, one participant felt that while the device induced relaxation, there were limited effects on pain and the novelty decreased over time.

CONCLUSIONS: In this small sample of participants who piloted VR for chronic pain at home, participants noted positive overall experiences with some potentially durable effects. Further research with more participants to elucidate potential mechanisms for the effectiveness of VR in chronic pain.

EXPERIENCES USING AT-HOME VIRTUAL REALITY FOR CHRONIC PAIN: A QUALITATIVE STUDY

Genevieve Bryant²; Hector R. Perez¹

¹Division of General Internal Medicine, Albert Einstein College of Medicine, Bronx, NY; ²General Internal Medicine, Montefiore Health System, Bronx, NY. (Control ID #3877401)

BACKGROUND: Virtual reality (VR) is a promising nonpharmacological option for chronic pain management. However, little is known about the quality and depth of patients’ experiences with at-home VR for chronic pain. As part of an ongoing pilot clinical trial, we sought to elucidate patients’ experiences using an at-home VR device for pain management in a qualitative study.

METHODS: We conducted a qualitative survey of participants in the Bronx, NY who used a VR device for chronic pain at-home for one week. Patients were eligible if they were English-speaking adults with at least moderate chronic pain. Patients were excluded if they had contraindications to using the device, including epilepsy. Participants were recruited from electronic medical record databases and from prior pain clinical trials. The study utilized a device marketed for chronic pain and used in prior clinical trials, the RelieVRx (AppliedVR, Van Nuys, CA). The RelieVRx uses immersive mindfulness and cognitive behavioral interactive visuals and VR activities, including breathing exercises with visual cues, to reduce pain. After using the device for a week, we conducted a qualitative interview focused on how patients used the device, their thoughts and feelings during use, and how they thought VR helped or hindered their pain. Interviews were coded and analyzed by two authors using thematic analysis.

RESULTS: Of 5 people that completed the at-home study, 4 participants agreed to a qualitative interview. Median age was 55, 3 were female, 1 was non-Hispanic Black, and 3 were Hispanic. The median pain level was 4.8/10. All participants had never used VR prior to the study. In interviews, participants reported an overall positive experience. All participants described how the immersion provided by the device facilitated relaxation (“It’s like I’m a part of [nature] as I do those exercises, because I breathe along with the flow and the movement of the picture inside the VR.”) Three participants also elaborated on how relaxation had beneficial effects on mood and sleep (“It’s teaching you how to breathe, to relax your muscles, your nerves. I did feel good when I was doing it.”). Lastly, two participants reported how the device led to positive effects on pain and relaxation even when not using the device (“Any time my body experienced pain I can go right to that state of mind and start to visualize that I am wearing the machine.”). In contrast, one participant felt that while the device induced relaxation, there were limited effects on pain and the novelty decreased over time

CONCLUSIONS: In this small sample of participants who piloted VR for chronic pain at home, participants noted positive overall experiences with some potentially durable effects. Further research with more participants to elucidate potential mechanisms for the effectiveness of VR in chronic pain.

EXPLORING THE BURDEN OF MESSAGES IN EHR INBOX AMONG PRIMARY CARE PHYSICIANS

Richa Bundy²; Adam Moses²; Elisabeth Stambaugh⁴; Paschal Stewart⁴; Lauren Witek²; Lindsey E. Carlasare³; Gary Rosenthal^1,2; Christine Sinsky³; Ajay Dharod^1,2

¹Internal Medicine, Wake Forest University School of Medicine, Winston-Salem, NC; ²Informatics and Analytics, Wake Forest University School of Medicine, Winston-Salem, NC; ³Professional Satisfaction, American Medical Association, Chicago, IL; ⁴Atrium Health Network, High Point, NC. (Control ID #3875895)

BACKGROUND: Physicians spend a substantial portion of their workday in the Electronic Health Record (EHR) and have cited time spent within their inbox to be particularly burdensome (Arndt et al, Annals of Family Medicine, 2017). Measures of inbox time have been mainly characterized and presented in aggregate (Melnick et al, Journal of the American Medical Informatics Association, 2021; Moore et al, Applied Clinical Informatics, 2021) However, understanding types of messages that contribute most to inbox burden is imperative for the development of interventions to target those message workflows and ultimately reduce time spent in the inbox. To our knowledge, no studies have characterized and quantified the types of messages received in the inbox across provider characteristics or by time on inbox.

METHODS: We collected one year of primary care physicians’ inbox volume and time by message type from the EHR data warehouse and the EHR vendor. The volume and time datasets were merged and aggregated by provider, message type, and calendar week. Providers scheduled less than 250 total hours were excluded from the analysis. We used descriptive statistics to assess overall volume and time per message by inbox message type. We also used Wilcoxon Rank-Sum tests to assess differences in median message volume and time by physician sex.

RESULTS: We included 326 primary care physicians’ inbox messages received between November 2021 and October 2022. Of the 69 unique message types in the inbox, three contributed to 54% of the message volume and 77% of the total time spent in the inbox: patient calls, patient medical advice request (PMAR), and results. Female physicians had a significantly higher median volume of PMAR messages than males, adjusted for scheduled hours. However, female physicians spent the same time per message as their male colleagues for these message types (Table 1).

CONCLUSIONS: Most of the time spent in the inbox can be attributed to only three message types. Many message types contribute to inbox volume but involve limited physician time. Additionally, female physicians received a disproportionate volume of messages from patients but spent similar amounts of time per message as their male colleagues.

GAPS IN OSTEOPOROSIS INFORMATION ONLINE

Irvind Buttar, Simran Parmar

Internal Medicine, Lenox Hill Hospital, New York, NY. (Control ID #3869329)

BACKGROUND: Patients are able to access an abundance of health care information via the internet. Online health information that is available should be comprehensive and inclusive. Osteoporosis is a condition effecting 500 million men and women all over the world. It is well established and noted that post menopausal women are affected. However, there is a significant population of patients who take immunosuppressants, have cancer, or metabolic conditions that place them at risk of developing osteoporosis. These populations should be included on osteoporosis informational pages to properly inform those at risk and help increase necessary screening.

METHODS: A review of fifty-eight academic Internal Medicine hospital systems was performed. We initially sought out if the hospital system had a patient education section on Osteoporosis. Out of the systems with patient education sections, we performed a cross sectional review to see if they mentioned screening/diagnostic methods and if they comprehensively included patient populations at risk. The primary outcomes were as follows: patient education section available, diagnostic testing information available, susceptible patient populations mentioned, and susceptible patient populations other than post menopausal women mentioned.

RESULTS: There were fifty-eight academic internal medicine hospital centers with accessible online information analyzed. Out of the fifty-eight main hospital centers, fifty-three (91%) had accessible osteoporosis websites with general information. Out of the fifty-three with these websites, forty-five (81%) mentioned screening/diagnostic modalities. Further, out of the forty-five hospital centers with diagnostic information, twenty-six (58%) mentioned the patient population of post menopausal women, and nineteen (42%) mentioned all other susceptible populations.

CONCLUSIONS: It is imperative to have accurate and comprehensive health care material available for our patients to promote literacy. Specifically focusing on osteoporosis, the above data search showed that only 42% of hospitals with websites convey extensive information on osteoporosis to patients who are at risk other than post menopausal women. This condition can be so lifechanging and limiting if not properly screened for and diagnosed. Proper patient education can increase screening, doctor -patient conversations while creating more informed patients.

HOW MUCH DOES INTRAVENOUS FLUID CAUSE HEMATOCRIT TO DROP?

Nabeel Qureshi^1,2; Joshua Pevnick²; Johan Carrascoza-Bolanos²; Matthew Bloom²

¹RAND Health, RAND Corporation, Santa Monica, CA; ²Cedars-Sinai Medical Center, Los Angeles, CA. (Control ID #3876042)

BACKGROUND: Hematocrit often drops after surgery. To differentiate drops from post-operative bleeding vs intravenous fluids, we sought to estimate expected post operative hematocrit (post-OH) values accounting for net fluid intake and intraoperative estimated blood loss (EBL).

METHODS: We reviewed patient-level data from the electronic health record (EHR) of an academic medical center in the California-Hawaii region for all non-pregnant adults undergoing elective knee or hip arthroplasty from November 2013 to September 2022. We used ordinary least squared regression to evaluate the association between an outcome variable of post-OH and predictors including: clinical [pre-operative hematocrit (pre-OH), intraoperative fluid inputs and outputs, and blood volume]; surgical (time from surgery to lab testing, EBL); and patient height and weight. We only included patients with a pre-OH result in our EHR within 7 days of the scheduled surgery date. We conducted sensitivity analysis using different specifications for estimated circulating blood volume.

RESULTS: 6,648 patients met the inclusion criteria. Of those, 58.6% were female. Mean BMI was 25.5 (SD=6.5). Mean net fluid intake was 1,121.6 mL (SD=795.6). Mean EBL was 143.3 mL (SD=178.1). Our model had an R2 of 0.553 compared to a model of pre-OH only of 0.280. Pre-OH was the strongest predictor of post-OH. Each 1% increase in pre-OH was associated with a 0.70% increase in post-OH. More time between the surgery and the post-OH collection reduced the strength of the relationship between pre-OH and Post-OH. On average, estimated blood loss and a positive fluid balance were associated with a decrease of 5.15% and 0.24% in post-OH, respectively. Our estimates were robust to sensitivity analyses and all variables included in the model were statistically significant with p-values < 0.005.

CONCLUSIONS: Compared to using the pre-OH alone, expected post-OH values may be better estimated by accounting for patient-specific EHR data. Unexpected drops in post-OH should prompt further clinical investigation.

IMPACT OF ARTIFICIAL INTELLIGENCE ON COVID 19 VACCINATION

RAJMOHAN RAMMOHAN¹; Tulika Saggar¹; Atul Sinha²; Melvin Joy¹; Dilman K. Natt¹; Charlene Curtis- Thomas¹; Susan Bunting¹; Prachi Anand¹

¹INTERNAL MEDICINE, NASSA UNIVERSITY MEDICAL CENTER, East Meadow, NY; ²Internal Medicine, Nassau University Medical Center, East Meadow, NY. (Control ID #3875196)

BACKGROUND: Artificial Intelligence is a device that responds to human voices and can be commanded to do various tasks. Artificial Intelligence can be found on smart speakers, smartwatches, mobile phones, tablets, and other devices. The most well-known are Alexa (Amazon), Siri (Apple), Google Assistant (Google), and Cortana (Microsoft). AI is being used to obtain health information, which has become a critical point of analysis for researchers in terms of question understanding and quality of response. Particularly, the COVID-19 pandemic has and still is severely affecting people worldwide, which demands studies on how VAs can be used as a tool to provide useful information.

METHODS: Universal Artificial Intelligence software Apple Siri, Amazon Alexa, Google Assistant, and Microsoft Cortana, were tested for medical information reliability. Voice recordings were done for 5 commonly asked questions in four different languages, English, Spanish, Mandarin, and French, regarding COVID vaccination. Likert’s scale was used to grade the response depending on the credibility of two blinded authors. The study was performed in two phases; (I) Comparisons between 4 AI were performed for the English language; (II) Comparing the outcome between different languages. The accuracies of each answer provided through audio output and the quality of the source supporting each answer were analyzed. The authors decided if each of the four AI provided clinically reliable advice.

RESULTS: Regarding the English language, Apple Siri’s responses were significant for reliability compared to other peers (95% vs. 70% vs. 65% vs. 15%, P: 0.012). Reliability was significant in AI using English as the Primary language as compared to the other Languages (73.2% vs. 60% vs. 55% vs. 10% p<0.01)

CONCLUSIONS: Artificial Intelligence continues to remain an important part of the development of Humanity. Most of the AI performed well in our study when English was the base language, but we believe there is a need for improvement for AI in other languages, especially with the technology becoming more ingrained in our everyday lives.

IMPACT OF EARLY TELEMEDICINE FOLLOW-UP ON 30-DAY HOSPITAL READMISSIONS

Anne Grauer¹; Talea Cornelius¹; Marwah Abdalla¹; Nathalie Moise²; Ian Kronish¹; Siqin Ye³

¹Medicine, Columbia University Irving Medical Center, New York, NY; ²Department of Medicine , Columbia University Medical Center, New York, NY; ³Medicine, Columbia University Medical Center, New York, NY. (Control ID #3868684)

BACKGROUND: Telemedicine is increasing in popularity but the impact of this shift on patient outcomes has not been well described. Prior data has shown that early post-discharge office visits can reduce readmissions. However, it is unknown if routine use of telemedicine visits for this purpose is similarly beneficial.

METHODS: We conducted a retrospective observational study using electronic health records data to assess if the rate of 30-day hospital readmissions differed between modality of post-discharge follow-up visits.

RESULTS: A total of 1,295 index discharges from 1,170 patients were included for analysis. Of these, 397 (31%), 481 (37%), and 417 (32%) discharges had telemedicine, in-person or cancelled/no-show follow-up visits within 7 days of discharge, respectively. The 30-day readmission rate for discharges with in-person, telemedicine, or cancelled/no-show follow-up visits within 7 days were 16%, 15%, and 17% (P=0.64), respectively. Compared to discharges with completed in-person follow-up visits, the adjusted odds of readmission for those with telemedicine follow-up visits was not significantly different (odds ratio [OR] 0.81, 95% confidence interval [CI] 0.53 to 1.25, P=0.34) (Table 1).

CONCLUSIONS: Our study is among the first to examine the impact of primary care telemedicine use on 30 day readmission rates when used for post-discharge follow-up after hospitalization in the general population. Our study showed that 30-day readmission rate did not differ significantly according to the modality of visit. These results provide reassurance that telemedicine visits are a safe and viable alternative for post-hospitalization follow-up. Future studies, including larger randomized controlled studies, will need to confirm our findings.

IMPACT OF SOCIAL MEDIA SENTIMENT ANALYSIS (SA) TO PREDICT POPULATION OPINION TOWARDS ANNUAL SCREENING COLONOSCOPY

RAJMOHAN RAMMOHAN¹; Melvin Joy¹; Atul Sinha²; Tulika Saggar¹; Dilman K. Natt¹; Charlene Curtis- Thomas¹; Susan Bunting¹; Prachi Anand¹

¹INTERNAL MEDICINE, NASSA UNIVERSITY MEDICAL CENTER, East Meadow, NY; ²Internal Medicine, Nassau University Medical Center, East Meadow, NY. (Control ID #3875287)

BACKGROUND: Sentiment analysis is an emerging trend to understand patients’ thoughts in multiple situations, especially in healthcare settings, regarding patients undergoing screening procedures such as colonoscopy. Given the growing use of social media, public health interventions to improve cancer screening are being widely implemented. The primary aim is to analyze the population response toward the annual colonoscopy, and the secondary goal is to analyze the population sentiment toward earlier colonoscopy screening.

METHODS: We used the Twitter application programming interface (API), Hypertext preprocessor (PAP), and Response capture interface (RAI) to collect the response across Twitter, YouTube, Reddit, and Facebook. Lexicon-based sentiment analysis of the tweets was done using the VADER sentiment analysis tool. Descriptive statistics were used to summarize absolute numbers, frequencies, means, and standard deviations (SD). Independent t-tests were used to compare the average positive and negative sentiment. Finally, correlation analysis was used to examine the relationship between the sentiment across the three years. The data were analyzed with SPSS.

RESULTS: We recorded 7,651,350 responses from December 2017 - November 2022. 3,520,673 responses between December 2018 - November 2020 and 4,130,677 responses between January 2021 - November

2022. Positive sentiment toward the Annual colonoscopy has improved significantly from 2018 to 2022 (59% vs. 69% P=0.022 95% CI). Positive sentiment increased from annual colonoscopy from age 50 to age 45 between 2018 to 2022 (53% vs. 67% p<0.01 95%CI).

CONCLUSIONS: Our study showed an increase in the frequency of Positive sentiment toward Annual Colonoscopy and Earlier colonoscopy surveillance. Positive sentiments are likely attributed to data showing that the population favors the benefit of earlier colonoscopy screening. The negative sentiments could be due to side effects, lack of knowledge, specific cultural backgrounds, and social media misinformation. Furthermore, the polarization in the Social media users, based on sentiment polarity, showed that the users were well connected, highlighting that such issues bond our

IMPLEMENTING A STANDARDIZED PROGRESS NOTE TEMPLATE TO REDUCE NOTE LENGTH

Peter Kleinschmidt¹; Adam Rule²; Heidi Twedt¹

¹School of Medicine and Public Health, University of Wisconsin-Madison, Madison, WI; ²Information School, University of Wisconsin-Madison, Madison, WI. (Control ID #3871960)

BACKGROUND: Progress note length in the US far exceeds that of notes in other countries. Note length has increased dramatically over the last decade which is driven largely by duplication of redundant information. Introduction of new evaluation and management (E/M) guidelines in January of 2021 provided an opportunity for reducing extraneous information yet adjusting decades-old habits has been slow to occur. Operational tools such as providing standard documentation templates for notes may encourage this transition to shorter and, hopefully, more readable notes.

METHODS: We developed a standard note template designed to reduce duplication of data, increase readability, and streamline note writing while also providing embedded tools to assist with coding and compliance. The template was made available to general internists writing office progress notes in July of 2021. Adoption of the note was optional but encouraged on a regular basis with involvement of physician informaticists, division heads, and executive-level clinical leadership. Twelve months after implementation of the template we analyzed impacts on progress note length and compared this the 12 months prior to implementation of the template. Inclusion for analysis required authors being staff physicians or advanced practice providers, having data available for at least 3 months in both pre and post change periods, and having completed at least 100 encounters both pre and post change.

RESULTS: Among the 80 users included in this analysis, 29 were frequent users (>50% of encounters written with template), 18 were moderate users (1-50% of encounters), and 33 were non-users (i.e. <1% of encounters). Overall adoption of the template plateaued at 37% of encounters written with the standard template. Frequent users reduced median note length by 902 characters (95% confidence interval: 782,1080), moderate users reduced by 343 characters (28, 879) and non-users reduced by 101 characters (24,162). Frequent users reduced note length significantly more than non-users (P=0.00008, Mann-Whitney U test)

CONCLUSIONS: An active campaign implementing a standard note template for provider documentation substantially shortened documentation length among frequent users of the template. Interestingly, note length shortened even in users who did not adopt the standard template. Although some shift in note characteristics could be attributed to the change in evaluation and management guidelines that occurred in January 2021, previous work has shown that this change had no impact on note length at least in the early months after the changes were put in place. While policy changes alone may not change note writing behaviors, a focused project involving note template standardization can set the tone for an organizational culture on reducing note length, hopefully reducing extraneous or duplicative information, and thereby making notes easier to read and write. Future study will focus on improvement in time to document, readability, and provider satisfaction.

INLINE CLINICAL DECISION SUPPORT: DOCUMENTATION OF PAIN MANAGEMENT WITHIN CLINICAL NOTES

Matthew K. Lee, Archana Saxena, Katherine Moawad, Eduardo Iturrate, Jonathan Austrian, Hardev Randhawa, Roland Casem, Adam J. Goodman, Katherine A. Hochman, Joseph Weisstuch

Jonah Feldman NYU Langone Health, New York, NY. (Control ID #3834245)

BACKGROUND: Pain is a ubiquitous experience across inpatient settings that can greatly shape patients’ hospital course and overall quality of life. This project aimed to provide inline clinical decision support (CDS) for documentation of a pain management assessment and plan. Thorough documentation enhances interdisciplinary therapeutic treatment for pain and allows for frequent provider re-assessment of inpatient pain management. Given that pain may vary in its clinical presentation and responsiveness to treatments, a CDS intervention was designed keeping a broad spectrum of patients in mind. This study aims to increase provider documentation addressing patients’ pain management assessment and plan.

METHODS: An inline CDS disappearing tip was developed to appear in the Assessment and Plan (A&P) section of note templates. The tip is populated in notes for all admitted medicine/surgery patients who had a pain score ≥ 5/10 or received pain medication within the last 24 hours. The tip prompts the provider to document a pain management assessment and plan. On 2/17/22 tracking of the tip commenced across four hospitals using an Epic smart data element.

After tracking, 2703 notes met tip criteria from 2/17/22 to 3/4/22 and were categorized into one of two groups depending on if the CDS was used or not. Retrospective chart review was performed on a random sample of 400 notes (200 from each group), excluding those that did not have an A&P section. Clinical notes were primarily evaluated for documentation of the source of pain and the plan for pain management. Additional information regarding specific features of pain management plans and author role was collected.

RESULTS: After implementation of CDS within clinical notes, 63.5% (127/200) of notes documented a source of pain compared to 59.5% (119/200) without the integrated CDS tip (p=0.206). With the integrated CDS tip, 37% (74/200) of notes explicitly documented a pain management plan compared to 44.5% (89/200) without the integrated CDS tip (p=0.937). All clinical notes with and without CDS tip integration had similar rates of pain management plans that addressed pain severity, breakthrough pain, and consultation.

CONCLUSIONS: In this pilot study of 400 inpatient notes, the pain management inline CDS tip was associated with an increase in the rate of pain documentation. The labeling of notes from the study will facilitate a natural language processing (NLP) evaluation over thousands of published notes. This NLP evaluation will be powered to more definitively assess the impact of the CDS tip on documentation outcomes and identify opportunities for provider education and optimization of the tip. Clearer documentation of a patient’s pain source, assessment, and management plan helps providers address pain more attentively and effectively.

NECESSARY BUT NOT SUFFICIENT: IMPLEMENTATION AND EFFECTIVENESS OF VA’S DIGITAL DIVIDE INITIATIVE DURING THE COVID-19 PANDEMIC

Donna Zulman^1,2; James Van Campen²; Jacqueline M. Ferguson²; Zainub Dhanani³; Alison L. Greene^2,1; Rachel Kimerling^4,2; Cindie Slightam²

¹Medicine, Stanford University, Stanford, CA; ²Center for Innovation to Implementation, VA Palo Alto Health Care System, Palo Alto, CA; ³Health Policy, Stanford University School of Medicine, Stanford, CA; ⁴National Center for PTSD, VA Palo Alto Health Care System, Menlo Park, CA. (Control ID #3877092)

BACKGROUND: Virtual care holds promise for overcoming access barriers, but the "digital divide" impedes many individuals' use of technology. We sought to evaluate the implementation and effectiveness of VA’s national initiative to distribute video-enabled tablets to Veterans with access barriers during the COVID-19 pandemic.

METHODS: The VA began distributing video-enabled tablets to Veterans in 2016, but demand for this resource increased dramatically during the COVID-19 pandemic. In response, the VA implemented a Digital Divide consult in order to scale up tablet distribution and direct tablets and subsidized data plans to Veterans with documented access barriers and care needs. Guided by the RE-AIM Framework, we evaluated this initiative's Reach (i.e., number and characteristics of tablet recipients), Adoption (i.e., rates and volume of video-based service use among tablet recipients), and Effectiveness (i.e., patient-reported experiences with tablets, based on a mailed and phone survey of 3,399 tablet recipients with subsequent video visits). We assessed Implementation by examining reasons for tablet referral and changes over time in the percentage of Veterans who had a video visit within 90 days of tablet receipt.

RESULTS: Between 2020-2022, >130,000 Veterans received VA-issued video-enabled tablets, one-third of whom resided in rural locations. Tablet recipients (compared to the general VA population) were more likely to be <65 yrs (53% vs 47%), non-white (34% vs 28%), and have >5 chronic conditions (45% vs 18%), a mental health condition (70% vs 33%), priority status for high disability (52% vs 41%), and a history of homelessness (19% vs 4%). The most common criteria cited in tablet referrals were a mental health diagnosis (64%), social isolation (24%), and distance from VA >30 miles (23%). Following implementation of the Digital Divide consult, the percentage of tablet recipients who had a video visit within 90 days of tablet receipt rose from 57% in October 2020 to 74% in April of 2022. After consult implementation, we observed that many high-need patient populations had high rates of video visit engagement (i.e. ≥4 video visits in six months after tablet receipt) including 51%, 40%, and 42% of Veterans at high-risk for suicide, with substance use disorder, and with a hospitalization in the 90 days prior to tablet receipt, respectively. However, a moderate proportion (22-30%) did not have any video visits in the six months after tablet receipt. Among 1,000 survey respondents (response rate 35.7%), 91% reported that video visits helped them get care in a timely manner and 80% reported receiving the same quality of care during video-based and in-person visits

CONCLUSIONS: Although VA’s tablet initiative successfully targeted Veterans with complex medical and social needs during the pandemic, video visit rates remained low for many recipients. Future implementation strategies should focus on facilitating tablet adoption and use for high-value virtual care services.

NOVEL LARGE LANGUAGE ARTIFICIAL INTELLIGENCE MODEL ACCURATELY DIAGNOSES CLINICAL CASE VIGNETTES DESPITE NOT BEING DESIGNED FOR MEDICAL CARE

John T. Menchaca, Sara Turbow

Internal Medicine, Emory University School of Medicine, Atlanta, GA. (Control ID #3868723)

BACKGROUND: Patients are increasingly turning to the internet to research their health questions and concerns. The utility of online symptom checkers (ex. WebMD) has been limited by their deficits in accurately diagnosing and triaging medical conditions. Novel artificial intelligence (AI) language models trained with expansive textual data may provide a new means of facilitating self-diagnosis and triage for the general public. The goal of this analysis is to evaluate the diagnostic and triage accuracy of ChatGPT, a publicly available large language AI model designed to helpfully interact with users on a wide range of topics.

METHODS: Forty-five standardized patient vignettes previously used in the literature to evaluate online symptom checkers were provided verbatim to ChatGPT. The model was sequentially asked to provide 1) the most likely diagnosis, 2) the 5 most likely diagnoses, and 3) triage advice regarding urgency of care (emergent, non-urgent, self-care). The browser cache for the ChatGPT website was cleared between each vignette. Descriptive statistics were used to calculate percentages of correct responses across categories.

RESULTS: Of the 45 standardized patient vignettes, all vignettes provided symptoms with duration, 27% included comorbidities, and 78% included objective data such as vital signs. The most likely diagnosis provided by ChatGPT was the correct diagnosis for 89% (95% CI 76% to 96%) of vignettes. The top 5 most likely diagnoses included the correct diagnosis in 96% (95% CI 85% to 99%). ChatGPT accurately triaged the urgency of care in 15 of 15 (95% CI 78% to 100%) emergency cases, 14 of 15 (95% CI 68% to 100%) non-urgent cases, and 5 of 15 (95% CI 11% to 62%) self-care cases. All inaccurate triage recommendations for self-care cases were to seek non-urgent medical care. The single inaccurate triage for a non-urgent case recommended self-care and included a caveat about possible inaccuracy with an additional recommendation to discuss treatment with a healthcare professional.

CONCLUSIONS: A large language AI model was highly accurate identifying the correct diagnosis in 45 standardized patient vignettes despite not being designed for medical care. The model accurately triages conditions requiring emergency and non-urgent care, though is risk-averse in cases when self-care may be appropriate. These findings suggest that large language AI models may be a promising patient and provider resource in the future, particularly if finetuned for medical use. Many important steps must be taken prior to its practical use in healthcare, among them including 1) demonstration of accuracy in more varied and complex clinical cases and when responding to non-clinical language, 2) quantifying how risk-averse triage may affect healthcare utilization, and 3) better understanding patient and provider perceptions of such technology.

NUTRI: PRELIMINARY RESULTS FROM A FEASIBILITY PILOT TRIAL OF A CLINICAL DECISION SUPPORT FOR COLLABORATIVE DIET GOAL SETTING IN PRIMARY CARE

Brandon S. Altillo^3,1; Madalyn Rosenthal^1,2; William M. Tierney^5,4; Lola Okunade^6,1; Mariana Rendon Flores²; Eesha Nayak¹; Steven Andrews⁷; Marissa Burgermaster^1,2

¹Population Health, The University of Texas at Austin Dell Medical School, Austin, TX; ²Nutritional Sciences, The University of Texas at Austin, Austin, TX; ³Internal Medicine, The University of Texas at Austin Dell Medical School, Austin, TX; ⁴Indiana University School of Medicine, Indianapolis, IN; ⁵Indiana University Richard M Fairbanks School of Public Health, Indianapolis, IN; ⁶Lone Star Circle of Care, Georgetown, TX; ⁷University of Colorado Anschutz Medical Campus, Aurora, CO. (Control ID #3877284)

BACKGROUND: Diet is a major driver of chronic disease, but primary care providers (PCPs) often lack the time and training to adequately address it during visits. We developed a clinical decision support, Nutri, that guides PCPs through efficient, data-driven, collaborative, behavioral science-based nutrition goal setting with patients. Nutri collects patients’ diet history via an Automated Self-Administered 24-Hour Dietary Recall and suggests goals based on the USDA Dietary Guidelines for Americans.

METHODS: We tested Nutri’s feasibility in a pilot randomized controlled trial in 10 high-volume primary care clinics of a federally qualified health center network. We randomized PCPs to receive Nutri or control (usual care) stratified by training (MD/DO vs NP/PA) and self-reported need for interpreter for Spanish- speaking patients. Both groups completed a validated diet counseling competency scale and self-efficacy items at baseline. English- or Spanish-speaking patients ages 18-64 with BMI ≥25 kg/m² plus a diet-related comorbidity were recruited via text message sent to patients with upcoming appointments. PCPs randomized to Nutri were trained and received Nutri reminders prior to and at the time of each appointment. Nutri PCPs reported their satisfaction after each encounter using Nutri and assessed usability with the System Usability Scale (SUS) at the trial’s conclusion. Both groups completed 3-month diet counseling self-efficacy and competency posttests. We used t-tests to assess pre-post differences in self-efficacy and competency between Nutri and control PCPs.

RESULTS: We randomized 12 PCPs (Nutri: 4 MD/DO, 2 NP/PA; 2 Spanish-speaking; mean 11.5 years practicing; control: 3 MD/DO, 3 NP/PA; 1 Spanish-speaking; mean 3.7 years practicing). Nutri PCPs used Nutri with a mean of 3 patients each (range 1-7). PCPs completed the Nutri workflow in 100% of 18 eligible encounters. In 78% of encounters, the PCP and patient chose a Nutri-recommended goal. Mean PCP satisfaction post-encounter was 3.7±0.7 (5-point scale) and SUS score was 75±17.0 (above average). Table 1 describes competency and self-efficacy results by group.

CONCLUSIONS: PCPs successfully completed the Nutri workflow and reported moderate satisfaction and high usability ratings. Diet counseling competency significantly increased among Nutri PCPs compared to control and self-efficacy trended positively. Nutri can play an important role in supporting efficient, evidence-based dietary counseling in primary care.

PREDICTORS OF TELEMEDICINE USE FOR PATIENTS WITH DIABETES IN TWO HEALTH SYSTEMS

Elaine Khoong¹; Magdalene Kuznia⁴; Elizabeth B. Sherwin²; Kathryn E. Kemper²; Sarah Nouri³; Courtney R. Lyles¹

¹Medicine, University of California San Francisco, San Francisco, CA; ²Epidemiology and Biostatistics, University of California San Francisco, San Francisco, CA; ³Medicine, University of California, San Francisco, San Francisco, CA; ⁴School of Nursing, University of California San Francisco, San Francisco, CA. (Control ID #3877306)

BACKGROUND: Few studies have investigated how health system telehealth implementation impacts use of telemedicine for vulnerable patients. Safety-net systems provide more audio-only visits, and low-income, older, or minoritized patients are more likely to use audio-only visits, but the interplay between system- and patient-level factors is understudied. Therefore, we assessed if system telehealth priorities were associated with disparities in telemedicine use.

METHODS: We studied patients with diabetes (DM) empaneled as of 3/19 to an academic (site A) or safety- net (site S) system in the California-Hawaii region. Telemedicine visits were >95% video at site A and >95% audio-only at site S. Using EHR data, we assessed: (1) if patients had 1+ visit in a “hybrid” period from 7/20-3/21 when in-person & telemedicine visits were available; (2)whether patients had 1+ telemedicine (telephone/video) visit. We performed multivariable logistic regression analyses (adjusting for sex, insurance, comorbidities, and baseline DM control) focused on four predictors associated with digital health inequities: age, race/ethnicity, language, and socioeconomic status. To evaluate system-level factors, we assessed for interactions between site and primary predictors.

RESULTS: We studied 15148 patients (5268 site A; 9880 site S). 1+ visit: 10201 (67.3%) had 1+ visit in the hybrid period. Relative to 75+ yo, 18-34yo had lower odds of 1+ visit and 50-74yo had higher odds of 1+ visit. Black and non-English speaking had higher odds vs white and English-speakers, respectively. There were no significant interactions. 1+ telemedicine visit: 8305/10201 (81.4%) had 1+ telemedicine visit. Site S and younger age were associated with higher odds of 1+ telemedicine visit while Non-Hispanic Asian and Hispanic race/ethnicity were associated with lower odds. Younger patients at site S had lower odds than those

at site A of having 1+ telemedicine visit, but non-Hispanic Asian and Chinese-speaking patients at site S had higher odds than those at site A. (See table.)

CONCLUSIONS: After considering patient-level factors, system-level factors impact telemedicine access for patients with similar demographic traits. We found that a system that prioritized telephone visits was associated with higher telemedicine use for Asian or Chinese-speaking patients but lower use for younger patients compared to a system that prioritized providing video visits. Patient-level predictors of telemedicine use should be considered in the context of the system where they receive care.

PREVALENCE AND NATURE OF CLINICIAN DISBELIEF IN ADMISSION NOTES OF HOSPITALIZED PATIENTS

Courtney R. Lee^1,2; Xinwei Chen^1,2; Ari Klein¹; Davy Weissenbacher³; Graciela Gonzalez-Hernandez³; Gary E. Weissman^1,2

¹University of Pennsylvania, Philadelphia, PA; ²Leonard Davis Institute of Health Economics, Philadelphia, PA; ³Cedars-Sinai Medical Center, Los Angeles, CA. (Control ID #3877192)

BACKGROUND: In ambulatory settings, electronic health records (EHR) of Black patients are more likely to contain language used by clinicians to communicate disbelief in their clinical history. However, less is known about the use of such language in hospital settings where time constraints and fatigue are more likely to contribute to bias. To characterize the prevalence of language used by clinicians to communicate disbelief of a patient’s history in admission notes and examine whether the use of such language varied by patient race and ethnicity.

METHODS: We conducted a retrospective cohort study using the Medical Information Mart for Intensive Care (MIMIC-III, version 1.4) study, a deidentified EHR database from patients hospitalized in intensive care units (ICU) between June 1, 2001 and October 31, 2012. Because no applicable lexicon reflecting clinician disbelief has been identified, we applied the linguistic construct epistemic stance, which refers to a writer’s commitment to the certainty of information. Using this construct, we manually assembled a list of terms from published sources. We identified a total of 62 terms (e.g., reports, claims), and used regular expressions, an exact string-matching technique, to identify these terms within our sample. Three coders separately annotated a random set of 25 snippets for each term, then a consensus-based gold standard was used to confirm the presence or absence of epistemic stance. Terms were eliminated if they were considered correctly identified <80% of the time. Thirteen terms were included in the final analysis. The main outcome was a binary indicator for presence of at least one term in a note. The primary exposure was patient race and ethnicity (non-Hispanic White [NHW], non-Hispanic Black [NHB], Hispanic, Asian/Pacific Islander, or Other), which reflected self-report or a determination by the administrator who registered the patient. We used a mixed- effects logistic regression model to account for clustering of notes within admissions and admissions within patients. We adjusted for age, sex, preferred language, and insurance provider.

RESULTS: The sample included 17,551 admission notes written about 5,079 patients in 5,996 unique admissions. Of these patients, the mean (SD) age was 64.5 (18.0), 2348 (46.2%) were female, 3798 (74.8%) were non-Hispanic White, 565 (11.1%) were non-Hispanic Black, 188 (3.7%) were Hispanic, and 134 (2.6%) were Asian/Pacific Islander. In comparison to NHW patients, NHB patients had similar odds of containing at least one disbelief term (OR 1.17, 95% CI 0.95-1.43; p=0.72), as did Asian/Pacific Islander patients (OR 1.20, 95% CI 0.76-1.89); p=0.44), and Other racial/ethnic groups (OR 0.85, 95% CI 0.52-1.39; p=0.52)

CONCLUSIONS: While patient race and ethnicity have been linked to greater prevalence of clinician disbelief in the outpatient setting, this study suggests that the prevalence of disbelief does not vary by race and ethnicity among patients in the hospital.

PRIMARY CARE AND TELEHEALTH VISITS ACROSS ENGLISH LANGUAGE DURING THE COVID-19 PANDEMIC IN THE LOS ANGELES SAFETY NET

Alejandra Casillas¹; Anshu Abhat³; Griselda Gutierrez²; Carmen Mendez³; Stefanie D. Vassar⁴; Yu-Chuang D. Huang⁴; Arleen F. Brown⁵

¹Medicine, David Geffen School of Medicine @ UCLA, Los Angeles, CA; ²Medical Administration, Harbor-UCLA Medical Center, Torrance, CA; ³Internal Medicine, Harbor-UCLA Medical Center, Torrance, CA; ⁴Division of Internal Medicine & Health Service Research, UCLA, Los Angeles, CA; ⁵Medicine, University of California Los Angeles, Los Angeles, CA. (Control ID #3875314)

BACKGROUND: Telehealth has the potential to expand health care in safety net health systems—settings which disproportionately serve racial/ethnic minority, low-income, and/or Limited English Proficient (LEP) persons, and other patients who already experience social barriers to health care services.

At the outset of Coronavirus-19 disease (COVID-19), the Centers for Medicare and Medicaid Services granted payment for video and telephone visits. However, little is known about telehealth trends during the pandemic among safety nets. We describe this in the Los Angeles County Department of Health Services (LAC DHS)— the second largest safety net system in the US.

METHODS: We aggregated electronic health record data, comparing visit types (in-person, video, telephone) for unique patients with primary care visits during three study periods: pre-pandemic (Mar 2019- Feb 2020, n=161,087), pandemic Year 1 (Mar 2020-Feb 2021, n=164,752) and Year 2 (Mar 2021-Dec 2022, n=185,592). We examined differences between patients with English as primary language in the record, versus those with non-English (LEP).

RESULTS: In the pre-pandemic period, English and LEP patients had a mean 3.09 and 3.64 primary care visits, with no significant changes in Year 1 or 2. Figure 1 shows monthly primary care visits per 1,000 patients: in-person declined during Years 1 and 2, with increasing telehealth (combined telephone/video; video less than 1%) for both language groups. In the pre-pandemic period 18% (English) and 17% (LEP) of patients in primary care had a telehealth visit. In Year 1, this significantly changed to 89% (English) and 91% (LEP). In Year 2, 74% (English) and 76% (LEP) patients in primary care had a telehealth visit.

CONCLUSIONS: Despite COVID-19, the LA safety net delivered stable primary care visits. LEP patients, who already face barriers that negatively impact health outcomes—did not experience decreases. Telephone visits made up a significant portion of primary care for English and LEP groups after March 2020. One policy implication is that eliminating telephone visits as a billable visit would disproportionately affect underserved populations and threaten the ability of safety nets to meet need.

PROTECTION CONFERRED BY DELTA AND BA.1/BA.2 INFECTION AGAINST BA.4/BA.5 REINFECTION AND HOSPITALIZATION: A RETROSPECTIVE COHORT STUDY

Nicole Winchester¹; Nabin Shrestha³; Priscilla Kim²; Larisa G. Tereshchenko⁴; Michael B. Rothberg⁵

¹Cleveland Clinic Lerner College of Medicine of Case Western Reserve University, Cleveland, OH; ²Education, Cleveland Clinic Lerner College of Medicine of Case Western Reserve University, Cleveland, OH; ³Infectious Disease, Cleveland Clinic, Cleveland, OH; ⁴Department of Quantitative Health Sciences, Cleveland Clinic, Cleveland, OH; ⁵Department of Quantitative Health Sciences, Cleveland Clinic, Cleveland, OH. (Control ID #3868122)

BACKGROUND: SARS-CoV-2 immunity has declined with subsequent waves and accrual of viral mutations. In vitro studies raised concern for immune escape by BA.4/BA.5, and a study in Qatar showed moderate protection, but these findings have yet to be reproduced.

METHODS: This retrospective cohort study included individuals tested for COVID-19 by PCR during Delta predominance (July 1, 2021 to December 25, 2021), or BA.1/BA.2 (December 26, 2021 to June 24, 2022) and then retested during BA.4/BA.5 (June 25, 2022- August 18, 2022). The preventable fraction (PF) was calculated as the ratio of the infection/hospitalization rate for initially positive patients divided by infection/hospitalization rate for initially negative patients, stratified by age, and adjusted for age, gender, comorbidities, and vaccination status using logistic regression.

RESULTS: 20,987 patients were tested during Delta and/or BA.1/BA.2 and retested during BA.4/BA.5. Prior Delta infection provided no significant protection against BA.4/BA.5 infection (Adjusted PF: 11.9% (95% confidence interval [CI], 0.8-21.8); p=0.036) and minimal protection against hospitalization (Adjusted PF: 10.7% (95%CI, 4.9-21.7); p=0.003). In adjusted models, prior BA.1/BA.2 infection provided 45.9% (95%CI, 36.2-54.1) (p <0.001) protection against reinfection with BA.4/BA.5 and 18.8% (95% CI, 10.3-28.3) (p<0.0001) protection against hospitalization with BA.4/BA.5. Older adults (>65 years of age) derived greater protection from BA.4/BA.5 by prior BA.1/BA.2 infection than younger adults (<65) (p<0.001). Up- to-date vaccination also provided modest protection against reinfection with BA.4/BA.5 and hospitalization.

CONCLUSIONS: Prior infection with BA.1/BA.2 and up-to-date vaccination provided modest protection against infection with BA.4/BA.5 and hospitalization, while prior Delta infections provided minimal protection against hospitalization, and no protection against reinfection.

RELATIONSHIP BETWEEN THE CLINICAL LEARNING ENVIRONMENT AND INTERNAL MEDICINE RESIDENT CLINICAL REASONING DOCUMENTATION PRACTICES

Verity Schaye²; Matthew Haller²; Ilan Reinstein²; Daniel Sartori²; David J. DiTullio²; Erin Geraghty²; Kevin Hauck¹; Benedict Guzman²; Jesse Burk Rafel²

¹Medicine, NYU Langone Health, New York, NY; ²NYU Langone Health, New York, NY. (Control ID #3872088)

BACKGROUND: The clinical learning environment (CLE) – the organizational structures and cultures that trainees encounter – can have a significant impact on resident performance which can be difficult to quantify. We explore the relationship between the CLE at one academic internal medicine program and resident clinical reasoning (CR) documentation assessed by a previously developed machine learning (ML) model that classifies CR documentation as high- or low-quality based on a validated gold standard.

METHODS: We conducted a retrospective cross-sectional analysis of trends in resident CR documentation quality determined by the ML model among all general medicine admissions on resident teams (n = 12,605 admissions, n = 302 residents) across four academic years (AY) (July 2018-June 2022). Descriptive statistics and chi squared testing, including tests for linear trend, were performed with the R statistical package.

RESULTS: Resident CR documentation quality was lower during day shifts (59% high-quality vs 66% high- quality during night shift (8pm-7:59am), p<0.001) and quality decreased with increasing number of notes in the shift (note #1 in shift 65% high-quality vs note #5 in shift 55% high-quality, p=0.01). There was a trend in improvement by AY from 2018 to 2022 with a decrease in AY 2019 which included the initial COVID-19 surge in NYC (63% high-quality AY 2018 vs 58% high-quality AY 2019 vs 64% high-quality AY 2020 vs 67% high-quality AY 2021). CR documentation quality during the initial NYC COVID-19 surge was significantly lower than other times (51% high-quality March-May 2020 vs 65% high-quality other times during study period, p<0.001).

CONCLUSIONS: In this retrospective analysis of trends in resident CR documentation, aspects of the CLE such as day shift (vs night), volume of admissions in a shift, and the COVID-19 surge all were associated with lower quality CR documentation. We hypothesize that quality was higher during the night shift as residents can focus more on new admissions without the workload of routine non-emergent care. We observed an increase in note quality by AY and one hypothesis for this observation is there was an increase in number of resident teams and a decrease in team caps during this time period. Limitations of this work include its cross-sectional design, which precludes causal claims, and absence of controls for confounding by patient factors. Next steps will be to analyze the relationship between CLE and CR documentation quality in a nested regression model accounting for patient comorbidities. Our findings highlight the utility of using an automated ML model to explore the relationship between the CLE and resident performance measures given the ability to create large datasets for analysis in minimal time with ML which would not be feasible with human review. Retrospective analyses such as this one could inform studies with causal design to further understand impact of the CLE and inform quality improvement efforts that target hypothesized root causes.

ROLE AND IMPACT OF REAL-TIME PHARMACY BENEFIT CHECK TOOLS IN THE INPATIENT SETTING

Bradley D. Very¹; Priyanka Solanki²; Kristian Feterik³

¹Medicine, University of Pittsburgh Department of Medicine, Pittsburgh, PA; ²Department of Medicine, UPMC, Pittsburgh, PA; ³Department of Medicine, University of Pittsburgh School of Medicine, Pittsburgh, PA. (Control ID #3877034)

BACKGROUND: Real-time pharmacy benefit check (RTPBC) tools estimate outpatient out-of-pocket (OOP) prescription costs, and have been integrated into electronic health records (EHR). These components are underutilized, and little is known about their application in the inpatient setting.

METHODS: We surveyed 250 internal medicine trainees at an academic center to assess the importance of accessing OOP costs, tool awareness, and its effect on workflow. A system-based educational effort followed the pre-test with a tutorial and demonstration. Two months later, trainees completed a post-test. Results were assessed for significance with an unpaired Wilcoxon test.

RESULTS: There were 38 and 22 respondents in the pre and post-test groups, respectively. Trainees felt that access to outpatient OOP prescription costs is valuable, and considering cost is important when selecting medications continued after discharge. Tool awareness increased, with 26.3% of pre-test respondents reporting familiarity, compared to 50% in the post-test group. There was a statistically significant improvement in the ease of access to OOP costs (pre Z = 1.97, 95% CI, 1.49-2.39; post Z = 2.72, 95% CI, 2.04-3.40), and an increase in utilization (pre Z = 1.94, CI, 1.49-2.39; post Z = 2.72, CI, 2.04-3.40). There was a non-statistically significant increase in the ability to integrate OOP costs into workflow (pre Z = 2.42, CI, 2.08-2.76; post Z = 2.95, CI, 2.44-3.46). See Figure 1.

CONCLUSIONS: While trainees agreed that considering outpatient OOP prescription costs for hospitalized patients is valuable, the role of RTPBC tools remains unclear. Further research is needed to assess how system-based implementation of RTPBC tools can improve workflow, and evaluate the potential effect on clinical outcomes, including prescription abandonment and adherence.

THE IMPACT OF WIDESPREAD TELEHEALTH IMPLEMENTATION ON PATIENT-CLINICAL TEAM INTERACTIONS AMONG TWO COHORTS OF PATIENTS WITH DIABETES

Elaine Khoong¹; Sarah Nouri²; Elizabeth B. Sherwin⁵; Magdalene Kuznia⁶; Anna D. Rubinsky^3,5; Kathryn E. Kemper¹; Oanh K. Nguyen¹; Urmimala Sarkar⁴; Dean Schillinger¹; Courtney R. Lyles¹

¹Medicine, University of California San Francisco, San Francisco, CA; ²Medicine, University of California, San Francisco, San Francisco, CA; ³Medicine, University of California, San Francisco, San Francisco, CA; ⁴Medicine, UCSF, San Francisco, CA; ⁵Epidemiology and Biostatistics, University of California San Francisco, San Francisco, CA; ⁶School of Nursing, University of California San Francisco, San Francisco, CA. (Control ID #3875214)

BACKGROUND: COVID-19 changed primary care utilization. Utilization studies focus on billed visits, but care of patients with chronic diseases includes visits with other team members or between-visit interactions. These non-billed interactions are understudied and are more frequent in safety-net settings. We evaluated the impact of telehealth implementation on patient-clinical team interactions for patients with diabetes (DM) at an academic and safety net health system.

METHODS: Using EHR data, we studied visits (in-person, telehealth) and between-visit interactions (telephone calls, patient portal messages) for adults with DM empaneled to an academic (site A) or safety-net (site S) system in the California-Hawaii region. We included encounters with any member of the primary care team (e.g., clinician, pharmacist, social worker) and divided our study period into three: pre-COVID (4/19-3/20), strict shelter-in-place (4-6/20), hybrid (7/20-3/21). We assessed four outcomes reported as mean per person per month: all encounters; all visits; billed visits; between-visit interactions. Using an interrupted time series analysis adjusted for sex, insurance, comorbidities, and baseline DM control, we evaluated changes over time and reported marginal means to assess differences in each period by race/ethnicity, language preference, socioeconomic status (SES), and age. Due to telehealth implementation differences (telephone vs video visits), we analyzed sites separately.

RESULTS: Among the 15148 patients (5268 site A; 9880 site S), most encounters were between-visit interactions (72% site A; 59% site S). There was a significant growth in total encounters from pre-COVID to hybrid (site A: 1.14 to 1.21; site S: 0.84 to 1.00), which was driven at both sites by between-visit interactions. At site A, patient messages increased (0.30 to 0.47) vs telephone calls at site S (0.47 to 0.54). In contrast, all visits and billed visits significantly decreased at site A (0.35 to 0.30; 0.31 to 0.26). At site S, all visits increased (0.36 to 0.39), but billed visits were stable (0.26 to 0.27). When evaluating differences by demographic traits, we found fewer differences in visit frequency by age, race/ethnicity, language, and SES in the hybrid period vs pre-COVID. However, when evaluating between-visit interactions, non-English speakers had significantly fewer than English speakers in the hybrid period vs pre-COVID.

CONCLUSIONS: Telehealth implementation changed patient-clinical team interactions. Between-visit interactions grew more than visits, but the modality of the interaction diverged in our cohorts (patient messages vs telephone). While differences in visit usage by race/ethnicity, language, age, and SES closed compared to pre-COVID, differences in between visit interactions by language preference persisted. To fully understand care utilization for patients with chronic diseases, studies should expand beyond billed visits, particularly since disparity patterns differ in billed visits vs between-visit interactions.

USE OF VIRTUAL SCRIBES ASSOCIATED WITH SIGNIFICANT DECREASES IN ELECTRONIC HEALTH RECORD TIME BY AMBULATORY PHYSICIANS

Lisa Rotenstein³; Jianyi Zhang²; Christine Iannaccone¹; Christopher Holland²; Michael Healey³; David Y. Ting⁴; Christine Sinsky⁵; David W. Bates²

¹General Internal Medicine, Brigham and Women's Hospital, Boston, MA; ²Brigham and Women's Hospital, Boston, MA; ³Medicine, Brigham and Women's Hospital, Belmont, MA; ⁴Medicine, Massachusetts General Hospital, Lexington, MA; ⁵Professional Satisfaction, American Medical Association, Chicago, IL. (Control ID #3873893)

BACKGROUND: Ambulatory physicians spend much of their days using the electronic health record (EHR) and on EHR-based documentation. Virtual documentation support, such as virtual scribes, has the potential to enhance the physician experience while reducing the barriers associated with in-person documentation assistance. We characterized the impact of virtual scribes on physician EHR use patterns at two large academic medical centers.

METHODS: We identified physicians who had used a virtual scribe at Brigham and Women’s Hospital or Massachusetts General Hospital (MGH) between 01/2020 and 09/2022. For each physician, we extracted data about EHR use patterns from the Epic Signal database.

We assessed physicians’ specialty and hospital affiliate, as well as whether they used a synchronous vs. asynchronous virtual scribe service. We then used paired t-tests to compare EHR time metrics pre- and post- virtual scribe use at a physician level. We compared each physician’s total EHR time per appointment, pajama time (5:30 PM-7 AM) per appointment, note time, percent of note written by physician, average note length, and percent of orders with team contribution in the three months pre- and post- use of virtual scribes.

RESULTS: Our sample included 116 physicians (89% from MGH). Scribe users were distributed across specialties, with 22% internal medicine, 17% family medicine, and 11% dermatology physicians.

On a per-physician basis, virtual scribe use was associated with significant decreases in EHR time per appointment, note time per appointment, and pajama time per appointment in the three months post versus the three months prior to virtual scribe use. Total average EHR time was 39.3 minutes pre-scribe use vs. 32.6 minutes post (p=0.02), average note time per appointment was 10.8 minutes pre-scribe use and 9.1 minutes post-scribe use (p<0.001), and pajama time per appointment was 10.1 minutes pre-scribe use vs. 8.6 minutes post-scribe use (p<0.001). The proportion of the note contributed by the physician significantly decreased with scribe use (80% prior to scribe use versus 50% after).

Note length significantly increased with scribe use (mean 6395.4 characters post-scribe vs 5953.0 characters pre-scribe use; p<0.001). There was no significant change in the proportion of orders with contribution from another member of the care team with scribe use.

CONCLUSIONS: In this longitudinal study across two academic medical centers, we demonstrate that use of virtual scribes is associated with significant decreases in multiple EHR time metrics, and slight increases in note length. Future studies should segment analyses by whether scribe offering is synchronous vs. asynchronous, qualitatively characterize physician and patient experiences with virtual scribes, and assess the quality of scribed notes.

VIRTUAL REALITY FOR CHRONIC PAIN IN A COMMUNITY WITH LOW ACCESS TO TECHNOLOGY: A USABILITY AND PILOT STUDY

Lesly Sanchez Alvarez¹; Genevieve Bryant²; Hajar Traiba¹; Hector R. Perez¹

¹Division of General Internal Medicine, Albert Einstein College of Medicine, Bronx, NY; ²General Internal Medicine, Montefiore Health System, Bronx, NY. (Control ID #3876048)

BACKGROUND: Evidence suggests virtual reality (VR) technology may be beneficial for reducing pain symptoms in patients with chronic pain. However, studies to date have yet to recruit from communities with low access to technology. We sought to conduct a usability study and to pilot test a VR device among participants recruited from a community with low access to technology.

METHODS: We conducted in-person usability testing and pilot testing of a VR device in patients with chronic pain in the Bronx, NY, whose inhabits have among the lowest access to technology in New York State. Participants were enrolled if they were English-speaking adults with at least moderate chronic pain. Participants were excluded if they had contraindications in using the VR headset, including epilepsy. Participants were recruited from electronic medical record databases and from prior pain clinical trials. The study utilized a VR device marketed for chronic pain and used in prior clinical trials, the RelieVRx (AppliedVR, Van Nuys, CA). The RelieVRx uses immersive mindfulness and cognitive behavioral interactive visuals and VR activities, including breathing exercises with visual cues, to facilitate improved pain. First, to measure usability, participants were asked to operate the device and researchers compiled notes on usability tasks (turning on/off the device, fitting the headset, operating the machine, completing activities). We rated performance on each usability task on a 0-3 point scale, with values 1 and below considered usability barriers; we report proportions here. Lastly, in our pilot study, participants completed VR activities

for 10 minutes and reported pain intensity (using a 10-point Likert scale) and mood (using the Brief Mood Introspection Scale) both before and after. Paired t-tests were used to calculate differences in pain intensity and mood before and after VR testing.

RESULTS: We recruited 16 participants (mean age: 58.2 years; female: 75%; non-Hispanic Black: 62.5%; high school education or lower: 53%). Almost all (94%) had never used VR. On usability tasks, participants had little difficulty with most tasks; where there were issues, these were easily remedied. While only 38% had no difficulty turning on the device, 87% of participants were able to fit and operate the device, and 100% had no difficulty completing VR activities. On pilot testing, participants reported large reductions in pain intensity after VR activities (32.0% mean reduction in pain intensity, p=0.005). Feelings of liveliness and calmness increased after VR activities (30% improvement, p=0.006; 16.2% improvement and p=0.05, respectively), while feelings of tiredness decreased (31.8% reduction, p = 0.01).

CONCLUSIONS: In this study of VR for chronic pain in a sample of patients recruited from a community with low access to technology, we found no major usability issues and large effects on pain and mood after one session. Further research can determine whether effects seen in this study are durable.

Scientific Abstract - DEI, Health Equity, and Social Determinants of Health

"SHIFTING THE CULTURE AND THE WAY THAT WE PRACTICE": PERINATAL CLINICIANS' COGNITIVE AND BEHAVIORAL CHANGES IN RESPONSE TO ANTIRACISM AND ANTIBIAS INTERVENTIONS

Sarah B. Garrett¹; Fiona Miller⁵; Linda Jones⁴; Julie Harris⁴; Breezy Powell⁴; Erica Chan⁶; Stephen Zamarripa¹; Daniel Dohan¹; Melissa Simon²; Brittany D. Chambers³

¹Philip R. Lee Institute for Health Policy Studies, University of California San Francisco, San Francisco, CA; ²Northwestern University Feinberg School of Medicine, Chicago, IL; ³Department of Human Ecology, University of California Davis, Davis, CA; ⁴California Preterm Birth Initiative, University of California San Francisco, San Francisco, CA; ⁵University of California San Francisco School of Medicine, San Francisco, CA; ⁶University of Colorado Anschutz Medical Campus School of Medicine, Aurora, CO. (Control ID #3874037)

BACKGROUND: Many health systems are implementing interventions to address racism and bias in maternal healthcare. There is little research on their effects. We characterize how perinatal clinicians report they and colleagues have changed following antiracism/antibias interventions.

METHODS: As part of larger study about antibias training, we conducted in-depth interviews with perinatal clinicians (n=20) at two hospitals in Northern California. Using inductive/deductive thematic analysis, we developed a taxonomy of described changes.

RESULTS: Respondents were nurse midwives (6), physicians (6), registered nurses (5), and other staff (3). They self-identified as Black (4), multiracial (4), or white (12) women; two identified as Latinx or Hispanic. Twelve respondents discussed changes in themselves or colleagues following interventions such as antiracism talks, antibias trainings, and colleagues’ requests to reform biased behavior. We characterized these as: 1. Cognitive changes, including: Recognizing one’s own biased thinking and behavior (“It was a very subtle shift for me…Instead of saying, ‘I’m not a racist, I don’t have biases,’ saying, ‘Yes, you do, and that’s okay. Let’s acknowledge them… and then you can start shifting them’”); and better understanding the role of race and racism in disparities (“It’s made a deep impression on me to learn some of the data about the discrepant perinatal outcomes… It's not education and income. Blackness, that makes the difference in poor outcomes, and in not believing women”). 2. Individual behavior changes, including: Evaluating whether one is providing unbiased care (“I think for an extra second about, like, would I treat this patient differently were they to be more similar to me”); working to change biased speech and behavior (“[Colleague] told me about the things people have said that are actually showing levels of racism… I’m trying to un-train myself out of it because of the racial connotations”); communicating concern about colleagues’ biased speech or behavior (“If I hear something I definitely say something”); and identifying ways to better support patient autonomy. 3. Team behavior changes, including: More intra-colleague discussion of racism, bias, and respectful care; and collaborating to assess and optimize treatment for individuals in racial/ethnic groups with higher rates of perinatal complications (“The doctor stopped, ‘Is there anything that we’re not covering to make sure that we’re decreasing her risk?’ I hadn’t heard a doctor verbalize it before”).

CONCLUSIONS: Perinatal clinicians observe cognitive, behavioral, and individual- and team-level changes following antiracism/antibias interventions, some of which exceed mere allyship. Future research should investigate which type(s) of changes benefit the care and outcomes of patient populations disproportionately burdened by maternal health inequities, and identify which interventions, in which contexts, produce these.

DIVERSITY, EQUITY, AND INCLUSION PROGRAMS AT US MEDICAL SCHOOLS: A NATIONAL SURVEY

Melinique Walls²; Isra Hasnain²; Monica Vela³; Paulina Zajac³; Norma Poll-Hunter⁴; Wei Wei Lee¹.

¹General Internal Medicine, University of Chicago, Chicago, IL; ²University of Chicago Pritzker School of Medicine, Chicago, IL; ³Medicine, University of Chicago Pritzker School of Medicine, Chicago, IL; ⁴Association of American Medical Colleges, Washington, DC. (Control ID #3876866)

BACKGROUND:A diverse physician workforce strengthens health equity via improved healthcare access for underserved patients. Despite efforts to increase diversity at US medical schools, Black, Hispanic, and American Indian students remain underrepresented in medicine and are more likely than their white peers to experience mistreatment and bias. While many schools have diversity, equity, and inclusion (DEI) offices, little is known about the scope of these offices.

METHODS: Between Aug 2021-Feb 2022, 154 US medical schools were surveyed about their DEI offices. The survey assessed the presence of a DEI office, characteristics of leaders, budgets/FTE support, and programming. Chi square and Mann-Whitney U tests explored associations between budget, and FTE for staff and faculty by school type (public vs. private), size and diversity (%URM students enrolled).

RESULTS:Of 154 schools,103 responded (67%). 92% reported having a DEI office. Of 64 schools reporting an annual budget, 28% reported <$50,000, and the maximum was >$500,000; there was no difference by school type (p=0.45) and diversity (p=0.80).

Of 94 respondents who reported demographic data, 71% were faculty, 47% were deans, and 11% were directors. Of 61 respondents who identified as the “DEI leader,” 68% were women, 59% were Black, 69% were non-Hispanic, and 39% were in their roles for <5 years. Median FTE for DEI leaders was 0.50 (IQR 0.5-1.0) and 3.0 (IQR 2.0-4.5) for all DEI faculty/staff. Median faculty/staff FTE per URM student was 0.03 (IQR 0.01-0.05) and while no differences by school type (p=0.13) or diversity (p=0.40) were found, higher median FTE per URM student was found at smaller (<590 students) vs. larger schools (0.04 [IQR 0.02-0.07] vs. 0.02 [IQR 0.01-0.03], p=0.01) and at private vs. public schools (0.03 [IQR 0.02-0.06] vs. 0.02 [IQR 0.01-0.04], p=0.04). Schools with <20% URM students had higher FTE per URM student than those with greater diversity (0.04 [IQR 0.02-0.07]) vs. 0.02 [IQR 0.01-0.04], p=0.002). Median faculty/staff FTE per enrolled student was 0.005 FTE (IQR 0.003-0.009).

Of 95 respondents, most DEI offices promoted an inclusive culture (92%), advanced diversity of the medical profession (87%), and supported community service (71%). Of 81 responding schools, 31% evaluated their DEI office, with most using the AAMC graduate questionnaire diversity items (76%) and internal surveys of DEI events (68%).

CONCLUSIONS: To our knowledge, this is the first study to describe characteristics of DEI offices at US medical schools. Almost all responding schools had DEI offices, which were staffed by approximately 3 FTEs and approximately half had annual budgets of up to $100,000 to support a range of student and community based initiatives. Smaller schools, private schools, and those with less diversity devoted more FTE per URM student than larger, public, and more diverse schools. Few schools evaluate their DEI programming and we recommend more rigorous evaluation to guide resource allocation and strategic planning.

2022 PREVALENCE OF BURNOUT AMONG WOMEN RESIDENTS AND ASSOCIATION WITH DISCRIMINATION-BASED TRAUMA: A MULTI-INSTITUTIONAL STUDY

Tyra Fainstad¹; Adrienne Mann¹; Alexander Heilman¹; Pari Shah²; Christine D. Jones³; Lotte Dyrbye³

¹Internal Medicine, University of Colorado, Denver, CO; ²Graduate School of Social Work, University of Denver, Denver, CO; ³Medicine, University of Colorado Denver School of Medicine, Aurora, CO. (Control ID #3874489)

BACKGROUND: The Covid-19 pandemic has precipitated dramatic increase in physician burnout. Although physician burnout has been well documented, the current state of burnout among female residents remains underexamined.^1-3 Recent data shows mistreatment/discrimination by patients, families, and visitors in the workplace⁵ increases the odds of burnout among practicing physicians. Single specialty studies in residents suggest a similar relationship, but an understanding of the relationship between past experiences of discrimination outside of the workplace and burnout risk among residents is lacking. The purpose of this study is to describe the current prevalence of and risk factors for burnout amongst female physician trainees across multiple institutions and specialties.

METHODS: In September 2022, a baseline survey was administered to 1,017 female trainees who had volunteered to participate in a professional coaching program², across 26 GME training programs. The survey included demographics, the Trauma Symptoms of Discrimination Scale (TSDS),⁴ and the Maslach Burnout Inventory (MBI). The TSDS is a 21-item self-report measure that screens for trauma symptoms due to past discriminatory events. The scale includes questions about current symptoms including avoidance, negative cognitions, social fears, and worries about the future. Items are rated from 1 (never) to 4 (often) based on the amount of distress caused specifically by prior discriminatory acts.⁴ Burnout was defined as scoring >27 for emotional exhaustion (EE) and/or >10 for depersonalization (DP) on the MBI.

RESULTS: All 1,017 trainees responded to the baseline survey. The mean age was 31 years, 88% (843)

identified as heterosexual, 59.9% (573) identified as White. Most (81%) were non-surgical, and 20.7% (207), 19.8% (198) and 59.5% (596) were PGY-1, PGY-2, and >PGY-3, respectively. Overall, 66% had burnout, 66.0% had high EE, and 61.3% had high DP. Mean EE and DP scores were 30.58 and 11.83. Higher scores on the TSDS was associated with higher odds of burnout (OR, 1.30 for 10 units of change, 95% CI 1.11-1.53, p=0.001), high EE (OR 1.26 for 10 units of change, 95% CI 1.11-1.43, p<0.001), and high DP (OR 1.12 for 10 units of change, 95% CI 1.00-1.26, p=0.058). Multivariable analysis exploring independent predictors of burnout will be presented in the meeting.

CONCLUSIONS: Data from this large, multi-institutional, multi-specialty cohort show a higher level of burnout than reported in pre-pandemic large studies of US residents.^1,3 This study suggests that trauma from previous discrimination experiences increase burnout risk in female residents. Burnout among trainees is progressing. In addition to a GME-wide national wellbeing effort focused on system-level drivers, initiatives to help those who have suffered trauma from discrimination inside and outside the workplace may be useful. Future studies should explore how best to help trainees recover from previous experiences of trauma.

ACCESS TO CONTINUOUS GLUCOSE MONITORING SYSTEMS IN A RESIDENT RUN PRIMARY CARE CLINIC FOR UNDERSERVED MINORITY PATIENTS WITH TYPE 2 DIABETES: A PILOT STUDY

Rebecca G. Sgroi¹; Srikala Gumireddy¹; Wan Ying Tan¹; Leah Roseph¹; Diana Salama²; Rebecca Kaplan³; Yolande Milambwe³; Cynthia R. Feher¹; Nancy Rennert^4,51Internal Medicine, Norwalk Hospital Department of Medicine, Norwalk, CT; ²University of Vermont Larner College of Medicine, Burlington, VT; ³Primary Care Clinic, Norwalk Hospital, Norwalk, CT; ⁴Chief of Endocrinology, Norwalk Hospital, Norwalk, CT; ⁵Associate Clinical Professor, Yale School of Medicine, New Haven, CT. (Control ID #3868330)

BACKGROUND: Disparities in diabetes care are well documented. Continuous Glucose monitoring Systems (CGMs) have improved patient experience without the need for finger sticks and have been shown to effectively lower Hemoglobin A1C, decrease the number of hypoglycemic events and increase the serum glucose time in range. CGMs may also decrease clinical inertia and allow improved diabetes management. Medically underserved and historically marginalized persons are disproportionately affected by diabetes, yet limited data exist on use of CGMs in this population. We report a pilot initiative that brought CGM access (FreeStyle Libre 2) to our Federally Qualified Community Health Center (FQHC)’s resident physician primary care clinic. Our patients face unique challenges, as most are underinsured (with 40% having no insurance) and 50% report a preferred language other than English.

METHODS: We created a group of Clinical Champions (Resident Physicians, Attending physicians, Registered Nurses and Medical Assistants) to develop a Team Workflow and education plan for all clinicians and staff including CGM placement, trouble shooting and data interpretation. Patients with Diabetes Type 2 and an A1C >8.0% or on insulin were referred to be enrolled in CGM. Exclusion criteria were renal failure and pregnancy. The study visit schedule included an initial nurse visit for patient education and device placement of FreeStyle Libre (FSL). Follow up medical visits were scheduled at 2 weeks, 6-8 weeks, and 3-4 months where glycemic data was reviewed and shared decision making took place to make medication and behavioral changes. Baseline HBA1c was collected and repeated at the 3–4-month visit. Provider and patient surveys were distributed at each visit.

RESULTS: Fifty participants were enrolled in the CGM study and had FSL placed. The CGM active time was greater than 50% in 76% of participants and those who continued with CGM had an increase in active time. Initial mean A1C for the entire cohort was 9.64% and average glucose after 2 weeks was 195 mg/dl. Hypoglycemia, especially overnight, reduced over time. The use of CGM data was associated with clinician- initiated medication changes. Provider and patient surveys showed positive sentiment to using the device and revealed an increase in likelihood of medication changes based on glycemic trends. In addition, patients reported better understanding of their disease process overall.

CONCLUSIONS: This pilot study demonstrates the feasibility of utilizing diabetes technology (FreeStyle Libre 2) in underserved patients who often lack access and support required to manage their diabetes. CGM is becoming more common in outpatient settings to determine glycemic trends and make appropriate medication recommendations. This study shows that CGM can be successfully deployed during Primary Care visits by physicians in training who will likely continue to use CGM well beyond their training.

ACHIEVING EQUITY IN SCREENING MAMMOGRAPHY AMONG IMMIGRANT EAST AFRICAN WOMEN IN SEATTLE: ONE SIZE WILL NOT FIT ALL

Rebecca S. Gold¹; Rachel L. Yung³; ADELAIDE H. MCCLINTOCK²

¹Internal Medicine, University of Washington, Seattle, WA; ²Medicine, University of Washington, Seattle, WA; ³Oncology, University of Washington, Seattle, WA. (Control ID #3874130)

BACKGROUND: Compared to white women, Black women are diagnosed with breast cancer at advanced stages and have higher mortality rates. Decreased mammography utilization significantly contributes to breast cancer disparities. Immigrant Black women face additional barriers that compound these inequities; notably, mammography rates are among the lowest for recent immigrants. We aimed to describe experiences and preferences related to breast cancer screening among Somali and Ethiopian women, two growing immigrant populations in Seattle, to inform system-wide interventions to increase screening rates within a diverse patient population at a large, urban, academic medical center.

METHODS: Semi-structured focus groups with women who self-identified as Somali or Ethiopian, eligible for breast cancer screening in Seattle, WA. Each focus group discussed participant views on health care, breast cancer screening, and future outreach. Interviews were conducted in Somali or Amharic with a real- time certified bilingual interpreter and cultural mediator. Verbatim transcripts from audio recordings underwent constant comparison and iterative data reduction analysis by two authors.

RESULTS: 11 Somali women participated in 2 focus groups and 7 Ethiopian women participated in 2 focus groups. Several themes were shared between Somali and Ethiopian women: obstacles to screening (e.g., physical location of services, need for interpreters, competing caregiving responsibilities), the power and value of a doctor’s recommendation, the weight of information from within their own community, and the motivation to screen when hearing stories of people with cancer. Both groups recommended more presence in community outreach and recommended oral or visual materials as the primary mode of information sharing. Themes more common among Somali women included an expectation of pain, a cultural tradition of silence around health issues, and an emphasis on the role of religion in health (e.g., predestination). Ethiopian women described how although faith and religion including prayer were important in their lives or for coping with health issues, they desired more separation between religion and healthcare. Ethiopian women also described a desire for more information about breast cancer and screening and may place greater emphasis on the provider as a trusted source of information.

CONCLUSIONS: This study identified obstacles to screening and opportunities for developing tailored, culturally-relevant interventions to improve rates among East African communities in our region. Healthcare systems must be prepared to address the unique needs of both populations for interventions to meaningfully impact screening rates and ultimately reduce disparities in breast cancer outcomes.

ADAPTATION OF A BIDIRECTIONAL CRISIS AND EMERGENCY RISK COMMUNICATION FRAMEWORK BY COMMUNITY ENGAGED RESEARCH PARTNERSHIPS IN RURAL MISSISSIPPI DURING THE COVID-19 PANDEMIC

Rodney Washington⁷; Jane Njeru²; Lohr M. Abby²; Jennifer A. Weis⁹; Angela Omondi⁸; Mauda Monger⁴; Sandra Melvin⁵; Nakeitra Burse³; Caroline Compretta⁶; Irene G. Sia¹; Mark L. Wieland²

¹Division of Public Health, Infectious Diseases, and Occupational Medicine, Mayo Clinic Minnesota, Rochester, MN; ²Division of Community Internal Medicine, Geriatrics, and Palliative Care, Mayo Clinic Minnesota, Rochester, MN; ³Six Dimensions, LLC, Ridgeland, MS; ⁴My Brother’s Keeper, Inc, Ridgeland, MS; ⁵Institute for the Advancement of Minority Health, Jackson, MS; ⁶Department of Preventive Medicine, The University of Mississippi Medical Center, Jackson, MS; ⁷School of Population Health, The University of Mississippi Medical Center, Jackson, MS; ⁸Department of Behavioral and Environmental Health, School of Public Health, College of Health Sciences, Jackson State University, Jackson, MS; ⁹Center for Clinical and Translational Science, Mayo Clinic Minnesota, Rochester, MN. (Control ID #3874056)

BACKGROUND: Crisis and emergency risk communication (CERC) frameworks have been used to promote public participation in coronavirus 2019 (COVID-19) mitigation efforts, but equitable implementation depends on reaching those who have been disproportionately affected by COVID-19 disparities. African American and Hispanic populations as well as immigrant and refugee populations have experienced disproportionate COVID-19 incidence, hospitalizations, and deaths due to pandemic-associated environmental and socioeconomic factors framed by structural racism. Community engagement is important for reaching populations at risk for health inequities in the COVID-19 pandemic.

METHODS: In March 2020, a community-engaged research partnership in Southeast Minnesota adopted a CERC framework to address COVID-19 prevention, testing, and socioeconomic impacts within health disparity groups. Bidirectional communication between Communication Leaders (trusted leaders in the community selected by community partners for their extensive social connections) and their social networks collaborated with the partnership to refine messages, leverage resources, and advise policy makers. This community-engaged risk communication intervention demonstrated high acceptability, feasibility, perceived efficacy, and sustainability. The framework was adapted by a community-academic partnership with rural African American populations in three Mississippi counties with high COVID-19 disparities. Intervention reach was assessed by the number of messages delivered by Communication Leaders to members of their social networks. Perceived scalability of the intervention was assessed by the Intervention Scalability Assessment Tool. Bidirectional communication between Communication Leaders and community members within their social networks was used by the partnership to refine messages, meet resource needs, and advise statewide decision makers.

RESULTS: Communication Leaders in all three counties used a variety of platforms to share information, reaching more than 8,482 individuals during the first 3 months, in the three counties. The intervention was deemed to be highly scalable by partnership members. Scalability was perceived to be facilitated by the magnitude and urgency of the problem, the high perceived efficacy of the intervention that is adapted for local contexts, high acceptability through the tailoring process and co-ownership by community partners, and the relatively low start-up costs of partnering with existing communication leaders in each region.

CONCLUSIONS: Adaptation of a community-engaged pandemic CERC intervention is feasible and scalable, and it has the potential to reduce COVID-19 inequities across heterogeneous populations. This approach may be incorporated into current and future pandemic preparedness policies for community engagement.

ADDRESSING 30-DAY READMISSION RATES FOR PNEUMONIA: WHAT ARE THE RISK FACTORS?

Elizabeth M. Jean-Marie, Kaniza Z. Abbas, Willie L. McClure, Jihane Faress

Medicine, University Hospitals, Cleveland, OH. (Control ID #3874490)

BACKGROUND: Pneumonia is a severe health problem in the United States (US), responsible for over a million annual hospital admissions and more than 140,000 annual hospital readmissions. Studies show that approximately 1 in 5 patients with pneumonia are readmitted to the hospital within 30 days of discharge. This leads to a severe burden on the healthcare system, both in terms of resources and costs. The goal of this study was to assess factors associated with 30-day readmission for patients admitted with pneumonia our Northeast Ohio hospital system.

METHODS: We obtained data from patients in our healthcare network through the EMR. Using 1CD-10 codes, we generated a list of patients with pneumonia as their admission diagnosis between September 2020 and August 2021. We then filtered the list to include patients with an additional encounter in the system within the next 30 days (an admission, ED visit, or observation). Patients without an additional encounter were compared as a control group. The medical charts for these patients were reviewed for information on demographics, comorbidities, readmission diagnosis, discharge, and follow-up appointments. These data points were collected and manually input into a secure RedCap database.

RESULTS: A total of 296 patient encounters were reviewed in our healthcare system–96 encounters for our 30-day readmission group (a rate of 32%) and 200 encounters for a comparison group. Readmission/ED visits were largely within 7 days (36, 42%) of discharge. Of the readmission encounters reviewed, only 19 were related to pneumonia (19.7%), while over 80% returned to the hospital for other reasons. Patients who were readmitted when compared to the control group were most likely to be Black (76%, 60%, respectively) and female (58%, 47%, respectively). The most common comorbidity in the readmission group was COPD (38.7%, 9.4%, respectively), whereas in the control group it was malignancy (33.9%, 53.7%, respectively).

CONCLUSIONS: At our large hospital network, nearly 1 out of 3 of our patients admitted for pneumonia return to the hospital within 30 days. Our study highlights at risk populations for pneumonia readmission at our institution are those who are Black, Female and/or have COPD. Special care and follow-up focused on pneumonia patients who fall into these high risk populations might result in decreased post-hospital readmissions.

ADVANCING HEALTH EQUITY THROUGH PARTNERSHIPS OF STATE MEDICAID AGENCIES, MEDICAID MANAGED CARE ORGANIZATIONS, AND HEALTH CARE ORGANIZATIONS

Anna L. Thorndike⁷; Lauren Peterson¹; Sivan Spitzer²; Shilpa Patel³; Anne Smithey³; Jennifer Moore^5,4; Marshall H. Chin⁶

¹Crown Family School of Social Work, Policy, and Practice, University of Chicago, Chicago, IL; ²Azrieli Faculty of Medicine, Bar-Ilan University, Tel Aviv, Israel; ³Center for Health Care Strategies Inc, Hamilton, NJ; ⁴Department of Obstetrics & Gynecology, University of Michigan, Ann Arbor, MI; ⁵Institute for Medicaid Innovation, Washington, DC; ⁶Department of Medicine, University of Chicago Division of the Biological Sciences, Chicago, IL; ⁷University of Chicago Pritzker School of Medicine, Chicago, IL. (Control ID #3859661)

BACKGROUND: To reduce health inequities, care transformation to address medical and social needs, especially for marginalized populations including people with Medicaid insurance, should be supported and incentivized by payment reform methods. Further, collaboration across government agencies, managed care organizations, health care organizations, and communities is essential to overcome health system fragmentation and implement sustainable reform. This study examines how multi-stakeholder teams operationalized the Roadmap to Advance Health Equity model to (a) build cultures of equity and (b) integrate health equity into care transformation and payment reform initiatives.

METHODS: Advancing Health Equity: Leading Care, Payment, and Systems Transformation is a national program funded by the Robert Wood Johnson Foundation that brings together multi-stakeholder teams to design and implement initiatives to advance health equity. Each team consists of representatives from state Medicaid agencies, Medicaid managed care organizations, and health care organizations in seven U.S. states (DE, IL, ME, NJ, PA, TN, WA). Through didactic and technical assistance support, teams were encouraged to implement the Roadmap to Advance Health Equity model to build a culture of equity, identify a health equity focus, diagnose root causes of the health equity focus, implement care transformation, and reform payment to advance health equity. In 2021, semi-structured interviews were conducted with representatives (n=23) from all seven teams about experiences implementing the Roadmap. Interviews were coded for themes using NVivo Qualitative Data Analysis software.

RESULTS: Facilitators of building cultures of equity included 1) build upon preexisting intra-organizational cultures of equity, 2) recruit and promote diverse staff, and 3) train staff on health equity and anti-racism. Teams faced challenges building inter-organizational cultures of equity. Facilitators of identifying a health equity focus area and its root causes included 1) use data to identify health equity focus, 2) overcome stakeholder assumptions about inequities, and 3) partner with Medicaid members and community. Facilitators of implementing care transformation and payment reform were 1) tailor preexisting care delivery and payment structures to focus on health equity, 2) partner with Medicaid members and individual providers, and 3) incentivize equitable care transformation and payment reform. Facilitators of sustainability planning included 1) identify evidence of improved health equity focus and 2) maintain relationships among stakeholders. Teams faced challenges determining the role of the state Medicaid agency.

CONCLUSIONS: Multi-stakeholder teams shared practical strategies for implementing the Roadmap to Advance Health Equity that can inform future efforts to build intra- and inter-organizational cultures of equity and integrate health equity into care delivery and payment systems for Medicaid programs.

ADVERSE FINANCIAL OUTCOMES BEFORE AND AFTER COVID-19 INFECTION

Nora V. Becker¹; Carlton F. Erin⁶; Theodore Iwashyna²; John W. Scott⁴; Michelle H. Moniz³; John Z. Ayanian^5,1

¹Internal Medicine, University of Michigan, Ann Arbor, MI; ²Johns Hopkins University, Baltimore, MD; ³Obstetrics and Gynecology, University of Michigan, Ann Arbor, MI; ⁴Surgery, University of Michigan, Ann Arbor, MI; ⁵Institute for Healthcare Policy and Innovation, University of Michigan, Ann Arbor, MI; ⁶Pediatrics, University of Michigan, Ann Arbor, MI. (Control ID #3874518)

BACKGROUND: Survivors of COVID-19 infection may experience financial hardship, and credit outcomes are increasingly recognized as a powerful way to assess patients’ financial wellbeing. Our objective was to compare credit outcomes between two cohorts of COVID-19 survivors, the first with credit outcomes measured before their infections, and the second with credit outcomes measured after their infections.

METHODS: We utilized medical claims from adult enrollees in a large state-level commercial insurance network linked to their Experian credit report data in January 2021. We identified two cohorts of COVID-19 survivors: a “post-infection” cohort with infections diagnosed during March-July 2020 (credit outcomes observed ≥6 months after infection), and a “comparison” cohort with infections diagnosed during February- October 2021 (individuals with credit outcomes observed prior to infection). The comparison cohort included individuals who had a COVID-19 infection, but whose credit outcomes could not have been affected by their infection because these outcomes were measured prior to infection. All patients with identified COVID-19 infections were characterized as either non-hospitalized or hospitalized for COVID-19.

Logistic regression models were used to compare financial outcomes between the post-infection and comparison cohorts. Primary outcomes included having medical debt in collections, non-medical debt in collections, or a low credit score (<660, range 300-850). All models adjusted for age group, gender, COVID-19 hospitalization, and four zip-code level covariates. All models included an interaction term between cohort and COVID-19 hospitalization to test whether changes in proportions of individuals with adverse credit outcomes between cohorts differed among patients who were hospitalized versus not hospitalized.

RESULTS: The study cohort included 132,109 patients with COVID-19 infections. We found significantly higher rates of adverse financial outcomes among individuals after COVID-19 infection compared with individuals prior to COVID-19 infection, and even greater increases among those hospitalized with COVID-19. Adjusted rates of adverse credit outcomes increased by 1, 1.7, and 3.5 percentage points for medical debt in collections, non-medical debt in collections, and low credit score for non-hospitalized COVID-19 patients, compared with absolute increases of 7.6, 6, and 9 percentage points for hospitalized COVID-19 patients, respectively (all p<0.001).

CONCLUSIONS: To our knowledge, our study is among the first to use credit report data to examine objective patient financial outcomes before and after COVID-19 hospitalization. We demonstrate that adverse financial outcomes were significantly more common among commercially-insured adults after compared to prior to COVID-19 infection, particularly among those hospitalized with COVID-19. Further research is essential to define the mechanisms of this association and design policies to improve financial outcomes for COVID-19 survivors.

AN EVALUATION OF RACE STRATIFICATION IN THE ATHEROSCLEROTIC CARDIOVASCULAR DISEASE POOLED COHORT RISK EQUATIONS USING A SEQUENTIAL MODELING STRATEGY

Arnab K. Ghosh¹; Sara Venkatraman¹; Monika M. Safford²; Parag Goyal³; Lisandro D. Colantonio⁴; Todd Brown⁴; Samprit Banerjee¹

¹Medicine, Weill Cornell Medical College, New York, NY; ²Department of Medicine, Weill Cornell Medical College, New York, NY; ³Medicine, Weill Cornell Medicine, New York, NY; ⁴Epidemiology, The University of Alabama at Birmingham, Birmingham, AL. (Control ID #3857055)

BACKGROUND: The 2014 atherosclerotic cardiovascular disease (ASCVD) Pooled Cohort risk equations (PCREs) employ race stratification. However, the use of race-specific risk prediction has been questioned due to increasing recognition of the specious relationship between race and the biology of disease, and that race differences in risk prediction may be a consequence of differences in social determinants of health (SDOH).

METHODS: We analyzed data from 13,198 REGARDS study participants aged 45-79 years, without ASCVD[SV1] [AG2] , and with low-density lipoprotein (LDL) cholesterol 70-189 mg/dL or non-high- density lipoprotein cholesterol (non-HDL-C) 100-219 mg/dL at baseline in 2003-2007. Participants were followed for up to 10 years for incident ASCVD, including myocardial infarction, coronary heart disease death, and fatal and nonfatal stroke. We analyzed the discrimination using C-statistics and calibration comparing observed to predicted events of the original PCREs based on baseline covariates. Then, we compared the discrimination and calibration of 3 versions of the PCRE: Model 1. Best-fit race stratified equations, Model 2. Best-fit equations without race stratification, and Model 3. Best-fit equations where race was replaced by indicators for SDOH.

RESULTS: Across all race-sex strata (Black females [BF], Black males [BM], White females [WF], and White males [WM]), C-statistics nor calibration did not improve after removal of race stratification in Model 2 (BF 0.7, 95% confidence interval [CI]: 0.67-0.74; BM 0.67, 95% CI: 0.63-0.71; WF 0.76, 95% CI: 0.72-0.79; WM 0.68, 95% CI: 0.65-0.7) nor after the addition of relevant SDOH variables in Model 3 (BF 0.72, 95% CI: 0.68-0.75; BM 0.68, 95% CI: 0.64-0.72; WF 0.76, 95% CI: 0.73-0.79; WM 0.68, 95% CI: 0.66-0.71) compared to Model 1 (BF 0.71, 95% CI: 0.68-0.74; BM 0.68, 95% CI: 0.64-0.72; WF 0.76, 95% CI: 0.73-0.79; WM 0.68, 95% CI: 0.65-0.71).

CONCLUSIONS: The removal of race-stratification from the PCREs is unlikely to impact performance, and thus the removal of race-stratified PCREs may be warranted

AN INNOVATIVE APPROACH TO COMBATING FOOD INSECURITY AT THE GRASSROOTS LEVEL

Shiva A. Nuti²; Vivek S. Patel²; Roshan B. Tom²; Jed Janecek²; Alicia Cano²; Jacob Alaniz²; Jenieve Chapa²; Nikhil Gogineni³; Sagar Kamprath¹

¹Department of Family Medicine, The University of Texas Medical Branch at Galveston, Galveston, TX; ²The University of Texas Medical Branch at Galveston School of Medicine, Galveston, TX; ³Texas Tech University Health Sciences Center School of Medicine, Lubbock, TX. (Control ID #3876689)

BACKGROUND: According to a nationwide database, our community has a food insecurity rate of 14.1%, higher than the national average of 11.8%. We hypothesize that though numerous nonprofits address food insecurity in our local community, our target population is largely unaware of them due to inadequate communication between the community and these existing nonprofits; and so naturally, one potential facet of the food insecurity issue in our city relates more to poor awareness rather than resource availability. To address this, we founded a student-led nonprofit where our focus is on community integration of existing food resources and direct delivery of food donations.

METHODS: Our organization is a nonprofit lead by 16 volunteers that connects underserved residents to food resources. We first identified 30 home-bound families who indicated that they were food insecure through established databases provided to us by free health clinics. We then collected goods in partnership with various nonprofits and major grocery chains. Furthermore, we compiled a comprehensive guide of available resources via online research to distribute during our food delivery drive. During our food drive, our goal was to connect these residents to existing resources and raise awareness about nearby resources to help bridge the gap between the underserved and those willing to help. Lastly, we distributed a survey to our residents, analyzing trends such as annual income, monthly rate of food scarcity, and their knowledge of organizations that can help.

RESULTS: We distributed a survey to our residents to determine the number of food-based organizations they were familiar with and then compare that number with the actual number of surrounding food-based organizations. We determined that in our city and its nearby areas, there are at least 11 food pantries, but our pilot survey (22/30 family responses) results show that a majority of our residents are scarcely aware of them - 16 families knew between 0-1 food organizations, and 6 families knew between 2-4 food organizations. This indicates that there is a gap between the available resources and knowledge of these resources among families.

CONCLUSIONS: Our organization behaves as a medium which facilitates communication and integration between the existing food-centric nonprofits and the population in need, thus bridging the knowledge gap regarding available resources.

To improve our organization, in the future, we would like to create a centralized database containing all of the available resources that encompass basic necessities ranging from food to medical care. Through this database, we can contact families about their needs and address them efficiently. Our goal of resource integration will be accomplished in partnerships with free clinics, churches, and volunteer physicians so that they can spread the word to the underserved. This model will hopefully reduce inefficiency in obtaining help in turn improving long-term health outcomes in our community.

A PILOT STUDY OF THE EFFECT OF PATIENT-PHYSICIAN LANGUAGE DISCORDANCE ON WEIGHT LOSS-RELATED OUTCOMES IN ADULT PATIENTS WITH OBESITY AND OVERWEIGHT ATTENDING A PHYSICIAN-SUPERVISED WEIGHT MANAGEMENT CLINIC IN NEW ENGLAND

Marina Spinelli¹; Varalakshmi Niranjan^2,1

¹University of Connecticut School of Medicine, University of Connecticut School of Medicine, Farmington, CT; ²Medicine, University of Connecticut, West Hartford , CT. (Control ID #3875158)

BACKGROUND: Patient-physician language concordance or discordance describes whether patient and physician speak the same or different languages. Several studies exploring patient-physician language concordance/discordance find worse health outcomes with discordance; others find no differences. To our knowledge, this study is the first investigating the relationship between patient-physician language concordance/discordance and weight loss. With obesity and overweight being prevalent and having serious comorbidities, we aim to understand impacts of patient-physician language concordance/discordance on weight loss and related outcomes.

METHODS: We randomly selected 80 language-concordant and 20 language-discordant patients (using documented “preferred language”) from the adults with obesity and overweight attending a physician- supervised weight management clinic in New England from January 2017 through August 2022. From chart review, we recorded weights and related variables, calculated changes from first to last visits, calculated mean changes for both groups, and compared group means by two-tailed independent Welch t-tests with alpha of 0.05.

RESULTS: Preferred languages were English (N=80), Spanish (N=17), Arabic (N=1), Punjabi (N=1), and American Sign Language (N=1). Analyses showed the following means (p-values) for the language- concordant versus discordant groups, respectively: number of appointments attended was 3.6 versus 2.8 (0.30); weight decreased by 2.5% (of initial) versus 1.5% (0.34); skeletal muscle mass decreased by 1.4% (of initial) versus 0.06% (0.06); body fat decreased by 1.5% versus 1.8% (0.76); systolic blood pressure increased by 0.2 mmHg versus 1.8 mmHg (0.47); diastolic blood pressure decreased by 1.6 mmHg versus 0.03 mmHg (0.47); hemoglobin A1c decreased by 0.13 versus 0.06 (0.81); total cholesterol decreased by 2 versus increased by 7 (0.20); LDL cholesterol decreased by 5 versus increased by 5 (0.09).

CONCLUSIONS: There were no statistically significant differences in weight loss or related outcomes between language-concordant and discordant groups. Lack of difference may reflect limited sample size and/or the clinic’s efforts (professional medical interpreters via telephone language lines, “after visit summaries” in preferred language, and culturally sensitive weight loss meal plans) to decrease potential impacts of patient-physician language discordance and potentially coexisting cultural differences. If further studies support that current practices effectively address potential impacts of patient-physician language discordance, the clinic should strive to continue these practices.

APPLYING CRITICAL RACE FEMINISM TO EXPLORE BARRIERS AND FACILITATORS TO POSTPARTUM PRIMARY CARE UTILIZATION AMONG BLACK WOMEN WITH CARDIOVASCULAR DISEASE RISK FACTORS: A QUALITATIVE STUDY

Serena M. Ogunwole¹; Habibat A. Oguntade²; Kelly M. Bower³; Lisa A. Cooper⁴; Wendy Bennett⁵.

¹General Internal Medicine, Johns Hopkins University, Baltimore, MD; ²University of Minnesota School of Public Health, Minneapolis, MN; ³Johns Hopkins University School of Nursing, Baltimore, MD; ⁴Medicine, Johns Hopkins University School of Medicine, Baltimore, MD; ⁵Medicine, Johns Hopkins University, Baltimore, MD. (Control ID #3874253)

BACKGROUND: Black women with pregnancy complications, including excess gestational weight gain, gestational diabetes and hypertension, have a high risk for future chronic disease. Effective transitioning to primary care in the postpartum period may reduce chronic disease risk factors and thus improve racial disparities in maternal health and long-term health. Utilization of postpartum primary care is lower for Black (vs. White) women, but few studies have explored the unique experiences of Black postpartum women, while considering the intersecting influences of racism, sexism, and classism on their lived experiences. Our objectives are to: 1) identify barriers and facilitators to postpartum primary care utilization among Black women with cardiometabolic risk factor and 2) explore the cultural, social, and historical context that may influence postpartum primary care utilization among Black postpartum women with cardiometabolic risk factors.

METHODS: We recruited 18 Black women within one year of delivery with cardiometabolic complications prior to or during pregnancy (pre-pregnancy chronic hypertension, type 2 diabetes, obesity, preeclampsia, or

gestational diabetes) from early home visiting programs in Baltimore, Maryland. We conducted in-depth interviews from May 2020-June 2021. Two researchers trained in qualitative analysis conducted and analyzed the data. We created predetermined codes guided by critical race feminism theory and the research objectives. We used a hybrid inductive and deductive thematic analysis to identify codes and themes.

RESULTS: We identified eight main themes: 1) limited knowledge on the concept and need for postpartum primary care 2) gaps in care transitions and communication between the obstetric and primary care providers 3) pregnancy and childbirth as critical life experiences that influence postpartum health and future well-being 4) unique mental health and wellness needs that emerge postpartum 5) patient-provider relationship affecting care experiences 6) structural and personally mediated racism not explicitly articulated, but present and ubiquitous influences 7) intersectional experiences of Black motherhood/womanhood as sources of unique stress and resilience and 8) sociocultural ideals about health and healthcare influence care-seeking.

CONCLUSIONS: Black women’s experiences before, during, and after pregnancy provide valuable insights into facilitators and barriers to utilizing postpartum primary care. Interventions at the patient, provider, and system levels that are attentive to intersectional identities and lived expereinces, may promote health equity for Black women’s postpartum primary care and ultimately improve maternal and cardiovascular health.

A PRAGMATIC, STAKEHOLDER-INFORMED APPROACH TO PRIORITIZING SOCIAL RISKS FOR HEALTH CARE SYSTEM-BASED SCREENING AND INTERVENTION

Mayuree Rao^1,2; Matthew Maciejewski^3,4; Cindie Slightam⁵; Camila Chaudhary⁵; Karin Nelson^1,2; Donna Zulman^5,6

¹Health services research & development, VA Puget Sound Health Care System, Seattle, WA; ²Medicine, University of Washington, Seattle, WA; ³HSR&D, Durham VA Medical Center, Durham, NC; ⁴Duke University, Durham, NC; ⁵Health Services Research & Development, VA Palo Alto Health Care System, Palo Alto, CA; ⁶Medicine, Stanford University, Stanford, CA. (Control ID #3877272)

BACKGROUND: Screening for social risks – individual-level social and economic conditions – is increasingly common in clinical care. However, there exists little formal guidance on how to select among social risks for systematic screening within a health system.

METHODS: We developed and piloted a pragmatic approach for prioritizing social risks for screening

within a random sample of Veterans receiving care in the Veterans Affairs (VA) health care system. First, we identified social risks from existing screening tools and VA resources. Social risks were excluded if they did not meet our working definition of a social risk (from the US Preventive Services Task Force) or were not relevant to Veterans, based on duplicate review. Next, 15 VA stakeholders including researchers, operations leaders, and Veterans completed an online survey, rating the extent to which they perceived each social risk

1) impacts health outcomes, 2) impacts patient experience and 3) is actionable. Stakeholders were also asked to identify the 12 highest priority social risks. Survey results were presented to stakeholders during 2 hour- long meetings, serving as a basis for discussion of social risk prioritization. Finally, we used survey results and qualitative comments to identify high-priority social risks within and across criteria.

RESULTS: We identified 37 social risks, narrowed to 31 after exclusion based on definition and relevance to Veterans. Among the 31 social risks presented to stakeholders, discrimination experiences, health care access, housing instability, race/ethnicity, and transportation barriers were rated highly (i.e., were among the 12 highest rated social risks) across all 4 questions on the online survey. During discussion, stakeholders described challenges they faced in prioritizing social risks given varied clinical contexts; actionability was notably difficult to evaluate. Synthesizing the information from survey and discussion, we selected 21 high- priority social risks (Table).

CONCLUSIONS: Using a novel, pragmatic approach combining quantitative and qualitative data, we engaged diverse stakeholders to prioritize 21 social risks for health system-based screening and intervention.

This process highlights key criteria by which health care organizations may evaluate candidate social risks for systematic screening.

AREA DEPRIVATION AND WALKABILITY INFLUENCE CARDIOVASCULAR OUTCOMES AMONG INDIVIDUALS WITH HEART FAILURE WITH REDUCED EJECTION FRACTION Timothy M. Bober¹; Emily Guhl⁶; Scott D. Rothenberger²; Kwonho Jeong³; Kaleab Z. Abebe²; Amy Anderson²; Jared W. Magnani⁴; Bea H. Belnap²; Amber E. Johnson⁴; Bruce L. Rollman⁵

¹General Internal Medicine, UPMC Presbyterian-Shadyside, Pittsburgh, PA; ²Medicine, University of Pittsburgh, Pittsburgh, PA; ³Biostatistics, University of Pittsburgh, Pittsburgh, PA; ⁴Medicine, University of Pittsburgh, Pittsburgh, PA; ⁵Medicine, University of Pittsburgh, Pittsburgh, PA; ⁶Cardiology, Heritage Valley Health System, Sewickley, PA. (Control ID #3868804)

BACKGROUND: Neighborhood-level measures like Area Deprivation Index (ADI) and walkability are associated with increased cardiovascular disease (CVD) morbidity, higher utilization of health services, and mortality. We sought to understand the impact of these neighborhood-level factors on hospital readmissions and mortality among recently hospitalized depressed and non-depressed heart failure patients with reduced ejection fraction (HFrEF) enrolled in the NHLBI-funded Hopeful Heart Trial of telephonic collaborative care for HFrEF and depression in primary care.

METHODS: From 3/14-10/17, Hopeful Heart enrolled 756 hospitalized HFrEF patients (EF ≤45%) and NYHA class II-IV symptoms including 629 depressed (screen-positive 2-item Patient Health Questionnaire (PHQ-2) prior to discharge and score ≥10 on the PHQ-9 administered via telephone two weeks post- discharge) and 127 non-depressed HFrEF patients (PHQ-2 negative and PHQ-9 <5). We determined participants’ ADI score (0-100 scale, higher scores are worse) and Walk Score (0-100 scale, higher scores are better) using home addresses at the time of hospital discharge. We categorized participants into the top quartile (Q4) for ADI (most deprived) and Walk Score (most walkable) and then compared Q4 with all other quartiles (Q1-Q3). Methods included unadjusted nonparametric Kaplan-Meier analyses for survival curves, log-rank tests for differences in survival curves for 12-month mortality, and unadjusted Chi-Squared tests for 12-month readmission rate. Baseline depression status and treatment assignment did not predict readmission in our primary analysis and was not included in this subsequent analysis.

RESULTS: In all, 756 individuals with HFrEF participated (44% female; 73% white; mean EF: 28.3 +/- 9.1%) with mean ADI of 66.7 +/- 23.5 and mean Walk Score of 33.4 +/- 25.9. The mean PHQ-9 score for both ADI and Walk Score Q4 groups was similar to Q1-Q3. Participants in the Q4 group for worst ADI had a higher 12-month incidence of readmission that those in Q1-Q3 (81% vs. 70%, p=0.004), while those in Q4 for best Walk Score were readmitted at a similar level to QI-Q3. People in Q4 for ADI and Walk Score also had a similar 12-month incidence rates of cardiovascular (CV) (14% for Q4) and all-cause mortality (23% for Q4) versus Q1-Q3 quartiles.

CONCLUSIONS: Among a predominantly depressed cohort of people with HFrEF who enrolled in a clinical trial, worse ADI but not Walk Score is predictive of higher risk of 12-month hospital readmission. Neither ADI or Walk Score is predictive of 12-month CV or all-cause mortality. Further studies are needed to elaborate how neighborhood factors affect outcomes for people with heart failure such that interventions can be designed to address these disparities.

ASSESSING BARRIERS TO PRIMARY CARE FACED BY RESETTLED BHUTANESE REFUGEES THROUGH A QUALITATIVE ASSESSMENT

Amy Zheng¹; Swetara Joshi¹; Bhuwan Gautam²; Aditi Sharma²; Heather Costigan³; Morgan Loeffler³; Heather Stuckey³; Megan Mendez-Miller⁴; Tanuja Devaraj⁵

¹Penn State College of Medicine, Hershey, PA;²Department of Public Health Sciences, Penn State College of Medicine, Hershey, PA; ³Qualitative and Mixed Methods Core, Penn State College of Medicine, Hershey, PA; ⁴Department of Family and Community Medicine, Penn State College of Medicine, Hershey, PA; ⁵Whitman Walker Health, Washington, DC. (Control ID #3872707)

BACKGROUND: The Nepali-speaking Bhutanese (NSB) community is one of the largest resettled refugee communities in the U.S. Resettled from 2007 to 2014, this population of 90,000 individuals has been reported to experience a high burden of chronic illness by several early studies. After almost a decade from their initial resettlement in the U.S, the NSB community continues to face disparities in health and challenges to receiving quality primary care. This study utilizes a qualitative approach to investigate the current barriers faced by the NSB community in receiving primary care.

METHODS: Participants who identified as NSB refugees were recruited from an Internal Medicine clinic at a tertiary medical center in Pennsylvania. Semi-structured interviews were conducted from August 2021 to March 2022. Questions reviewed demographics, socio-economic barriers, clinic visit experience, physician- patient relationship, care navigation and health beliefs. Interviews were conducted in Nepali, transcribed and coded via NVivo. Coded data were analyzed to identify significant emerging themes.

RESULTS: Eleven men and thirteen women participated in the study (N=24), with an average age of 50.3±17.8 years. The average number of years that participants spent in refugee camps was 18.5 years, and almost two-thirds (62.5%) of participants reported not completing high school. Major barriers identified included lack of communication in between appointments and miscommunication (e.g., abbreviated paraphrasing or dialectical differences) related to language interpretation. Many patients also described frustration with accessing appointments with their primary care physician, difficulty coordinating long wait times with other personal obligations, and lack of continuity of care with their providers. Positive facilitators to care involved helpful interpersonal relationships with staff and providers as well as strong family support (e.g., transportation, interpretation, scheduling appointments, and picking up medicine).

CONCLUSIONS: The resettled NSB refugee population faces unique experiences and barriers in primary care. Key considerations to provide effective primary care include language and culturally appropriate communication, continuity of care with primary care providers, and care navigation.

ASSESSMENT OF DEPRESSION AND ADHERENCE TO ORAL ANTICOAGULATION IN INDIVIDUALS WITH ATRIAL FIBRILLATION

Matthew Lapa³; Gretchen Swabe³; Jared W. Magnani^3,2; Matthew Muldoon³; Bruce L. Rollman¹

¹Medicine, University of Pittsburgh School of Medicine, Pittsburgh, PA; ²Center for Research on Health Care, Department of Medicine, University of Pittsburgh, Pittsburgh, PA; ³University of Pittsburgh School of Medicine, Pittsburgh, PA. (Control ID #3876000)

BACKGROUND: Depression is associated with decreased adherence to essential medication and poor clinical outcomes. However, there is a lack of literature investigating the association of depression with adherence to oral anticoagulants commonly prescribed to patients with atrial fibrillation (AF). We hypothesized that individuals with depression would have decreased adherence to direct-acting oral anticoagulants (DOACs) compared to those without.

METHODS: We analyzed claims from Optum’s de-identified Clinformatics® Data Mart Database to identify individuals with incident AF initiating either DOACs or warfarin between 2013-2019. We quantified adherence using proportion of days covered (PDC), categorized as limited (PDC <80%), adequate (PDC 80 to <90%), or optimal (PDC ≥90%). We used logistic regression to relate depression to 12-month adherence to anticoagulation in models adjusting for demographics, medical and psychiatric comorbidities, household income, educational attainment, and insurance type. We conducted a secondary analysis ascertaining adherence stratified by DOACs or warfarin between individuals with and without depression.

RESULTS: We identified 101,041 individuals (age 74.5±8.9, 49.3% female, 28.9% non-White) with a new diagnosis of AF who initiated anticoagulation. In those with depression (n=16,108) the odds of adequate adherence to DOACs was 0.89 (95% CI: 0.85-0.93) and optimal adherence was 0.88 (95% CI: 0.84-0.92) compared to individuals without depression. There was no difference in adherence to warfarin

CONCLUSIONS: Comorbid depression in individuals with AF is associated with around 11% to 12% lower adequate and optimal 12-month adherence respectively to DOACs. Recognizing depression as concomitant with AF may guide interventions to improve adherence to essential oral anticoagulants.

ASSOCIATION BETWEEN AGE AND HEALTHCARE UTILIZATION IN ADULTS WITH CRIMINAL LEGAL INVOLVEMENT (CLI)

Sanjay Bhandari¹; Laura Hawks²; Rebekah J. Walker³; Leonard E. Egede⁴

¹Internal Medicine, Medical College of Wisconsin, Milwaukee, WI; ²General Internal Medicine, Medical College of Wisconsin, Wauwautosa, WI; ³Medicine, Medical College of Wisconsin, Milwaukee, WI; ⁴Medicine, Medical College of Wisconsin, Milwaukee, WI. (Control ID #3872186)

BACKGROUND: An exposure to the criminal legal system is associated with worse chronic disease control and increased healthcare utilization. While in the general population, older adults use acute and preventive services at higher rates than younger patients, little is known about the association between age and healthcare utilization in adults with CLI.

METHODS: Our sample included 37,279 US adults who participated in the National Survey of Drug Use and Health (2015–2019) and reported lifetime history of arrest or time spent on probation or parole. Our independent variable was age category: 18-49 years (younger adults); 50-64 years (middle aged adults); and 65+ years (older adults). Using negative binomial regression, we tested the association between these three age categories and healthcare utilization (ED, inpatient, and outpatient) controlling for relevant sociodemographic and clinical covariates.

RESULTS: Of the total participants, 59.9% were younger adults, 28.53% middle aged and 11.48% were older adults. Compared to the younger group, older adults were more likely to be male and identify as White race. Older adults had higher rates of most chronic noncommunicable diseases, but lower rates of past year substance use disorder or mental health condition. There were no difference in unadjusted mean number of ED visits between the three groups; older adults had more inpatient nights (1.2, 95% CI 0.9-1.4) than younger adults (0.5; 95% CI 0.4-0.5) and middle aged adults (0.8, 95% CI 0.7-0.9); there was a similar trend for outpatient visits. After adjustment, the older adults had lower frequency of ED visits (IRR 0.7, 95% CI 0.6-0.8) as compared to younger adults. The increased rate of inpatient utilization for older respondents compared to younger adults was completely attenuated in adjusted analyses (IRR 1.1; 95% CI 0.7-1.7). Middle aged respondents had higher outpatient utilization (IRR 1.1, 95% CI 1.1-1.2), whereas older adults had similar utilization pattern as compared to younger adults (IRR 1.0, 95% 0.9-1.1). Other independent predictors for higher acute healthcare utilization (ED visits and/or inpatient nights) included female gender, Black race, being widowed/divorced or never married, living in small- or non-metropolitan areas, insurance coverage by Medicaid or Medicare, and presence of various comorbidities.

CONCLUSIONS: Healthcare utilization patterns for adults with lifetime CLI are distinct, with younger and middle adults adults using more overall healthcare. After adjustment, younger adults are more likely to visit the ED than the older two groups, all three are equally likely to be inpatient and middle aged adults have highest rates of outpatient utilization. Interventions to reduce acute healthcare utilization in individuals with lifetime CLI should consider addressing the physical and behavioral health needs of young and middle aged adults with low education, low social support, residing in rural areas, and include multidisciplinary approach to manage their comorbidities.

ASSOCIATION BETWEEN FOOD INSECURITY IN YOUNG ADULTHOOD AND INCIDENT CARDIOVASCULAR DISEASE IN CARDIA: A LONGITUDINAL STUDY

Jenny Jia¹; Mercedes R. Carnethon²; Mandy Wong²; Cora E. Lewis³; Pamela Schreiner⁴; Namratha R. Kandula¹

¹Medicine, Northwestern University Feinberg School of Medicine, Chicago, IL; ²Preventive Medicine, Northwestern University, Chicago, IL; ³Medicine, University of Alabama at Birmingham, Birmingham, AL; ⁴Epidemiology and Community Health, Regents of the University of Minnesota, Minneapolis, MN. (Control ID #3876822)

BACKGROUND: Food insecurity, a condition of inconsistent access to sufficient food for an active, healthy lifestyle, affects 1 in 10 Americans annually. Food insecurity is cross-sectionally associated with cardiovascular disease (CVD) and its risk factors, such as hypertension, obesity, and lifestyle behaviors. However, few studies have examined these relationships longitudinally.

METHODS: We included participants from the Coronary Artery Risk Development in Young Adults (CARDIA) study with food insecurity and other exposure data and no CVD event as of 2000-2001, baseline for this analysis. The primary exposure was food insecurity at year 15. Participants who endorsed “We have enough food to eat, but not always the kinds of food we want to eat”; “Sometimes we don’t have enough food to eat”; or “Often we don’t have enough food to eat” in the past year were categorized as having some food insecurity. The outcome was CVD events, a composite of coronary heart disease, stroke/TIA, heart failure, peripheral arterial disease and CVD death through August 31, 2020. Follow-up time was calculated as timefrom baseline to time at which CVD event was identified or censored at 2020. We used multivariable Cox proportional hazards models to estimate hazard ratios (HRs) and 95% CIs for incident CVD as a function of food insecurity, adjusted for age, sex, socioeconomic factors (education, marital status, income, and usual source of medical care), and race. We also assessed sex, education, and race as effect modifiers by including multiplicative interaction terms between the primary exposure and candidate modifiers in separate models.

RESULTS: In our sample of 3,626 adults at baseline, mean age was 40.2 years (SD 3.6), and 531 (14.6%) had some food insecurity. In those with some food insecurity, 360 (68%) identified as Black and 151 (31%) had household incomes <$16,000 per year, compared to 1341 (43%) and 184 (10%) in those with no food insecurity, respectively. Over 20 years of follow up, 272 CVD events occurred, with 208 (6.7%) in those with no food insecurity and 64 (12.1%) in those with some food insecurity. After adjusting for age, sex, and field center, food insecurity was associated with a greater risk of incident CVD (HR 1.46; 95% CI 1.09, 1.94). In subsequent models, this association was attenuated after adjusting for socioeconomic factors (HR 1.14; 95% CI 0.83, 1.58) but not appreciably different after further including race (HR 1.09; 95% CI 0.72, 1.65). We found no interactions between food insecurity and sex, education, or race in the relationship between food insecurity and incident CVD.

CONCLUSIONS: In CARDIA, food insecurity in young adulthood was not associated with CVD events in middle adulthood, after accounting for other socioeconomic factors. This relationship did not vary by sex, education, or race. Food insecurity appears to be a marker of socioeconomic status and is a useful single measure in screening for social determinants of health in clinical settings.

ASSOCIATION BETWEEN GOVERNMENTAL SPENDING ON SOCIAL SERVICES AND HEALTH CARE USE AMONG LOW-INCOME MEDICARE BENEFICIARIES

Carlos Irwin Oronce^3,1; Ninez A. Ponce²; Catherine Sarkisian^3,1; Yusuke Tsugawa^1,2

¹General Internal Medicine & Health Services Research, David Geffen School of Medicine at UCLA, Los Angeles, CA; ²Health Policy and Management, UCLA School of Public Health, Los Angeles, CA; ³Medicine, VA Greater Los Angeles Healthcare System, Los Angeles, CA. (Control ID #3870245)

BACKGROUND: Low-income older adults experience a high burden of adverse social risk factors that contribute to worse access to primary care, frequent hospitalizations, and higher mortality. Governmental programs and social services mitigating these risks may lead to better outcomes. This study aims to evaluate the effect of county-level social spending on health care outcomes of dual-eligible Medicare beneficiaries.

METHODS: We analyzed a 20% random sample of Medicare beneficiaries 65 years and older dually enrolled in Medicaid from 2016 through 2018 linked to county-level governmental expenditures data from the US Government Finance Database. Four types of social spending comprised the exposures of interest: per capita spending on (1) public welfare, (2) housing and community development, (3) public transit, and (4) infrastructure. Outcomes included annual rates of primary care visits, emergency department (ED) visits, and preventable hospitalizations. We used separate multivariable negative binomial regression models with standard errors clustered at the county to measure the association between spending type and beneficiary- level outcomes, adjusting for individual and county covariates.

RESULTS: The analytic sample consisted of 607,857 low-income Medicare beneficiaries residing in 3,121 counties (mean age, 78.0 [SD 9.0] years; 66.1% female). Median county social spending in the four categories was $533 per capita (range: $312 to $1,016). On average, dual-enrollees had 3.5 (SD 3.8) primary care visits, 2.0 (SD 3.5) ED visits, and 0.2 (SD 0.5) preventable hospitalizations per year. After adjustment, higher welfare spending was associated with 12% higher primary care visits (adjusted incidence rate ratio [IRR] comparing the highest vs. lowest quintiles, 95% CI 1.04–1.20), 6% lower ED visits (IRR 95% CI 0.89–0.99), 8% lower preventable hospitalizations (IRR 95% CI 0.87–0.97) and 28% lower preventable hospitalizations from acute conditions (IRR 95% CI 0.75–0.90) per year. We observed that higher housing and community development spending was associated with 15% higher primary care visits (IRR 95% CI 1.07–1.23), 8% lower preventable hospitalizations (IRR 95% CI 0.87–0.98), and 14% lower preventable hospitalizations from acute conditions (IRR 95% CI 0.79–0.93) per year. Higher public transit spending was associated with 8% lower preventable hospitalizations (IRR 95% CI 0.88–0.96) and 14% lower preventable hospitalizations from acute conditions (IRR 95% CI 0.81–0.91) per year. Living in the highest quintile of infrastructure spending was not associated with any outcomes.

CONCLUSIONS: In a national sample of low-income Medicare beneficiaries, greater governmental spending on welfare, housing and community development, and public transit was associated with increased rates of primary care use and lower rates of acute care utilization. Public investments that address health- related social needs of low-income older adults may facilitate access to outpatient services while reducing costly acute care.

ASSOCIATION BETWEEN SOCIAL RISK PROFILES AND CLINICAL OUTCOMES IN ADULTS WITH TYPE 2 DIABETES: A LATENT PROFILE ANALYSIS

Rebekah J. Walker¹; Joni S. Williams²; Sebastian Linde³; Leonard E. Egede⁴

¹Medicine, Medical College of Wisconsin, Milwaukee, WI; ²Medicine, Medical College of Wisconsin, Milwaukee, WI; ³Medicine, Medical College of Wisconsin Department of Medicine, Milwaukee, WI; ⁴Medicine, Medical College of Wisconsin, Milwaukee, WI. (Control ID #3876193)

BACKGROUND: Recent evidence suggests combining medical care with interventions that target social risk factors can improve health outcomes for adults with diabetes. However, no clear process exists for identifying how best to categorize social risks in a way that informs effective intervention development. In addition, it is unclear if combinations of different risks are important to target concurrently as opposed to addressing risks separately.

METHODS: The aim of this analysis was to investigate social risk profiles and associations of those profiles with clinical outcomes in adults with diabetes using latent profile analysis. 615 adults with diabetes completed validated questionnaires for 26 social risk factors within six domains of social determinants of health: socioeconomic, neighborhood and built environment, education access and quality, food environment, social and community context, psychological risk, and behavioral risk. A three-step latent profile analysis was used to identify different subgroups within the sample using Stata v16. Continuous variables were created for each social risk indicator. First, 2, 3, 4, 5, 6, and 7-class profiles were investigated with a 5-class profile identified as the best fit profile based on AIC and BIC scores. Second, posterior probability was used to classify individuals into their true latent class, with accuracy of classification determined by entropy score (0.94). Finally, based on marginal probabilities, we tested the profiles against the outcomes of HbA1c, blood pressure, and quality of life.

RESULTS: Five latent class profiles were identified. The lowest risk group accounted for 36% of the sample and had limited social risk but indicated concerns about their neighborhood. This group had significantly higher mental health related quality of life compared to the other four groups. The second group accounted for 34% of the sample and had overall low risk, but considered their neighborhoods to have low aesthetics, low food access, and high social isolation. This group had significantly higher blood pressure. Group 3 accounted for 6% of the sample and had high social risk in terms of neighborhood crime, violence, and financial instability but low psychological and behavioral risk. The fourth group accounted for 14% of the sample and had high psychological and behavioral risks but low socioeconomic and neighborhood risks. This group had significantly higher HbA1c (0.5%) and lower mental health related quality of life. The highest risk group accounted for 8% of the sample and indicated high risk in all domains. This group had significantly higher HbA1c (1%) and lower mental health related quality of life.

CONCLUSIONS: Social risk profiles can be identified that can guide development and implementation of interventions based on the social determinant domain of highest risk and the health outcome of concern in adults with diabetes.

ASSOCIATION OF FOOD INSECURITY AND CARDIOMETABOLIC DISEASE IN PATIENTS OF FEDERALLY-QUALIFIED HEALTH CENTERS: A CROSS-SECTIONAL STUDY

Jenny Jia¹; Andrew L. Owen²; Matthew J. O'Brien²

¹Medicine, Northwestern University Feinberg School of Medicine, Chicago, IL; ²Center for Community Health, Northwestern University Feinberg School of Medicine, Chicago, IL. (Control ID #3876746)

BACKGROUND: Federally qualified health centers (FQHCs) are important sources of accessible medical care to low-income families in underserved areas. Many low-income families are affected by food insecurity, which is associated with chronic cardiometabolic diseases requiring primary care. Few studies have explored whether the association between food insecurity and cardiometabolic diseases exists in FQHC patients and whether it varies by sex.

METHODS: We included adult patients from community health centers in the Alliance of Chicago Community Health Services national network who were screened for food insecurity from 2017 to 2021. We measured the primary exposure, food insecurity, using the 2-question Hunger Vital Signs, categorizing responses of “yes,” “sometimes true,” and “often true” to either question at any time point as food insecure and “no,” “never true,” and “does not apply” at all time points as food secure. Outcomes were prevalent diabetes/prediabetes and cardiovascular disease (CVD), which we measured using diagnostic codes from electronic health records. To study the association between food insecurity and clinical outcomes, we used multiple logistic regression with food insecurity and sociodemographic covariates on the presence of each clinical outcome. In interaction models, we included multiplicative interaction terms between food insecurity and sex for each clinical outcome. We set statistical significance at α=0.05.

RESULTS: Our sample included 221,284 adult patients, 66% female, with mean age 36 years (SD 17.6). Food insecure patients, who accounted for 9% (n=20,460) of the sample, were more likely than food secure patients to identify as Black (49% vs. 41%) and less likely to have private health insurance (8% vs. 14%). In age- and sex-adjusted models, food insecurity was associated with diabetes/prediabetes (aOR= 1.09, p<0.001) and CVD (aOR=1.10, p<0.01). After further adjusting for health insurance, language, and race, which acted as a proxy for social determinants of health, food insecurity was associated with diabetes/prediabetes (aOR=1.07, p<0.01) but not with CVD (aOR=1.04, p>0.05). The interaction between food insecurity and sex on diabetes/prediabetes was significant, and food insecurity had a greater association with diabetes/prediabetes in women (aOR=1.17, p<0.001) than men. No interaction was found with CVD.

CONCLUSIONS: In a national sample of FQHC patients, food insecurity was associated with diabetes/prediabetes; this association was stronger in women than men. The association between food insecurity and CVD was null after accounting for proxies of social determinants of health. The magnitudes of these associations were lower than in prior similar studies in general populations, suggesting that further research should explore the interaction of social determinants of health and access to and use of FQHCs.

ASSOCIATION OF HIGH SENSITIVITY C-REACTIVE PROTEIN LEVELS WITH ELECTRONIC CIGARETTE USE COMPARED TO TRADITIONAL CIGARETTE, DUAL USERS, AND NON- SMOKERS AMONG US ADULTS.

Judith Gutierrez, Jacqueline Hirth, Roger Zoorob

Family and community medicine, Baylor College of Medicine, Houston, TX. (Control ID #3860528)

BACKGROUND: The use of electronic cigarettes (EC) has exponentially increased since their introduction to the US market in 2007, especially among young adults. Smoking is known to increase the risk of cardiovascular disease, but information on the impact of EC use on the cardiovascular (CV) system remains scarce. The purpose of this study is to evaluate differences in high sensitivity C-reactive protein (hs-CRP) levels, as a surrogate measure of cardiovascular risk, among EC users, traditional tobacco smokers (TTS), dual EC, and TTS as compared to non-smoking adults.

METHODS: This is a population-based secondary analysis of cross-sectional data from the country-wide representative, National Health and Nutrition Examination Survey (NHANES), which included data collected between 2016 and 2018. Each cycle, conducted across 2-year intervals, consisted of demographic, dietary,

and health-related data as well as a medical examination component in which physiological measurements are taken. We included all participants 21 to 60 years old in this study. Self-reported smoking and EC use were considered. Smoking status and nicotine-containing EC use were confirmed through serum cotinine levels. Serum hs-CRP was measured among patients who agreed to medical exam and blood draw. In addition to evaluating hs-CRP levels as continuous outcomes, we also created cutoff points that indicate low and moderate/high levels. Analyses were conducted using weights, clusters, and strata according to NHANES instructions to provide nationally representative estimates. Poisson regression was used to test the association between smoking and/or EC use and hs-CRP. Models were adjusted for covariates (age, gender, race/ethnicity, BMI, Diabetes, and hypertension).

RESULTS: A total of 6,219 participants were included. Low-level hs-CRP, indicating low CV risk, was present in 1,959 (32%) participants, and 4,260 (68%) had moderate/high CV risk. For nicotine product use, we found 508 (8%) smoked, 677 (11%) used ECs only, 823 (13%) both smoked and used ECs, and 4,211 (68%) were non-smokers. Smokers had the highest CV risk as indicated by hs-CRP [IRR=2.17 (95% CI= 2.17-2.18), p<0.001], and by dual smokers and EC users [IRR=2.17 (95% CI= 2.16-2.17), p<0.001], compared to non-smokers. E-cigarette smokers had a lower risk than both traditional and dual smokers, however, the risk was higher than non-smokers [IRR=2.08 (95% CI= 2.07-2.08), p<0.001].

CONCLUSIONS: Traditional tobacco and dual smoking are associated with increased CV risk, as indicated by hs-CRP. E-cigarettes are associated with lower CV risk than traditional smokers and dual smokers, but the risk is still higher than non-smokers.

ASSOCIATIONS BETWEEN SOCIAL DETERMINANTS OF HEALTH AND TOTAL HEALTHCARE COST IN AN INTEGRATED DELIVERY NETWORK

Ariel R. Silverman, Mark Duggan, Paul J. Chung, Robert Nocon

Health Systems Science, KaiserPermanente Bernard J Tyson School of Medicine, Pasadena, CA. (Control ID #3869414)

BACKGROUND: The U.S. is increasingly recognizing impacts of social risks on health and other outcomes. Meanwhile, healthcare costs are astronomically high, yet have not shown improved outcomes compared to peer nations with lower healthcare spending. Although social risk is widely believed to contribute to healthcare cost, this association has not yet been firmly established in a general population, as most research has focused on small, disease-specific populations or patients with high cost and high medical need. Additionally, research in this area has been challenging as there has not yet been standardized widespread adoption of social determinants of health (SDoH) screening. This study aims to identify associations between patient-reported social risk and total healthcare cost in an organization-wide sample of Kaiser Permanente (KP) members.

METHODS: A survey conducted by KP Social Needs Network for Evaluation and Translation (SONNET) was administered in 2020. Patients from all 9 states were surveyed using stratified sampling with oversampling for patients residing in census blocks with median household incomes at or below the 25^th percentile, Medicaid members, and applicants for financial assistance. The survey asked about a subset of SDoH including financial, housing, food, and transportation needs from the prior year. Survey data were linked to administrative electronic health record data including demographics, medical risk, and total direct healthcare cost for all services paid for by the health plan from 2019. Multivariable negative binomial regressions were performed to determine associations between social risk and total cost controlling for medical risk using DxCG, a widely used risk adjustment tool, and demographic variables.

RESULTS: Survey responses from 10,226 members were collected (23% response rate). Respondents were categorized into groups of no, moderate, or severe social risk based on existing literature, local operational team subject matter experts, and response to Likert-style questions. In a model without medical risk, moderate and severe risk were associated with healthcare cost 13% (p=0.02) and 48% (p<0.001) higher, respectively, than no risk, controlling for demographic covariates such as age group, gender, race, and region. When medical risk was added to the model, a 10% increase in medical risk was associated with 2.8% higher total cost (p<0.001), but the association between social risk and total cost was attenuated.

CONCLUSIONS: These results demonstrate a positive relationship between social risk and total healthcare cost, while confirming that medical risk is a potentially powerful mediator of this relationship. Further research is needed to better understand how medical risk might mediate the relationship, as well as potential impacts of social risk interventions. As payers, insurers, and providers build strategies to invest in SDoH at a broader scale, it is essential that we better understand relationships between social risk and healthcare cost.

AUTOMATED IDENTIFICATION AND DESCRIPTIVE ANALYSIS OF PATIENTS’ UNMET SOCIAL NEEDS IN CLINICAL TEXT: A CASE REPORT

Jane Njeru¹; Sungrim Moon²; Jay B. Doughty⁴; Mark L. Wieland¹; Jungwei W. Fan^2,3

¹Division of Community Internal Medicine, Geriatrics and Palliative Care, Mayo Clinic Minnesota, Rochester, MN; ²Department of Artificial Intelligence & Informatics, Mayo Clinic Minnesota, Rochester, MN; ³Division of Epidemiology, Department of Quantitative Health Sciences, Mayo Clinic Minnesota, Rochester, MN; ⁴Health Care Administration, Mayo Clinic Minnesota, Rochester, MN. (Control ID #3874360)

BACKGROUND: Social determinants of health (SDoH) influence health-related social needs, and are associated with suboptimal access to timely care, poor control of chronic diseases, and gaps in preventive care. Identifying patients with concomitant medical and social complexity is pivotal to funneling finite resources to address the needs of patients at the highest risk. However, systematically identifying patients who need assistance is difficult, and while data on medical comorbidities and prior healthcare utilization is easily accessible through electronic health records (EHR) and administrative databases, data on social complexity and unmet social needs are often buried in unstructured clinical text. The goal of this project was to develop natural language processing (NLP) solutions for identifying patients’ unmet social needs to facilitate referral to a community health worker (CHW) program for resource alignment.

METHODS: We sampled clinical notes and portal messages of 200 age/sex-stratified patients, retrieved from electronic health records. The base cohort consisted of adult patients seen at a large academic primary care practice in the midwest over a two-year period and referred to the CHW program for unmet social needs. Two nurse abstractors annotated five domains of unmet social needs in the clinical text. The annotations served as the gold standard for developing and evaluating a similarity- and a rule-based NLP program. We first explored a sentence similarity approach for efficiency of detecting the unmet social needs, and for comparison, a rule-based NLP algorithm was developed to identify the five domains of unmet needs. We evaluated the performance of both the approaches in identifying each as well as any of the five domains. Precision, recall, and F1-score were computed on the test set. Error analysis was performed at the patient level.

RESULTS: During the calibration stage, the two abstractors dually annotated 1,055 clinical notes and 100 portal messages. The annotated cohort consisted of 5,675 clinical notes and 475 portal messages, with an estimated kappa statistics agreement of 0.938. Among the 200 patients, 168 (84%) had at least one domain of the unmet needs annotated. The similarity- and rule-based NLP achieved an F1-score of 0.91 and 0.95, respectively. Patients identified through this AI/NLP algorithm will be referred to a CHW program for resource alignment related to the unmet need.

CONCLUSIONS: The clinical text contains rich information about patients’ unmet social needs. NLP can achieve satisfactory performance in identifying those needs and candidates for CHW referral.

BARRIERS AND FACILITATORS TO ACCESSING AND MAINTAINING PERMANENT SUPPORTIVE HOUSING FOR PEOPLE EXPERIENCING HOMELESSNESS WITH HIGH HEALTH NEEDS

Emily Scott¹; Thomas W. Gray³; Emma Allen-Morgan⁴; Alyssa Boral³; Sarah A. Stella²

¹Department ofMedicine, University of Colorado, Denver, CO; ²Department of Medicine, Denver Health, Denver, CO; ³Office of Research, Denver Health, Denver, CO; ⁴Quality Assurance, Colorado Coalition for the Homeless, Denver, CO. (Control ID #3875834)

BACKGROUND: People experiencing homelessness (PEH) have a large burden of both acute and chronic disease, and increased mortality rates compared to housed populations. A subset of PEH experience exceptionally high levels of unmet physical and behavioral health needs resulting in frequent utilization of high cost medical and other services. Permanent supportive housing is a critical foundation for addressing health and health-related social needs, yet PEH with high health needs may face additional barriers to accessing housing.

METHODS: We conducted semi-structured interviews with clients following entry into permanent supportive housing, and homeless service providers involved in the permanent housing process in the Mountain West region. Participants were recruited via snowball sampling through community and academic partners for both providers and clients. Interviews aimed to assess client and provider perspectives on barriers and facilitators to accessing and maintaining housing within our local coordinated entry system. Interviews were conducted by trained research staff using an interview guide. Interviews were audio-recorded, transcribed, and de-identified. Transcripts were coded in Atlas.ti and themes were inductively derived using grounded theory.

RESULTS: Interviews were completed with 15 service providers and 14 recently housed clients. Themes regarding the housing process for clients with high health needs included: 1) PEH with high health needs face increased barriers in accessing and navigating the coordinated entry system for housing services; 2) the standard assessment tool (i.e., VI-SPDAT) used to prioritize clients for housing does not adequately capture the vulnerability of many high needs clients, especially those with cognitive impairment; and 3) PEH with high needs are often placed with an inadequate level of support due to a lack of availability, have needs that exceed the level of care offered by the system, and/or experience housing loss rooted in a mismatch between client needs and preferences and services available. Recommendations for improvement included: 1) start the coordinated entry process during medical encounters (eg. hospitalizations) to improve access for PEH who frequently use medical services; 2) develop processes to utilize care team perspectives and existing contextual health data (in addition to self-report) to inform housing prioritization; and 3) partner with external agencies to address care gaps in supportive services in permanent housing.

CONCLUSIONS: PEH with the highest health needs face multiple barriers to permanent supportive housing and our findings highlight opportunities for improving the housing process for this at-risk group.

BEYOND SCREENING AND REFERRAL: A PRAGMATIC FRAMEWORK FOR HEALTH CARE SYSTEMS TO DETERMINE THE ACTIONABILITY OF SOCIAL RISKS

Mayuree Rao^1,2; Matthew Maciejewski^3,4; Karin Nelson^1,2; Leah M. Marcotte²; Donna Zulman^6,5

¹Health services research & development, VA Puget Sound Health Care System, Seattle, WA; ²Medicine, University of Washington, Seattle, WA; ³HSR&D, Durham VA Medical Center, Durham, NC; ⁴Duke University, Durham, NC; ⁵Medicine, Stanford University, Stanford, CA; ⁶Health Services Research & Development, VA Palo Alto Health Care System, Palo Alto, CA. (Control ID #3877226)

BACKGROUND: While health systems are increasingly screening patients for social risks, determining how to meaningfully address social risks identified through screening can be a challenge. Few frameworks exist to guide health system action in response to social risks.

METHODS: We conducted a narrative review of English-language literature indexed in PubMed from 2000 to 2021 that describes and classifies health system activities to address social risks. Based on critical appraisal and synthesis of the literature, we developed a pragmatic framework to characterize actionability of social risks within health systems.

RESULTS: The framework highlights 4 dimensions of social risk actionability: level of action, actor, purpose, and action (Figure). First, social risks can be addressed at several levels (e.g., during a patient’s visit, at the practice/institution level through programs, and at the community level through partnerships). Second, health system staff, or “actors” (including health professionals with direct patient contact, clinical and institutional leaders, and researchers) can take action in different ways depending on their role. Third, intervention on social risks can serve one or more purposes: to strengthen relationships with patients, tailor care, modify the social risk itself, and/or facilitate population health surveillance, research, and advocacy. Finally, specific actions on social risks vary by level, actor, and purpose. For example, at the encounter level, a primary care provider may schedule telehealth appointments for the purpose of tailoring care for a patient with a transportation barrier. At the institution level, an administrator may implement a program to provide transportation to medical appointments to reduce transportation barriers for all patients. At the community level, institutional leaders may advocate for public transit investment.

CONCLUSIONS: Drawing from the literature, the proposed framework offers a novel, structured approach to identify existing and potential actions to address social risks, for multiple purposes and across different roles and levels within a health care organization.

BUNDLING COLORECTAL CANCER SCREENING OUTREACH WITH SCREENING FOR SOCIAL RISK IN FEDERALLY QUALIFIED HEALTH CENTERS: A STEPPED-WEDGE IMPLEMENTATION-EFFECTIVENESS PILOT STUDY

Sanja Percac-Lima²; Gina R. Kruse¹; Daniel Gundersen³; Lynette Mascioli⁵; Deepinder Singh⁶; Mehezbin Munshi⁵; Karen Emmons⁴; Jennifer Haas^1,2

¹Medicine, Massachusetts General Hospital, Boston, MA; ²Division of General Internal Medicine, Massachusetts General Hospital, Boston , MA; ³Dana Farber Cancer Institute, Boston, MA; ⁴Harvard University T H Chan School of Public Health, Boston, MA; ⁵Massachusetts League of Community Health Centers, Boston, MA; ⁶University of Vermont Larner College of Medicine, Burlington, VT. (Control ID #3876061)

BACKGROUND: Bundling outreach for colorectal cancer (CRC) screening with screening for social determinants of health (SDOH) may enable patients to engage in preventive care by addressing SDOH and produce efficiencies in resource-constrained settings.

METHODS: In this clustered stepped-wedge trial, four Massachusetts federally-qualified health centers (FQHCs) were randomized to start implementation of a bundled outreach intervention over 8-week “steps” (10/2020-11/2021). The intervention consisted of outreach to 50-75 yr-olds not up-to-date on CRC screening to offer home fecal immunochemical testing (FIT) and SDOH screening. The implementation strategy used external facilitation and technical assistance for data reporting for each FQHC over two phases: 1) initial implementation (months 1-4) and 2) intervention adaptation guided by data to identify inequities in reach and effectiveness (months 5-8). The primary outcome was CRC screening completion by any guideline- recommended method. We compared screening rates in intervention and control FQHCs in each step by fitting generalized linear mixed effects models with random intercepts for FQHCs and patients to account for clustering of observations within FQHCs and multiple measurements among patients.

RESULTS: Each FQHC had a slightly different model of implementation and outreach depending on their unique infrastructure, workflows, capacity and pandemic-related staffing demands. Two implemented the pilot using population health approaches with outreach calls and two integrated the intervention within established teams and projects, such as pairing with high-risk patient community health worker programs and pre-visit planning. Of 32,524 unique patients who were not up-to-date with CRC screening in the 4 FQHCs, 53% were female; 22% Black, 14% Latino, 12% Asian, and 52% white; 48% were on public insurance, 33% private insurance, and 11% uninsured. The average rate of CRC screening completion was higher among intervention FQHCs than among control FQHCs. The odds ratio comparing CRC screening orders in intervention to control FQHCs was 2.5 (p<0.001), with an average marginal effect of 3 percentage point increase in CRC screening. A sensitivity analysis excluding pre-covid baseline periods found a slight attenuation of the effect but the same sign and significance at p<0.05.

CONCLUSIONS: This bundled intervention represents an effective, population-based solution to increase the delivery of cancer screening with consideration of the impacts of individual social risks. This strategy has potential to ameliorate CRC screening deficits associated with the pandemic.

BUT WHAT DOES INCLUSION LOOK LIKE?: INSIGHTS FROM HOSPITALISTS

Ashley Jenkins^1,2; Maya Narayanan³; Anne S. Linker⁴; Amelita Woodruff⁵; Archna Eniasivam^6,7; Benji Mathews⁸; David Sterken³; Gopi Astik⁹; Michelle Fletcher⁹; Nila Radhakrishnan¹⁰; Marisha Burden¹¹; Sarah Saari¹²; Andrew Auerbach⁶; Adriana Dhawan⁸; Areeba Kara¹³

¹Medicine, University of Rochester School of Medicine and Dentistry, Rochester, NY; ²Pediatrics, University of Rochester School of Medicine and Dentistry, Rochester, NY; ³University of Wisconsin-Madison, Madison, WI; ⁴Medicine, Icahn School of Medicine at Mount Sinai, New York, NY; ⁵Johns Hopkins Medicine, Baltimore, MD; ⁶Medicine, University of California San Francisco, San Francisco, CA; ⁷Pediatrics, University of California San Francisco, San Francisco, CA; ⁸University of Minnesota Medical School Twin Cities, Minneapolis, MN; ⁹Medicine, Northwestern University Feinberg School of Medicine, Chicago, IL; ¹⁰Medicine, University of Florida, Gainesville, FL; ¹¹Medicine, University of Colorado Denver School of Medicine, Aurora, CO; ¹²University of Washington, Seattle, WA; ¹³Department of Medicine, Indiana University School of Medicine, Indianapolis, IN. (Control ID #3861003)

BACKGROUND: Hospitalists often serve marginalized patients and have system-wide influence to promote diversity, equity, inclusion, and justice (DEIJ). Inclusion in particular has definition variability and no clear metrics of ‘success’. We sought to describe hospitalists’ perspectives about what constitutes an inclusive work environment.

METHODS: We conducted a rapid thematic analysis of open-ended questions from a cross-sectional electronically disseminated survey study examining DEIJ efforts in US hospital medicine groups. Survey participants were recruited from members of a US hospital medicine research collaborative in May 2021. Respondents were asked to elaborate upon survey item ratings (5-point Likert scale) related to their hospital medicine group’s inclusivity. Responses were summarized in a matrix and stratified by rating to allow for thematic analysis by a priori defined levels of self-rated inclusivity. High inclusivity (HI) contained extremely or moderately inclusive ratings. Low inclusivity (LI) comprised somewhat or not at all inclusive ratings.

RESULTS: Of the 436 respondents, 332 (76.5%) rated their site as having HI, 102 (23.4%) described their sites as having LI. 139 (32%) responded to the open-ended questions. Four themes emerged that informed perceptions of inclusion (Table): tangible efforts, diversity of the group, leadership, and personal experience and the workplace culture.

CONCLUSIONS: Frontline hospitalists identified investment in tangible efforts to further DEIJ for clinicians and patients as a necessary aspect of an inclusive hospital medicine group. Diversity within the ranks and leadership is valued as a marker of inclusivity. Workplace culture and day-to-day patient care experiences need alignment with the values of inclusion. Leaders play a significant role in furthering or thwarting these efforts.

CANCER INCIDENCE AMONG A SOMALI IMMIGRANT POPULATION IN MINNESOTA Ahmed A. Mohamed¹; Alanna M. Chamberlain²; Gregory Jenkins²; Lila J. Rutten^3,4; Jane Njeru¹; Mark L. Wieland¹

¹Department of Medicine, Mayo Clinic Minnesota, Rochester, MN; ²Quantitative Health Sciences, Mayo Clinic Minnesota, Rochester, MN; ³Exact Sciences Corporation, Madison, WI; ⁴Mayo Clinic Minnesota, Rochester, MN. (Control ID #3876771)

BACKGROUND: Somali immigrants and refugees constitute one of the largest African groups settling to the United States over the past three decades with the majority resettling in the state of Minnesota. In our primary care practice, we have documented significant disparity gaps in cancer screening among Somali patients compared to the general population. However, there is little known about cancer incidence among Somalis living in the US.

The objective of this study is to investigate cancer incidence among Somali patients as compared to non- immigrant patients in Olmsted County utilizing the Rochester Epidemiology Project (REP) medical records- linkage infrastructure.

METHODS: We determined the incidence of each cancer diagnosis using ICD-9 and ICD-10 codes and compared them between Somali and non-Somali populations during the years 2000-2020 in Olmsted County.

A person was defined as Somali if one or more of these conditions were met:

- one or more their records indicated that Somali was their country of origin

- one or more of their records indicated that they speak Somali

- one or more of their records indicated Somali race or ethnicity

- the text string “Somali” was found in a visit note from Olmsted Medical Center

Poisson regression was used to model the rate of a particular malignancy predicted by: year, age, gender and Somali descent. Relative risks for malignancy comparing Somali to the non-Somali Olmsted County population were calculated from the Poisson regression models.

RESULTS: Malignancies with statistically significant higher or lower incidence among Somali patients versus non-Somali patients are shown in Table 1. Liver and gastric malignancies have a higher incidence while hematologic, breast, cervical, lung, and melanoma have a lower incidence among the Somali population.

CONCLUSIONS: Malignancies related to infectious agents such as viral hepatitis have a higher incidence and prevalence in the Somali immigrant population of Olmsted County. Viral hepatitis is prevalent in the Somali population and is associated with hepatocellular carcinoma. Therefore, screening for and treating viral hepatitis can help reduce the incidence of liver cancer. There is a lower incidence and prevalence of malignancies related to lifestyle factors in the Somali population. Future studies on acculturation and cancer incidence in this population to evaluate this further are needed.

CAN HEALTH CONFIDENCE HELP PREDICT HOSPITAL LENGTH OF STAY OR READMISSION?

Ashley E. Brown, David O. Meltzer

Hospital Medicine, University of Chicago Division of the Biological Sciences, Chicago, IL. (Control ID #3873979)

BACKGROUND: It is well established that patients who are engaged or “activated” in their healthcare have better health outcomes and patterns of healthcare utilization. Additionally, patient activation may be associated with a range of socioeconomic factors affecting health outcomes and utilization. The Health Confidence Score (HCS) is a recently created tool that simply measures a component of health activation, health confidence. The HCS is a four-question survey at a basic literacy level, making it well-suited to monitor patient engagement within socio-economically vulnerable populations. We aim to understand if health confidence can help predict hospital readmission and length of stay.

METHODS: This prospective cohort study was performed at an urban academic medical center in the Midwest. The HCS was administered, as a part of a larger ongoing survey, to a sample of all inpatients admitted to the hospitalist service from January 21, 2022, to October 23, 2022. The survey data was combined with visit characteristics obtained from the electronic medical record. Population characteristics were defined by frequencies or mean and standard deviation (SD). We used bivariate analysis with a t-test to establish basic associations. Multivariate logistic regression controlling for sex, race, ethnicity, and age was used to assess the relationship of HCS with readmissions. Multivariate linear regression controlling for the same demographics was used to assess the relationship of HCS with hospital length of stay. Statistical significance was defined by p<0.05.

RESULTS: A total of 2,796 patients completed the HCS, of whom 55% were female, 76% were Black/African American, and 94% were not of Hispanic/Latino ethnicity. The mean age was 60 years (SD 17, range 19-105). During the study period, the mean HCS was 9.19 (SD 2.68, range 0-12) and 59% had an HCS≥9. Readmission occurred in 22% (601 patients) and the average length of stay was 7.32 days (SD 7.27, range 0-91). Patients with above average HCS (HCS≥9) had fewer readmissions (26% vs. 29%; p=0.162) with and without controlling for demographic confounders (OR 0.88; 95% CI [0.73, 1.06]; p=0.174), but these were not statistically significant. However, patients with an HCS≥9 had a significantly lower length of stay (6.71 vs. 8.33, p<0.001) compared to those with an HCS<9. This remained true when accounting for demographic confounders: patients with an HCS≥9 had an average length of stay 1.64 days (95% CI [2.2, 1.06]; p<0.001) shorter than those with an HCS<9.

CONCLUSIONS: To the best of our knowledge, this is the first study looking at the relationship between HCS, hospital readmission, and length of stay. While HCS was not significantly associated with readmission, we did find a strong association between HCS and hospital length of stay. We aim to assess whether HCS can have clinical utility in developing and targeting interventions to reduce hospital length of stay and improve outcomes for hospitalized patients.

CAN THE LANGUAGE BARRIER IMPACT THE READMISSION RATE IN COLON CANCER PATIENTS? – A SINGLE-INSTITUTION ANALYSIS

RAJMOHAN RAMMOHAN¹; Melvin Joy¹; Atul Sinha²; Tulika Saggar¹; Dilman K. Natt¹; Susan Bunting¹; Prachi Anand¹

¹INTERNAL MEDICINE, NASSA UNIVERSITY MEDICAL CENTER, East Meadow, NY; ²Internal Medicine, Nassau University Medical Center, East Meadow, NY. (Control ID #3875329)

BACKGROUND: Patients diagnosed with colon cancer face many challenges and need a good understanding of their diagnosis and proposed treatments to make informed decisions about their care. Health literacy is vital, and low health literacy may be associated with poorer outcomes. There is a concern that language barriers and literacy rates influence the readmission rate of colon cancer. We examine the 30-day readmission rate, 60-day readmission rate, and 90-day readmission rate regarding patients’ language and also to identify the health equity for any disparities.

METHODS: In our study, we performed a retrospective analysis of all Colon Cancer patients admitted to our hospital from 2009- 2022. Data on comorbidities, insurance status, surgical procedures, Cancer grade, colonoscopy reports, and baseline characters were collected. Patients were excluded if they died; Data was then divided based on the patient’s language. A Chi-Square test and odds ratio was performed on the entire cohort, and a matched subset was examined with propensity scoring matching to balance baseline characteristics.

RESULTS: A total of 701 patients were included in the study; 341(49%) were female, and the average median age group (was 55.6±5.9 years). Patients were divided into four groups based on their language. Propensity score matching was used to balance the baseline characteristics. Patients speaking English had significantly higher 30day readmission rates compared to the other Languages (47 vs. 7vs. 15 vs. 14, P=0.025, Correlation: 0.68). English had a significantly higher 60day readmission rate than the other Languages (25 vs. 3vs. 8 vs. 30, P<0.01, Correlation: 0.78). English had a significantly higher 90day readmission rate than the other Languages (10 vs. 2vs. 11vs. 10, P<0.01, Correlation: 0.55). Among Non- English speakers, Patients with insurance status had significantly higher readmission status than the non- insured patients (OR: 3.025, p<0.021).

CONCLUSIONS: This study is a promising start to study the impact of readmission rates in colon cancer patients based on language barriers. Understanding linguistic capabilities are crucial to bridge the gap between healthcare disparities. Further research on language discordance is needed to reduce the risk of language-discordant clinical encounters and the adverse health outcomes associated with these encounters.

CHALLENGES IN ACCESS TO COLONOSCOPY AT A STUDENT-RUN FREE CLINIC

Andrew B. Yang¹; Isaac A. Nathoo²; Ramia Rahman¹; Elizabeth K. Benitez²; Tiffany Merlinsky³; Ash*ta S. Batavia^2,5; Pamela Charney⁴

¹Weill Cornell Community Clinic, Weill Cornell Medicine, New York, NY; ²Medicine, Weill Cornell Medicine, New York, NY; ³Community Clinic, Weill Cornell Medicine, New York, NY; ⁴Internal Medicine, Weill Cornell Medical College, New York City, NY; ⁵Janssen Global Services LLC, Titusville, NJ. (Control ID #3875453)

BACKGROUND: This study looks at access to colonoscopies in an academic-based medical student-run free clinic that serves an uninsured population. Colonoscopy is critical for early identification of colorectal cancer, and delays in colonoscopy are associated with increased colorectal cancer (CRC) incidence and mortality. As such, we investigated referral and fulfillment rates at the clinic.

METHODS: This was a retrospective study of all colonoscopy referrals from 1/1/17 to 12/31/21. Demographic, colonoscopy referral, and colonoscopy procedure data were abstracted from the EHR by the authors. Up to May 2021, patients over 50 years were referred for screening colonoscopies, with the age changing to 45 years following a change in USPSTF guidance. Patients with a positive family history were referred for screening colonoscopies at age 40 or 10 years before the age that the immediate family member was diagnosed. Diagnostic colonoscopies, by contrast, are symptom driven. The time to procedure was defined as the number of days between the initial referral and the colonoscopy procedure.

RESULTS: During the study period, there were 150 referrals for 84 unique patients (33% male). 14 of these patients had a positive family history. The average number of referrals per patient was 1.8. 80% of referrals were diagnostic, and 55 referrals were repeats, meaning that multiple referrals were given for a single colonoscopy. This resulted in 95 remaining referrals, which we refer to as unique referrals. Amongst the unique referrals, 55% of diagnostic, 55% of screening, and 71% of referrals for a patient with relevant family history resulted in a colonoscopy. The median age at initial referral was 55 (interquartile range 50-61).

In total, 52 colonoscopies (79% screening, 21% diagnostic) were completed in 47 unique patients. 22 colonoscopies resulted in abnormal findings with a changed interval for a repeat colonoscopy, including 8 with tubular adenoma and 11 with polyps . There were 5 patients referred for 2 separate colonoscopies during the timeframe of the study due to a combination of family history of CRC and high-risk lesions. The average time between initial referral and colonoscopy was 329 days.

CONCLUSIONS: Our study found that 150 colonoscopy referrals resulted in 52 colonoscopies, for a 35% conversation rate from colonoscopy referral to completed procedure, which is worse than described for uninsured & underinsured Americans. Furthermore, the conversion rate is worse for diagnostic colonoscopy referrals as they accounted for 55% of unique referrals but only 21% of completed procedures. Furthermore, among the 47 patients that received colonoscopies, the time to procedure was nearly 3-fold longer that what has been found in the literature. In our population, the rate of pathologic findings was comparable to the literature. Given these findings, additional research is needed to understand the barriers to care that drive decreased and delayed follow-through with screening & diagnostic colonoscopies.

CHALLENGES IN REFUGEE HEALTH: LESSONS FROM GREECE IN BUILDING TRUST AND COLLABORATION IN THE INTERNATIONAL HUMANITARIAN SETTING

Natalie M. Colaneri¹; Essam Daod²; Hrayr Attarian³; Jess Ghannam⁴; Shannon Galvin⁵

¹Internal Medicine, University of California San Diego, La Jolla, CA; ²Humanity Crew, Haifa, Israel; ³Neurology (Sleep Medicine), Northwestern University Feinberg School of Medicine, Chicago, IL; ⁴Psychiatry and Global Health Sciences, University of California San Francisco, San Francisco, CA; ⁵Infectious Diseases, Northwestern University Feinberg School of Medicine, Chicago, IL. (Control ID #3876449)

BACKGROUND: Greece is one of many countries facing an ongoing humanitarian health crisis involving displaced people. More than 50,000 asylum seekers and refugees currently reside in Greece, and many suffer from physical as well as mental health concerns. Humanitarian aid workers from around the world are currently working together in an increasingly complex system in Greece, trying to provide support to asylum seekers and refugees from a diverse array of countries. This qualitative study explores ideas to build trust and collaboration in the international humanitarian setting.

METHODS: This qualitative study involved semi-structured interviews with 28 international aid workers who currently or previously worked in Greece. Subjects were recruited through email, word-of-mouth, and utilizing snowballing sampling. Responses to the questions “What can aid workers do to build more trust with refugees?” and “How can aid workers work together to support refugees?” were analyzed using thematic analysis. The study was approved by the Northwestern University Institutional Review Board.

RESULTS: Three major themes emerged regarding ways for aid workers to build more trust with displaced communities: 1. promoting empowerment/agency of asylum seekers and refugees, 2. learning about various aspects of their cultures, and 3. demonstrating certain character traits. Ideas for promoting empowerment/agency of individuals included: asking about needs of communities to guide the services provided, providing individuals with clear information in their language, providing options and choices to individuals, and aid workers viewing their work as "solidarity" rather than “helping” individuals. Many respondents also viewed learning about various aspects of culture as key to building trust, including learning language and exploring traditions and activities. Finally, certain character traits were integral to building trust, including openness, respect, honesty/authenticity, and professionalism. Respondents also suggested ways that aid workers from different backgrounds could work together to support displaced communities, including utilizing specific strengths of each team member, communicating openly, and sharing knowledge with each other.

CONCLUSIONS: This study explores different ways for aid workers to build trust with the communities they serve, including empowering individuals, prioritizing learning about the cultures of different communities, and demonstrating traits including openness and honesty. This research also suggests ideas for collaboration in the humanitarian field, including utilizing strengths of different team members, communicating openly, and sharing knowledge. This study will hopefully lead to further discussion regarding ways to build trust in the international humanitarian setting, and allow aid workers to more effectively work together to improve the health and well-being of asylum seekers and refugees in Greece and around the world.

CHANGES IN DISPARITIES IN OPIOID PRESCRIBING FOR LOW-ACUITY CONDITIONS BEFORE AND AFTER RANDOMIZED CLINICIAN FEEDBACK INTERVENTIONS

Aidan P. Crowley¹; Chuxuan Sun^1,2; Xiaowei Yan⁵; Amol S. Navathe^1,3; Joshua M. Liao^4,3; Mitesh S. Patel⁶; David Pagnotti¹; Zijun Shen⁵; M. K. Delgado^3,7

¹Medical Ethics and Health Policy, University of Pennsylvania Perelman School of Medicine, Philadelphia, PA; ²Division of General Internal Medicine, University of Pennsylvania Department of Medicine, Philadelphia, PA; ³Leonard Davis Institute of Health Economics, University of Pennsylvania, Philadelphia, PA; ⁴University of Washington Department of Medicine, Seattle, WA; ⁵Center for Health Systems Research, Sutter Health, Sacramento, CA; ⁶Ascension Health, St Louis, MO; ⁷Department of Emergency Medicine and Department of Biostatistics, Epidemiology, and Informatics, University of Pennsylvania Perelman School of Medicine, Philadelphia, PA. (Control ID #3877127)

BACKGROUND: Racial and ethnic minorities historically receive opioid prescriptions at lower rates and dosages than white patients. Stewardship interventions to decrease excessive opioid prescribing may either improve or exacerbate these disparities. However, there is little evidence about these potential differential effects. The objective of our study was to determine how a randomized trial of opioid stewardship clinician feedback interventions that reduced opioid prescribing overall affected patients differently by race and ethnicity.

METHODS: This study used data from a 4-arm cluster randomized clinical trial conducted among 438 clinicians from 21 emergency departments (EDs) and 27 urgent care clinics (UCs) in the western U.S. The clinical trial examined the effects of 3 interventions on opioid prescribing patterns by randomly assigning eligible physicians, physician assistants, and nurse practitioners by ED or UC site to 1 of 4 study arms: usual care, individual audit feedback, peer comparison feedback, or a combination of both forms of feedback. We analyzed data from a 6-month baseline period (Mar-Sep 2019) and a 6-month intervention period (Sep-Mar 2020). The primary outcome was number of pills prescribed (low ≤ 10 pills, medium = 11-19 pills, high ≥ 20 pills). Generalized mixed-effects logistic regression models were used to examine patient characteristics associated with low-pill prescriptions during the baseline period. These model types were also applied to intervention period data to determine whether pills per prescription varied by patient race and ethnicity obtained from the electronic health record (white, Black, Asian, Hispanic, other) between usual care and each of the 3 intervention arms.

RESULTS: Compared with white patients, Black patients were more likely to receive a low-pill prescription during both the baseline (AOR: 1.18, 95% CI: 1.06 to 1.31, p = .002) and intervention (AOR: 1.43, 95% CI: 1.07 to 1.91, p = .015) periods. No other racial or ethnic group was significantly more likely than white patients to receive a low-pill opioid prescription at baseline. While combined feedback led to an overall increase in low-pill prescriptions (AOR: 1.89, 95% CI: 1.28 to 2.78, p = .001), there were no significant differences in any intervention’s treatment effects by patient race and ethnicity as identified by individual interaction terms (Asian AOR: 0.83, p = .30; Black AOR: 0.88, p = .57; Hispanic AOR: 0.86, p = 0.24; Other AOR: 0.80, p = 0.21).

CONCLUSIONS: Combined individual audit and peer comparison feedback reduced the number of opioid pills per prescription equally by patient race and ethnicity. However, this intervention did not significantly ameliorate the baseline disparity we identified in prescribing by race. Future studies should determine whether behavioral interventions can reduce prescription disparities by evaluating standardized dosing strategies or stratified clinician feedback by patient characteristics.

CHASING THE OUNCE OF PREVENTION: THE EFFECT OF THE COVID-19 PANDEMIC ON PREVENTIVE IMAGING

Jenny L. Weon¹; Yin Xi²; Yee Seng Ng²; Lauren N. Cooper³; Marlon I. Diaz⁶; Robert W. Turer⁴; Samuel McDonald⁴; Duwayne L. Willett⁵; Christoph U. Lehmann³; Ling Chu⁵

¹Pediatrics, The University of Texas Southwestern Medical Center, Dallas, TX; ²Radiology, The University of Texas Southwestern Medical Center, Dallas, TX; ³Clinical Informatics Center, The University of Texas Southwestern Medical Center, Dallas, TX; ⁴Emergency Medicine, The University of Texas Southwestern Medical Center, Dallas, TX;⁵Internal Medicine, The University of Texas Southwestern Medical Center, Dallas, TX; ⁶Texas Tech University Health Sciences Center El Paso Paul L Foster School of Medicine, El Paso, TX. (Control ID #3873220)

BACKGROUND: The COVID-19 pandemic was associated with substantially reduced preventive screening activities. The 2022 AACR Cancer Disparities Progress Report urged researchers to identify how the COVID-19 pandemic affected health disparities in cancer. Our objective was to determine if the pandemic exacerbated health disparities in preventive screening (including cancer screening).

METHODS: We extracted weekly volumes of preventive screening exams (bilateral mammograms and bone density exams) between January 7, 2019 and May 1, 2022 from the radiology data warehouse at an academic medical center in the Southern region. Patients were grouped by race and ethnicity into Hispanic or Latino, non-Hispanic Black, and non-Hispanic White patients. The March 16-May 9, 2020 period was designated as the COVID-19 shutdown. Ratios of post- to pre-shutdown mean weekly volumes were estimated for each subgroup using binomial regression. All analyses were done in SAS 9.4.

RESULTS: For mammograms, post-shutdown volumes were similar to pre-shutdown volumes for all race and ethnicity groups (Hispanic and Latino, ratio of means 1.08, [95% CI 0.90, 1.29], p = 0.99; non-Hispanic Black, ratio of means 1.04, [95% CI 0.87, 1.24], p = 1.00; non-Hispanic White, ratio of means 1.03, [95% CI 0.86, 1.23], p = 1.00). For bone density exams, volumes significantly increased post-shutdown to more than 3-fold for Hispanic patients (ratio of means 3.40, [95% CI 2.77, 4.17], p < 0.0001) and to 2-fold for non- Hispanic Black patients (ratio of means 2.00, [95% CI 1.63, 2.44], p < 0.0001), while volumes significantly decreased post-shutdown to 67% of pre-shutdown volume for non-Hispanic White patients (ratio of means 0.67, [95% CI 0.56, 0.82], p = 0.002).

CONCLUSIONS: In our population there were no significant differences between post- and pre-shutdown volumes in race and ethnicity groups for mammograms and we observed a significant increase in weekly screenings for Hispanic or Latino and non-Hispanic Black patients in contrast to non-Hispanic White patients for bone density exams. The increase of bone density screening for Hispanic or Latino and Black patients over time remains unexplained and requires further examination.

COMMUNICATED BIAS IN THE ELECTRONIC HEALTH RECORD FOR PATIENTS PRESENTING WITH PAIN

Pooja A. Lagisetty^1,2; Adrianne R. Kehne³; Colin macleod¹; Christopher Fung⁴

¹Internal Medicine, University of Michigan, Ann Arbor, MI; ²Veterans Health Administration, Ann Arbor, MI; ³Center for Clinical Management Research, VA Ann Arbor Healthcare System, Ann Arbor, MI; ⁴University of Michigan Michigan Medicine, Ann Arbor, MI. (Control ID #3876534)

BACKGROUND: There are over 34 million pain-related ED encounters annually. Pain-related stigma and racial discrimination are increasingly recognized as causes of inequity in pain-related care. These biases may be documented in the medical record using stigmatizing language (e.g., “drug seeker,” “real” pain), and may negatively impact downstream general medical care by communicating biases about patients to other providers. Thus we aimed to measure the prevalence of stigmatizing language in notes written upon patients’ initial presentation to the emergency department and evaluate predictors for its presence within these notes.

METHODS: A lexicon of stigmatizing language was determined a priori from the existing literature, expert opinion, and investigator consensus and classified into seven categories: discounting symptoms, commenting on patient behavior, medical stigma, negative patient descriptors, socioeconomic indicators, substance use disorder stigma, and weight-related stigma. We included ED encounters from a large academic center (2017-2021) of patients presenting for the five most common pain complaints: chest pain, abdominal pain, headache, back pain, and injury. We used descriptive statistics to describe the frequency of stigmatizing language and multivariable analysis to determine predictors (demographics, acuity of presentation via the Emergency Severity Index, pain type, and insurance type) of such language.

RESULTS: There were 51,570 pain-related encounters in the sample. The mean triage note length was 48.4 words (SD = 20.5) and 2691 notes (5.2%) contained stigmatizing language. Of the pain conditions examined, chest pain had the highest prevalence of stigmatizing language in encounter notes (6.66% of chest pain encounters). The most common category of stigmatizing language was negative patient descriptors (e.g., “refuse,” “insists”) which comprised 30.4% of stigmatizing language. Language around substance use disorders (e.g., “habit,” “abuse”) was also common (19.7%). In a logistic regression model, notes were significantly more likely to have stigmatizing language when the patient was Black (OR=1.10, p=.04) compared to White patients, male (OR=1.15, p<0.001), a higher acuity score (ESI of 2 vs. 4; OR=2.48, p<0.001), or had Medicaid (OR=1.29, p<.001) or Medicare (OR=1.22, p<0.001) as compared to private insurance.

CONCLUSIONS: Stigmatizing language is present in roughly 1 in 20 ED encounters for pain and is disproportionately prevalent in specific patient populations. Future studies should explore whether this language is copied forward to general medical history and physical note documentation and if this impacts pain-related treatment outcomes such as time to analgesia and type of analgesics delivered.

COMMUNICATION AND ACTIVATION IN PAIN TO ENHANCE RELATIONSHIPS AND TREAT PAIN WITH EQUITY (COOPERATE): A RANDOMIZED CLINICAL TRIAL

Marianne S. Matthias^1,2; Joanne K. Daggy⁵; Matthew J. Bair^3,2; Diana J. Burgess⁴; Johanne Eliacin¹; Kevin L. Rand⁶; Laura J. Myers¹; Michelle Salyers⁶; Adam Hirsh⁶

¹HSR&D, Roudebush VAMC, Indianapolis, IN; ²Medicine, Indiana University School of Medicine, Indianapolis, IN; ³Medicine, Center for Health Information and Communication, Indianapolis, IN; ⁴Medicine, University of Minnesota and Minneapolis Veterans Affairs Health Care System, Minneapolis, MN; ⁵Biostatistics, Indiana University School of Medicine, Indianapolis, IN; ⁶Psychology, Indiana University Purdue University Indianapolis, Indianapolis, IN. (Control ID #3852665)

BACKGROUND: Chronic pain affects tens of millions of Americans and is a leading cause of disability. Racial disparities in pain treatment have been extensively documented, and Black patients exhibit lower levels of patient activation—having the knowledge, confidence, and skills to self-manage—than White patients. Lower patient activation leads to less effective clinical communication (e.g., sharing concerns, shared decision-making), which can adversely affect pain management.

METHODS: COOPERATE is a randomized trial of an intervention to increase patient activation for Black Veterans with chronic pain. It includes 6 individual coaching sessions delivered via phone over 3 months, focusing on two essential skill sets to facilitate patient activation: 1) goal setting/prioritization, 2) clinical communication. 250 Black Veterans were randomized to the intervention or to attention control. The primary outcome was patient activation; secondary outcomes included communication self-efficacy, pain, and psychological functioning. Outcomes were assessed at baseline, 3 (primary endpoint), 6, and 9 months (sustained effects). A linear mixed-model repeated measures approach was used to test change in patient activation compared to the control group at 3 months adjusting for baseline working alliance and number of comorbid conditions. Group differences in mean change from baseline at 6 and 9 months were estimated. Secondary outcomes were analyzed similarly, adjusting for multiple comparisons using the Šídák method.

RESULTS: Patient activation improved significantly relative to the control group from baseline to 3 months (+4.6 vs. +0.13, p=.03). Differences were sustained at 6 (p=.001) and 9 months (p=.03). Communication self- efficacy improved (+3.63 vs. +.92, p=.046) at 3 months. Although improvements persisted, they were not significantly different from the control group at follow-up points after Šídák adjustment. Pain intensity and interference improved more in the intervention group at 3 months (-0.65 vs. -0.04), but differences were not significant after adjustment. Depression and anxiety improved in intervention participants and worsened in in control participants, but group differences were not significant.

CONCLUSIONS: COOPERATE effectively increased patient activation and communication self-efficacy, with differences in patient activation sustained over time. Given its importance for effective self- management, finding ways to increase patient activation is a key path to improved pain management. This is especially important for historically marginalized groups, such as Black patients, who report greater pain severity, worse pain care, and lower quality clinical communication than White patients. Consequently, COOPERATE represents a promising approach to decrease disparities in pain care and may be an important component of efforts to achieve health equity.

COMMUNICATION DISABILITY ACCOMMODATION PROGRAMS IN US HEALTHCARE ORGANIZATIONS

Jennifer Y. Osh*ta¹; Ashley Couture³; Megan Morris²

¹Medicine, Clinical and Translational Sciences, University of Vermont Larner College of Medicine, Burlington, VT; ²University of Colorado, Denver, CO; ³University of Vermont Medical Center, Burlington, VT. (Control ID #3876379)

BACKGROUND: Adults with communication disabilities in hearing, expression, reception, and cognition (PwCD), make up roughly 14-16% of the US adult population. They experience poorer healthcare outcomes when compared to their peers without communication disabilities. The Americans with Disabilities Act and the Joint Commission mandate equitable access to clear and mutual understanding of healthcare information through reasonable aids and services (Communication Disability Accommodation Programs). Despite these national mandates, few studies have examined how healthcare organizations have operationalized delivery of these aids and services at a systems level. This study aimed to describe the communication aids offered by US organizations and how delivery of accommodations have been implemented into organizational workflows.

METHODS: We recruited participants through purposeful sampling of disability coordinators through the Disability Equity Collaborative LEADERs workgroup, a learning collaborative of disability coordinators representing over 30 large healthcare organizations across the US. We completed semi-structured interviews with 19 disability coordinators representing 15 healthcare organizations from five regions of the US. Interviews were conducted over Zoom and lasted roughly 60 minutes. The transcripts of these interviews were coded and analyzed using qualitative content analysis. Codes were organized and clustered based on their relationships and linkages. Themes were drawn from analyzing these clusters.

RESULTS: Communication Disability Accommodation programs did not typically have dedicated staff or funding. The disability types most often addressed were Deaf, hearing impaired, or visual CDs. Leaders of these programs were most often directors of interpreter services. In two cases, a dedicated disability services program existed. Most organizations relied on providers or patient initiation to flag accommodation needs. In the inpatient setting, workflows were embedded within pre-existing language access practice patterns. In the outpatient setting, workflows were determined by individual clinics. Aids were often provided using a “Toolkit” strategy of a box of basic aids (e.g. a hearing amplifier, communication board, magnifier, white board) readily available at key locations. Delivery of accommodations to patients required coordinated efforts by multiple staff members across departments. Programs did not audit whether patients received accommodations, rather they relied on feedback via patient complaints. Barriers to program implementation included a lack of dedicated staffing and funding, hospital staff turnover, lack of provider awareness of CDs and accommodations, and reduced patient awareness of available accommodations.

CONCLUSIONS: Implementation of communication accommodations in these 15 US Healthcare organizations were limited to a subset of the CD population, typically embedded within language access workflows, and in early stages of development.

COMMUNITY-BASED COVID-19 TESTING AND EDUCATION AMONG FORMERLY INCARCERATED INDIVIDUALS – PRELIMINARY RESULTS FROM THE MOSAIC STUDY Connor S. Holmes¹; Maxwell Ackerman¹; Jordy Rojas-Antigua¹; Lindsey Riback¹; Chenshu Zhang¹; Micaela Linder²; Ronald F. Day²; Aaron Fox¹; Matthew J. Akiyama¹

¹Medicine, Albert Einstein College of Medicine, Bronx, NY; ²Fortune Society, Long Island City, NY. (Control ID #3876617)

BACKGROUND: The largest U.S. single-site COVID-19 outbreaks have occurred in carceral facilities. Many criminal legal system involved (CLSI) individuals live in congregate settings upon reentry and receive limited COVID testing and education. Community based organizations (CBOs) and community health workers (CHWs) provide essential reentry services to CLSI individuals and can contribute to COVID testing and education efforts. Our objective in this randomized controlled trial is to test the effectiveness of a co- located CHW-led COVID testing and education intervention in a reentry CBO. In this preliminary analysis, we aim to examine the vulnerability of the cohort thus far and attitudes towards COVID.

METHODS: We are recruiting 250 CLSI individuals released from jail or prison in the last 90 days. Participants are randomized to onsite Point-of-Care testing and education (O-PoC) or Standard of Care (SoC). Over one year, O-PoC participants receive quarterly testing onsite at the CBO and a CHW provides COVID-related counseling and education. SoC participants are referred for quarterly offsite testing at community sites. Here we report data from baseline assessments, including sociodemographics, comorbidities, prior COVID infection and exposure, and concerns about future infection.

RESULTS: From 4/2022-12/2022 we enrolled 114 participants, mean age is 42, 90% are male, 56% are Black and 35% are Hispanic. Over half (59%) report homelessness and 56% are on probation or parole. Participants report a mean 13 years of lifetime incarceration and 18 months incarceration since March 2020. One-third (36%) report receiving outpatient healthcare in the past 30 days, 61% report active tobacco use, 24% are asthmatic, and 5% are HIV+. One-third (32%) report ever testing positive for COVID (25% while incarcerated) and 60% lived with someone who had COVID while incarcerated. Most (79%) received a COVID vaccine and 47% have at least one booster. On average, participants believed there was a 22% chance they would contract COVID in the next 3 months with only 19% reporting ongoing worry about contracting COVID. On average, more participants worried about family or friends contracting COVID and transmitting COVID to others – 49% and 35%, respectively.

CONCLUSIONS: Preliminary data demonstrate high rates of past COVID exposure and infection while incarcerated with limited concern for future infection. High rates of unstable housing and comorbidities could suggest increased risk of infection and poor outcomes. Elevated COVID-related risk in this CLSI population reflects a need for a tailored CHW-delivered testing and education intervention.

COMMUNITY-BASED ORGANIZATION PERSPECTIVES ON PARTICIPATING IN STATE-WIDE COMMUNITY CANVASSING PROGRAM AIMED TO REDUCE COVID-19 VACCINE DISPARITIES IN CALIFORNIA

Lisa Mansfield²; Nisha Sunku¹; Savanna L. Carson³; Stefanie D. Vassar³; Dale Slaughter⁵; Gloria Kim⁵; Alana G. Troutt⁶; Keith C. Norris⁴; Arleen F. Brown⁴

¹Medicine, UCLA Medical Center Olive View, Sylmar, CA; ²Division of General Internal Medicine and Health Services Research, Department of Medicine, University of California Los Angeles, Los Angeles, CA; ³Division of Internal Medicine & Health Service Research, UCLA, Los Angeles, CA; ⁴Medicine, University of California Los Angeles, Los Angeles, CA; ⁵CERP, University of California Los Angeles David Geffen School of Medicine, Los Angeles, CA; ⁶Government Operations Agency, State of California, Sacramento, CA. (Control ID #3874862)

BACKGROUND: Get Out the Vaccine (GOTVax), an innovative statewide government-funded COVID-19 vaccine canvassing program in California, aimed to reduce inequities in vaccine accessibility and reliable COVID-related information disproportionately affecting minoritized communities. GOTVax contracted with 34 community-based organizations (CBOs) between April 2021 and December 2021 for vaccine outreach, education, and vaccine registration to reduce structural barriers to vaccination. The academic institution served as the program oversight, contractor, and liaison and provided COVID-19 and vaccine educational training and materials for canvassers. GOTVax was modeled on voter registration and Census door-to-door canvassing approaches and served as an economic stimulation program that hired unemployed and underemployed residents. Communities selected for the program were in zip codes disproportionately impacted by COVID-19, in the lowest quartile of the California Healthy Places Index, and unreached by previous State-led initiatives. The GOTVax CBOs were selected as they already served the state-indicated high-risk zip codes for vaccine registration outreach. We sought to understand the perspectives of CBO leaders participating in GOTVax on the barriers and facilitators to their roles in COVID-19 vaccine outreach. METHODS: We conducted a qualitative community-partnered evaluation of GOTVax. Semi-structured, virtual interviews were conducted with CBO leaders from November 2021 to January 2022. Questions explored the pandemic impact on CBO operations, barriers and facilitators to implementation and management of GOTVax operations, and areas for program improvement. Online demographic surveys described CBO characteristics. Transcripts were analyzed using reflexive thematic analysis.

RESULTS: Thirty-one of 34 CBOs participated (91%). Most CBOs employed less than 25 people (56%), operated for more than 15 years (59%), and reported annual budgets of $1 to $9 million (44%). Identified themes encompassed both facilitators and barriers to program participation. Key facilitators included leveraging trust through recognized entities, promoting empathetic, tailored outreach, and flexibility of milestone-based CBO funding contracts for rapid program implementation. Barriers included navigating community sociopolitical, geographic, and cultural factors, managing canvasser’s safety, desiring performance metrics for self-evaluation of outreach success, mitigating canvassing technology challenges, and concerns of program infrastructure initially limiting outreach to registered voters. CBOs problem solved barriers with the academic and government partners.

CONCLUSIONS: Community-academic-government partnerships should support capacity building for continued COVID-19 outreach and future public health emergencies. Future public health programs should leverage partnerships with community-trusted CBOs and provide flexible funding for tailored outreach in local communities.

COMMUNITY-CENTERED OUTREACH LEADS TO RAPID UPTAKE OF MPOX VACCINES AT COMMERCIAL SEX VENUES IN A LARGE METROPOLITAN AREA OF THE MID-ATLANTIC REGION

Joseph Osmundson^5,6; Katherine Murphy²; Julian Watkins¹; Andrew Wallach^4,2; Adam Baran⁸; Chris Hawke⁷; Ashwin Vasan³; Theodore G. Long²

¹New York City Department of Health and Mental Hygiene, New York, NY; ²Office of Ambulatory Care and Population Health, New York City Health and Hospitals Corporation, New York, NY; ³New York City Department of Health and Mental Hygiene, New York, NY; ⁴NYU Langone Health, New York, NY; ⁵Biology, New York University, New York, NY; ⁶The Rockefeller University, New York, NY; ⁷N/A, N/A, New York, NY; ⁸N/A, N/A, New York, NY. (Control ID #3877336)

BACKGROUND: In May, 2022, the MPOX virus began spreading through global queer sexual networks. Many prior cases in Europe were linked to large queer events with sex on site. While a third-generation smallpox vaccine, JYNNEOS, is maintained in the US Strategic National Stockpile and can prevent MPOX infections, significant shortages remained through the Fall, 2022.

METHODS: We present an on-site vaccination strategy at locations where people meet for sex. All commercial events for queer people with sex on site in a large Mid-Atlantic metropolitan area voluntarily

closed from July 15 through September 3 to help prevent the spread of MPOX. Due to national vaccine shortages, in early September, only first doses of the two-dose JYNNEOS vaccine series were offered in our area. Following the reopening of commercial sex venues, mobile vaccination teams were dispatched to these sites to offer JYNNEOS vaccine on site, including second vaccine doess. Large metropolitan area healthcare officials engaged in bi-directional communication with event hosts and community advocates. Data were collected on doses administered, event size and participant demographics by the event hoses and planners.

RESULTS: From September 3rd through November 12th, mobile uunit healthcare workers administered nearly 400 doses of JYNNEOS vaccines at commercial sex venues. Over this same time period, MPOX cases continued to decrease, even as sexual behaviors returned to baseline. Once second JYNNEOS doses were available at commercial sex venues, vaccine update was high, with 40-60% of event participants receiving an initial dose on site. Mobile units further administered over 2000 doses at public events without sex on site, and over 90% of these were first JYNNEOS does. Mobile units at these event helped increase broad uptake among those who hadn't yet received a dose of MPOX vaccine, while mobile units at commercial sex venues ensured completion of the two-does regimen among those participating in sex with multiple partners. In fct, the mobile units administered nearly 75% of second JYNNEOS doses at commercial sex venues

CONCLUSIONS: We demonstrate the power of tailored community engagement, bringing healthcare to places where people already gather, and show that high uptake of helthcare at commercial sex venues can help decrease the spreaf of sexually transmitted infections, both established and emerging. Our results also emphasize that people attending commercial sex venues could become sexual health hubs that prevent STI transmission throughout entire sexual networks by meeting the needs of their most connected nodes to test for and prevent HIV and bacterial STIs.

COMMUNITY-INFORMED MOBILE OUTREACH AS A STRATEGY TO INCREASE HYPERTENSION AWARENESS AND SCREENING

Priya Sarin Gupta^1,2; Karla Chamorro Garcia³; Miriam Deukmejian³; Anne Fox³; Tavinder Phull³; Elsie Taveras³; Allison S. Bryant²

¹Medicine, Massachusetts General Hospital, Newton, MA; ²Mass General Brigham Inc, Boston, MA; ³Community Health, Mass General Brigham Inc, Boston, MA. (Control ID #3874923)

BACKGROUND: Hypertension is a modifiable risk factor directly related to the development of heart disease. Although hypertension is highly prevalent in historically marginalized communities, awareness and screening remain deficient. There is limited information on mobile health programs as an intervention for hypertension screening modality. This program leveraged a clinic on wheels where the van was co-located using community engagement and data on social vulnerability index. The locales serviced with the mobile clinics have a baseline population of ~40% of individuals born outside of the US, Hispanic ethnicity cited as one of the most common ethnicities and the median household income less than the US average. This approach aimed to evaluate the effectiveness of a mobile-based intervention tasked with increasing blood pressure screenings and awareness in a marginalized population.

METHODS: Partnership with community organizations and data determined where the clinics on wheels held sessions. The vans were typically out in the field for four 4-hour sessions a week. The full menu of services included other primary care services such as vaccination provision. Each van was staffed with a physician medical lead, an NP, MA, CHW and an operations manager. Patients who self-identified as either having an established diagnosis of hypertension or interested in a blood pressure screen in a mobile clinic setting on a walk-in basis were screened. Convenience sampling was used, and urban areas identified as historically marginalized were targeted. Patients found to have abnormal blood pressure readings at clinic were deemed eligible to receive an ambulatory automatic blood pressure monitor and follow-up via phone call. These patients received information pamphlets and verbal delivery of information to foment awareness of the condition. Patients served were asked questions about their demographics and offered a standardized Social Determinants of Health screener.

RESULTS: The population analyzed overall included 857 individual patients who sought walk-in blood pressure screenings from May to November 2022. Individuals were mostly female (64%), median age was 65 years (interquartile range 57-64yrs); ethnicity was 49% Not Hispanic, 44% Hispanic, 7% unavailable. Of the populations analyzed, 47% were white, 28% Other (which includes individuals who cited race as Hispanic/Latino population), 13% Black or African American. About 37% of individuals in this group had an abnormal blood pressure (systolic blood pressure > 140).

CONCLUSIONS: A walk-in low threshold model of blood pressure screening allowed us to reach individuals with elevated blood pressure in historically marginalized communities in the greater boston area. In this setting, women are more likely than men to access blood pressure screening. Mobile-based interventions implemented in partnership with community organizations to provide blood pressure screenings have potential to reach historically underserved communities in large proportion.

COMPARISON OF RACIAL AND ETHNIC DISPARITIES IN COVID-19 MORTALITY BETWEEN NATIONAL VA AND US POPULATIONS.

Evan Shannon^2,1; Taona P. Haderlein^2,3; W. N. Steers²; Michelle S. Wong²; Donna L. Washington^2,1

¹Division of General Internal Medicine and Health Services Research, University of California Los Angeles, Los Angeles, CA; ²VA Center for the Study of Healthcare Innovation Implementation and Policy, Los Angeles, CA; ³Veterans Emergency Management Center, North Hills, CA. (Control ID #3869274)

BACKGROUND: In the US, racial and ethnic disparities in COVID-19 outcomes have been attributed to structural racism resulting in inadequate healthcare access, more vulnerable social conditions, and more severe underlying comorbidities in minoritized racial and ethnic groups (e.g., Black and Hispanic) compared with non-Hispanic White individuals. The Veterans Health Administration (VA), the nation’s largest single- payer integrated healthcare system, addresses many social determinants of health to eligible patients. Despite this, VA studies have also identified COVID-19-related racial and ethnic disparities. However, it is unclear if the magnitude of disparities in COVID-19 outcomes differ between VA and general US populations. We aimed to determine the extent to which age-adjusted sex-stratified racial and ethnic variations in COVID-19 mortality are comparable between VA and non-VA US populations.

METHODS: We obtained measures of age, sex, and COVID-19 mortality for COVID-19-positive cohorts from VA (23,153 women, 202,077 men) and CDC US population data (5,783,792 women, 5,207,603 men) from 3/1/20–5/7/21. Our analysis included non-Hispanic Black, Hispanic, and non-Hispanic White (White) participants. We tested associations between race and ethnicity and COVID-19 mortality, between data source (VA vs. US) and COVID-19 mortality, and between sex and COVID-19 mortality using clustered logistic regression that controlled for age and included a race and ethnicity-by-data source-by-sex interaction term to test whether the magnitudes of the racial and ethnic disparities differed between VA/US data sources by sex. The racial and ethnic disparities were expressed as Black-White and Hispanic-White differences in

COVID-19 adjusted mortality proportions (AMP), and the differences between disparity estimates were expressed as VA/US Black/White or Hispanic/White difference-in-differences (DID) in AMPs. RESULTS: In the VA, Black/White mortality disparities were present for men (Black Veteran AMP 1.7 percentage points [pp] higher than for White Veterans, 95%CI 1.4,2.0), but not for women. US Black/White

mortality disparities were present for both men (AMP +3.1pp, 95% CI 3.1,3.2) and women (AMP +2.0pp, 95% CI 1.9,2.0). US Black/White disparities exceeded those in the VA for both men (AMP DID +1.5pp, 95% CI 1.2,1.8) and women (AMP DID +2.0pp, 95% CI 1.7,2.3). Hispanic/White mortality disparities were present for men in the VA (AMP +1.7pp, 95% CI 1.2,2.1) and US (AMP +1.5pp, 95% CI 1.5,1.6), but not for women in either population. VA/US Hispanic/White disparities comparisons did not differ for men or women.

CONCLUSIONS: COVID-19 mortality disparities were greater in the US compared with VA for Black men, but similar for Hispanic men. Disparities were present for Black women in the US, but not in the VA, and were not present for Hispanic women in either setting. VA healthcare access and attention to social determinants of health may partially address racial and ethnic COVID-19 mortality disparities.

CONTEXTUALIZING RISK, PATHWAYS, AND SOLUTIONS FOR THE RELATIONSHIP BETWEEN ADVERSE CHILDHOOD EXPERIENCES (ACES) AND TYPE 2 DIABETES: A QUALITATIVE ANALYSIS OF THE LIVED EXPERIENCE OF AFRICAN AMERICAN ADULTS LIVING IN AN INNER-CITY

Jennifer A. Campbell¹; Leonard E. Egede²

¹General Internal Medicine , Medical College of Wisconsin, Milwaukee, WI; ²Medicine, Medical College of Wisconsin, Milwaukee, WI. (Control ID #3874535)

BACKGROUND: ACEs confer significant risk for diabetes in adulthood. Within categories of ACEs, evidence shows sexual abuse is related to a 39% increased risk of developing diabetes. The objective of this study was to explore perceptions about risk for sexual abuse, potential pathways leading to diabetes, and solutions based on the lived experience.

METHODS: Three focus groups were conducted with 8-10 participants in each group in the city of Milwaukee, WI. A semi-structured interview guide was used. Focus was placed on sexual abuse based on evidence in the literature demonstrating the relationship with diabetes. Participants were asked to share their perspectives on the prevalence and causes of sexual abuse in their own lives and community and how these experiences relate to adult health, specifically diabetes. Each group ended with a discussion on possible solutions. Grounded theory was used for analysis of qualitative findings. Constant comparisons resulted in emergent themes that were indexed and compared with the rest of the data to establish analytical categories. Analytical categories were further refined as additional data was collected until groups of key themes or categories were identified.

RESULTS: Participants totaled 21 across 3 groups, with approximately 8 – 10 participants in each group. All participants were African American. Average age was 49, average number of years living with diabetes was 13 years. Participants included both men and women. Participants experienced one or more occasions of sexual abuse while growing up. The major themes can be classified as A. Cause, B. Pathway, and C. Treatment/Solution (Figure 1).

CONCLUSIONS: Based on the lived experience, 3 major themes were identified: family instability as a risk for sexual abuse, maladaptive coping as the pathway leading to type 2 diabetes, and social structure as a potential solution to prevent abuse and improve outcomes. Findings from this study provide preliminary data on perspectives of patients who have type 2 diabetes who have experienced sexual abuse. Additionally, these findings emphasize coping as an important factor for mitigating risk of developing diabetes among individuals who have experienced sexual abuse. Future research is needed to examine quantitatively, pathways to inform prevention at the primary, secondary, and tertiary levels of care for both the prevention of sexual abuse, mitigating risk for diabetes, and improving overall health.

COVID-19 TESTING EQUITY IN A LARGE METROPOLITAN AREA OF THE MID-ATLANTIC REGION

Dan H. Rosenfeld¹; Sean Brennan²; Shahrzad Divsalar³; Sarah Joseph Kurien⁴; Chris Keeley⁵; Andrew Wallach^6,4; Theodore G. Long⁴

¹Test & Treat Corps, New York City Health and Hospitals Corporation, New York, NY; ²Mailman School of Public Health, Columbia University, New York, NY; ³Test and Trace, New York City Health and Hospitals Corporation, New York, NY; ⁴Office of Ambulatory Care and Population Health, New York City Health and Hospitals Corporation, New York, NY; ⁵Ambulatory Care, New York City Health and Hospitals Corporation, New York, NY; ⁶NYU Langone Health, New York, NY. (Control ID #3877292)

BACKGROUND: COVID-19 has caused over 43,000 deaths in this large metropolitan area in the Mid- Atlantic, with a disproportional impact on certain communities. As part of the COVID-19 response, this area has directly administered over six million COVID-19 tests (not including millions of indirectly administered tests not covered in this analysis), at no cost to the individual, resulting in nearly half a million poositive results. Given that testing throughout the pandemic has tended to be higher in more affluent areas, these tests were targeted to areas with fewer resources. This study aims to evaluate the impact of this program, specifically, to review its ability to provide equitable testing in economically, geographically, and demographically diverse populations. Of note, in addition to the brick and mortar testing sites evaluated here, this program additionally conducted 2.1 million tests through mobile units, in order to further address testing inequity.

METHODS: Testing data was collected from the in-house Microsoft SQL Server Management Studio Clarity database, representing 6,347,533 total tests, and 449,721 positive tests. These tests were conducted at 48 hospital system locations. Per capita testing rates by zip code tabulation area (ZCTA) and COVID-19 positivity rates by ZCTA were used as dependent variables in separate regressions. Media income, median age, percent English speaking, and percent people of color were used as independent demographic variables, in order to analyze testing patterns across a number of intersecting identities. Negative binomial regressions were run in a Juypter Notebook using Python.

RESULTS: Per capital testing was inversely correlated with median income geographically. The overall pseudo r-squared value when comparing hospital system tests by ZCTA against the selected variables was 0.1101. The number of tests significantly increased as median income fell (SE=1.00000155, p-0.001). Positive tests by ZCTA rose at a significant level alongside the percentage of English-only speakers (SE=0.271, p-0.027).

CONCLUSIONS: This program was able to improve equity through the provision of no-cost testing focused in areas that were disproportionately impacted and had fewer resources. By finding higher numbers of positive tests in resource-poor neighborhoods, the major Mid-Atlantic area was able to deploy additional resources such as contact tracing and isolation/quarantine support including free food delivery and free hotel stays early in the course of hte COVID-19 pandemic. Equitable deployment of testing is feasible and should be considered early on in future epidemics or pandemics.

DELAYS IN CARE AMONG PATIENTS WITH MILD COGNITIVE IMPAIRMENT (MCI) AND MILD COGNITIVE IMPAIRMENT AND DEMENTIA (MCID) DURING THE FIRST WAVE OF THE COVID-19 PANDEMIC

Manying Cui¹; John N. Mafi^1,2; Julia C. Arbanas¹; Mei Leng¹; Chi-Hong Tseng¹; Nina Harawa¹; Catherine Sarkisian¹

¹Medicine – GIM & HSR, University of California Los Angeles David Geffen School of Medicine, Los Angeles, CA; ²RAND Corporation, Santa Monica, CA. (Control ID #3876713)

BACKGROUND: The COVID-19 pandemic and associated policies disrupted U.S. ambulatory care. With the growth in the U.S. population with mild cognitive impairment or dementia (MCID), it is important for policymakers to learn about the delays in care Americans with MCID face. We sought to examine delays in care among U.S. Medicare beneficiaries with MCID across Medicare coverage type, ambulatory service type, and racial and ethnic groups during the first wave of the pandemic. We hypothesized that delays in care would be least prevalent among (1) white participants and (2) participants with Medicare Advantage (MA), which allows for greater flexibility in care modalities and more resources for addressing social determinants of health than traditional Medicare (TM).

METHODS: We analyzed survey response data collected during the first 6 months of the pandemic through the Health and Retirement Study, a nationally representative survey of older Americans. To mitigate non-response bias, the participant’s proxy would respond if the participant was unable to do so. We included Medicare beneficiaries aged ≥65 years and used validated cognitive measures to identify participants with MCID. We calculated rates of delays in accessing overall care, medical visits, surgeries, dental care, and prescription drug use, and stratified delays in care across racial and ethnic groups of non-Hispanic Black, Hispanic, and non-Hispanic other participants. We also stratified by MA vs. TM. We accounted for survey clustering and adjusted results by survey weights for national representativeness and response rate.

RESULTS: We identified 1,568 survey respondents, representing an estimated 6,876,434 U.S. Medicare beneficiaries who are 65 and older with MCID; 57.2% were female with the following race/ethnicity: 61.0% white, 20.6% Black, 14.7% Hispanic, and 3.7% other. Approximately 25% of Medicare beneficiaries with MCID reported delays in care during the first wave of the pandemic. There were no statistically significant disparities in delays in overall care (28.7% vs 27.8%, p=0.75), dental care (4.1% vs 4.9%, p=0.52), and surgeries (16.3% vs 13.9%, p=0.31) between racial and ethnic minority and white patients with MCID. There were disparities in prescription drug access between racial and ethnic minority and white patients with MCID (4% vs 2%, p<0.001). We did not find differences in delays in overall care, medical visits, prescriptions, surgeries, or dental visits among patients with MA or TM.

CONCLUSIONS: In this nationally representative survey of older Americans, we found that one-quarter of Medicare beneficiaries with MCID reported delays in care during the first wave of the COVID-19 pandemic. MA coverage was not associated with improved access to care. Racial and ethnic minority patients with MCID reported more delays in prescription drug access than white patients, raising concerns about ensuring equitable access to prescription drugs during the pandemic.

DESIGNING AN IMPLEMENTATION STRATEGY TO INCREASE SOCIAL NEEDS SCREENING IN A RESOURCE LIMITED CLINIC WITH STAKEHOLDER INPUT.

Elena Byhoff¹; Amy LeClair²; Anayelly Medina³; Cara Smith⁴; Tuhin Roy⁵; Mari-Lynn Drainoni^6,7

¹Medicine, University of Massachusetts Chan Medical School, Worcester, MA; ²Department of Medicine, Tufts Medical Center, Boston, MA; ³Tufts University School of Medicine, Boston, MA; ⁴Family Medicine, Duke University Health System, Durham, NC; ⁵Greater Lawrence Family Health Center, Lawrence, MA; ⁶School of Public Health, Boston University, Boston, MA; ⁷Infectious Diseases, Boston Medical Center, Boston, MA. (Control ID #3868719)

BACKGROUND: Collection of social needs screening data in clinical care is increasing exponentially. In practice, broad implementation of data collection is difficult due to contextual factors across clinical practices. This is particularly true for resource-limited settings, such as Federally Qualified Health Centers (FQHCs) that serve predominantly low income populations who often have high prevalence of unmet social needs. We used the Practical Robust Implementation and Sustainability Model (PRISM) to examine internal (patient, staff and provider) and external (clinic and community referral) factors that impact implementation of social needs screening in clinical encounters, and design an implementation strategy to improve screening. METHODS: Between Jan – March 2022, we conducted a rapid ethnographic assessment with opportunistic interviews to understand current patient workflow and real time implementation processes of social needs screening in an FQHC. We then conducted focus groups (n= 5) to understand processes, perspectives, and context for improving social needs screening among clinic staff and patients. A stakeholder panel was convened from members of each group to design an implementation strategy for social needs screening across the clinic.

RESULTS: Our findings suggest three key barriers to implementation. First, the electronic medical record (EMR) design limited accessibility and feasibility of routine social needs screening and documentation within clinic workflow. Second, social needs screening is not prioritized across all clinic staff. Clinic staff were unable to keep up with data collection demands on top of providing health care, thus allowing social needs screening to “fall off” the routine patient intake. Third, clinic staff had limited knowledge of both patient perceptions of social needs screening and opportunities for social services referral to address identified needs. Implementation strategies developed in partnership with clinical and patient stakeholders addressed all three barriers: improved EMR accessibility and reminders, improved workflow for screening with a “no wrong door” approach, and a stakeholder derived peer education campaign to improve knowledge and buy-in for screening.

CONCLUSIONS: Policy change is necessary but not sufficient to facilitate screening for social needs in the clinical context. Resource-limited settings, such as FQHCs, where patients face disproportionate unmet social needs, may benefit from targeted stakeholder engagement to improve screening implementation.

DEVELOPING A PATIENT INFORMED BIAS REPORTING TOOL WITH INPUT FROM VULNERABLE PATIENT POPULATIONS

Delia Shen, Ashley Odai-Afotey, Sonya Davey, Esteban Gershanik. Internal Medicine, Brigham and Women's Hospital, Boston, MA. (Control ID #3874265)

BACKGROUND: Data show that vulnerable populations, including racial/ethnic minorities and those with limited English proficiency (LEP), have increased risk of a hospital safety event. However, when reporting systems are stratified by race/language, there are under-reported safety and bias events in vulnerable patients. Changes are needed to incorporate a health equity framework in reporting processes. It is essential to analyze bias incidents in non-punitive, anonymous methods to address implicit and explicit staff bias. We aim to develop an anonymous, voluntary patient-facing Event Reporting Tool within Internal Medicine to collect bias and safety incidents experienced by patients of color and patients with LEP to inform integration into existing hospital-wide reporting structures.

METHODS: This is a qualitative study where patients of color and LEP were surveyed at an academic medical center to: 1) understand experiences of racial and/or language discordance bias 2) elucidate barriers to report incidents and 3) gather feedback to inform tool development. Survey responses were recorded with Likert scales. Open ended questions were coded for themes. The tool was reviewed with the Quality and Safety department and iterative editing meetings with the Patient and Family Advisory Council, consisting of patient volunteers.

A convenience sample of patients admitted to Internal Medicine inpatient units was interviewed. Inclusion criteria were: self-identified non-White and/or non-English speaking patients age>18.

RESULTS: 27 patients were interviewed, with 2 declining to finish. 44% were female with ages 22-81 and mean age 48.5. 63% identified as Black, 33.3% Latino, 7.4% Asian, 7.4% American Indian, and 3.7% Pacific Islander. 22% cited bias incidents due to perceived race, ethnicity or language. The majority (89%) felt the care team explained in a way they could understand. However, 22% disagreed about always receiving enough details about their care. 15% overheard negative conversations about themselves. 11% did not trust their medical team decision making.

77.8% spoke English and 22.2% spoke Spanish as their primary language. Of those who spoke Spanish, they noted that staff only used interpreters 60% of the time. Patients with chronic conditions, including pancreatitis and sickle cell disease, cited bias incidents and perceived “pain-seeking” behavior due to their race. Patients noted care differences when roommates were White or primary English speakers. Patients expressed mixed feelings about their likelihood to report adverse events, citing lack of trust with individuals or the system. Most preferred phone call, internet or paper survey for reporting methods.

CONCLUSIONS: In this qualitative study, vulnerable patients cited bias when asked specifically about unfair incidents related to race or language. This Event Reporting Tool will be used to restructure the reporting system to better capture experiences of bias and to serve as a model for other healthcare institutions.

DEVELOPMENT OF A CONCEPTUAL FRAMEWORK TO EXPLAIN THE RELATIONSHIP BETWEEN ADVERSE CHILDHOOD EXPERIENCES (ACES) AND CLINICAL OUTCOMES IN ADULTS WITH TYPE 2 DIABETES: A MIXED METHODS STUDY

Jennifer A. Campbell¹; Leonard E. Egede²

¹General Internal Medicine , Medical College of Wisconsin, Milwaukee, WI; ²Medicine, Medical College of Wisconsin, Milwaukee, WI. (Control ID #3874559)

BACKGROUND: Prior work shows ACEs increase the risk for diabetes, with sexual abuse identified as a major driver of risk amongst ACE categories. Limited data exist on understanding the pathways between sexual abuse, diabetes risk, and diabetes outcomes. This study used mixed methods to identify risk constructs and test the relationships with clinical outcomes in adults with diabetes.

METHODS: Qualitative: Three focus groups were conducted with adults with type 2 diabetes. A semi-

structured interview guide was used to explore risk factors for sexual abuse and pathways leading to diabetes outcomes. Grounded theory was used for analysis of qualitative findings and model development. Major themes were constructed into a conceptual framework with risk constructs based on participant perspectives of how sexual abuse occurs and the factors that influence poor outcomes in diabetes (Figure 1). Quantitative: Wave 1 (1995-1996) and Wave 2 (2004-2006) of the Midlife in the US (MIDUS) study was used and analyzed cross-sectionally. The model was used for variable selection. Logistic and linear regression models were run to assess the relationship between risk constructs, sexual abuse, coping, physical health, and A1C. Separate models were run for each hypothesized relationship.

RESULTS: Qualitative: Focus group participants totaled 21 adults with diabetes, both men and women. Average age was 49. Participants experienced one or more occasions of sexual abuse while growing up. Major themes served as the risk constructs. Quantitative: The final sample included 464 adults with diabetes. In the adjusted models, family instability was significantly associated with increased odds of reporting childhood sexual abuse for foster care (OR=2.65; CI=1.07; 6.54) and having an absent father (OR=0.59; OR=0.35-0.99) but not financial hardship. Maladaptive coping was significantly associated with A1C for comfort eating (b=0.19; CI=0.06;0.31) only. For physical and mental health, fair physical health was significantly associated with HbA1c (b=0.64; CI=0.02; 1.26).

CONCLUSIONS: Certain risk constructs increase risk for odds of sexual abuse, while sexual abuse increases odds of maladaptive coping, and maladaptive coping is associated with worse A1C. Future research is needed to test the direct and indirect pathways and to identify targets for future interventions.

DIFFERENCES IN THE CHARACTERISTICS OF PHYSICIAN NETWORKS THAT TREAT WHITE AND BLACK PATIENTS WITH HEART DISEASE

Ioana Popescu^1,2; Cheryl Damberg²; Luke Matthews²

¹Division of General Internal Medicine, UCLA, Los Angeles, CA; ²RAND Corporation, Santa Monica, CA. (Control ID #3873971)

BACKGROUND: Black-White disparities in high quality hospital use for heart disease treatment may be due to Black and White patients seeing different physicians. The objective of this study was to map physician patient-sharing networks for Medicare patients and assess for Black-White differences in patient-sharing between primary care physicians (PCPs) and cardiologists.

METHODS: We used 100% Medicare Inpatient, Outpatient and Carrier claims for 2016-2017 to identify Black and White fee-for-service Medicare patients with heart disease and the PCPs and cardiologists treating them in hospital referral regions (HRRs) with at least 10 physicians sharing 2 or more Black and White patients (N=226). We used physician claims on shared patients to map separate HRR-level physician networks for Black and White patients focused on PCP to cardiologist patient sharing. We calculated two race-specific network measures: degree (number of cardiologists with whom a PCP shares patients) and transitivity (propensity of physicians to form tightly connected groups). Measures were adjusted for Black- White differences in physician patient panel size and calculated for all settings (hospital and office) and office settings only.

RESULTS: Compared to White patients, Black patients had more cardiology visits, overall (6.9 vs 6.6, p<.001) and with unique cardiologists (3.0 vs. 2.6, p<.001), and were treated less often in office settings (31.7% vs. 40.5%). The Black-White mean differential network degree was 23.4 in all settings and 3.6 in office settings (p<.001 for both), meaning that PCPs shared Black patients with more cardiologists than White patients. The mean differential network transitivity was -0.2 for all settings (p<.001) and near zero for office settings (p=.74), meaning that physicians were less connected in the Black vs. White patient-sharing networks in all but office settings.

CONCLUSIONS: Black-White differences in patient-sharing between PCPs and cardiologists likely reflect differences in referrals, care coordination and ultimately quality of care. Both physician network and patient factors may contribute to these patterns. and/or patients (e.g., regaining community trust).

DIRECT AND INDIRECT RELATIONSHIP BETWEEN HISTORIC REDLINING AND NATIONAL PREVALENCE OF DIABETES: A STRUCTURAL EQUATION MODELING ANALYSIS

Leonard E. Egede²; Rebekah J. Walker¹; Sebastian Linde³; Jennifer A. Campbell⁴

¹Medicine, Medical College of Wisconsin, Milwaukee, WI; ²Medicine, Medical College of Wisconsin, Milwaukee, WI; ³Medicine, Medical College of Wisconsin Department of Medicine, Milwaukee, WI; ⁴General Internal Medicine , Medical College of Wisconsin, Milwaukee, WI. (Control ID #3876093)

BACKGROUND: Structural racism has emerged as an important contributor to poor health outcomes. Historic redlining, the previously legal practice of systematically denying credit access and insurance, resulted from Home Owners’ Loan Corporation (HOLC) residential security maps graded on a color coded scale (A (Green)=Best; B (Blue); C (Yellow); D (Red)=Hazardous), hence the term redlining. The goal of this study was to investigate the direct and indirect relationships between historic redlining and prevalence of diabetes in a national sample.

METHODS: We combined census tract level data across multiple sources: a) diabetes prevalence from CDC PLACES 2019 database, b) HOLC grade from Mapping Inequality project, and c) incarceration risk, poverty, housing, education, employment, job environment, economic mobility, and demographics, from Opportunity Insights database. The assignment of redlining to present-day census tract was done based on overlap with historically HOLC graded areas. The final analytic sample consisted of 9,590 US census tracts. Structural equation modeling was then used to investigate direct and indirect relationships between redlining and diabetes prevalence through possible mediators of health care access, community resources, social capital, and social risk. Stata v16 was used for the analysis and analyses were adjusted for population.

RESULTS: Higher prevalence of diabetes was directly associated with more redlining within a census tract (0.28, p<0.001). All indirect paths were significant with the three strongest indicating diabetes prevalence is indirectly associated with redlining via health care access (0.48, p<0.001); community resources measured via education level (-0.44, p<0.001); and social risk measured via income (-0.23, p<0.001). Figure 1 shows standardized coefficients to allow comparison with higher numbers indicating stronger relationships. CONCLUSIONS: Consistent with our conceptual framework, redlining has significant direct and indirect relationships with diabetes prevalence. Indirect relationships include health care access, community resources including employment and education, social capital, and social risk factors of low income and high incarceration.

DISAGGREGATION OF ASIAN-AMERICAN ETHNICITY CATEGORY REVEALS DISPARITIES IN HOSPITAL LENGTH OF STAY AND DIAGNOSES

Yasuko M. Mano¹; Bianca J. Katnik²; Annie Ro²; Andrew Young¹

¹Medicine, University of Southern California Keck School of Medicine, Los Angeles, CA; ²University of California Irvine, Irvine, CA. (Control ID #3869511)

BACKGROUND: Asian-Americans are the fastest-growing minority in the United States and are comprised of different subgroups. Data for Asian subgroups are often aggregated in research, masking disparities in health and hospital outcomes. Previous studies describe disparities in cardiovascular disease and mortality among Asian subgroups. To our knowledge, no studies have evaluated differences among Asian subgroups in hospitalized patients. We aimed to evaluate differences in average length of stay (ALOS) and illness severity among Asian subgroups for inpatients at a public hospital in Los Angeles, California.

METHODS: We conducted a retrospective analysis of all inpatient encounters at LAC+USC Medical Center in 2019. We combined data from Cerner PowerInsight and Vizient Health System. All data were de-identified to conform to HIPAA requirements. The four Asian subgroups with over 100 encounters were included in the analysis: Chinese, Filipino, Korean, and Vietnamese. Multivariable linear regression was used to assess ALOS, and multivariable logistic regression was used to assess illness severity (reference group: Chinese). Insurance type, primary language, living status, and gender were included in our model to control for confounding. Supplementary regression analyses were performed to assess disparities among the groups in relation to diagnoses. The top five Clinical Classification Software (CCS) diagnostic categories among the cohort were selected as outcomes of interest: septicemia, hypertension, alcohol-related disorders, diabetes, and biliary tract disease. P-values less than 0.05 were deemed significant.

RESULTS: Filipinos were noted to have a significantly lower ALOS compared to Chinese (p-value <0.05). While all patients aggregated had an ALOS of 5.93 days, Chinese had a higher ALOS of 7.74 days. Our data show no significant difference in illness severity among Asian subgroups. Koreans and Filipinos had a larger probability of admission for complications from hypertension compared to Chinese (odds ratio of 2.99 and 5.25, respectively; p-value <0.01). The probability of being admitted for complications from biliary tract disease was larger for Filipinos compared to Chinese (odds ratio of 5.85; p-value <0.05). There were no significant differences in probabilities of being admitted for complications from diabetes, septicemia, or alcohol-related disorders among Asian subgroups.

CONCLUSIONS: Our data show significant differences in ALOS and probabilities of being admitted for complications from hypertension and biliary tract disease among Chinese, Filipinos, Koreans, and Vietnamese. Although limited by small sample sizes, our results show how disparities among Asian subgroups are masked by data aggregation. The Asian-American race category should be disaggregated when analyzing health disparities.

DISPARITIES IN RECEIVING TREATMENTS FOR ALCOHOL USE DISORDER AMONG ALL OF US PARTICIPANTS

Phuc H. Le¹; Jacob Rich^1,2; Eden Y. Bernstein³; Sudie E. Back⁴; Hamlet Gasoyan⁵; Thanh C. Bui⁶; Gina Ayers⁷; Michael B. Rothberg⁸

¹Center for Value-based Care Research, Medicine Institute, Cleveland Clinic, Cleveland, OH; ²Case Western Reserve University, Cleveland, OH; ³General Internal Medicine, Massachusetts General Hospital, Boston, MA; ⁴Medical University of South Carolina, Charleston, SC; ⁵Center for Value-Based Care Research, Cleveland Clinic, Cleveland, OH; ⁶College of Medicine, Department of Family and Preventive Medicine, The University of Oklahoma Health Sciences Center, Oklahoma City, OK; ⁷Pharmacy, Cleveland Clinic, Beachwood, OH; ⁸Internal Medicine, Cleveland Clinic, Cleveland, OH. (Control ID #3876051)

BACKGROUND: Alcohol use disorder (AUD) affects approximately 32 million US adults each year. Both medications and behavioral interventions (e.g. psychotherapy) are effective for AUD yet underutilized. Evidence on disparities in treatment utilization is inconsistent and limited. We examined racial/ethnic and socioeconomic disparities in receipt of AUD treatments in US adults.

METHODS: We conducted a retrospective analysis using All of Us Research Program’s nationwide data. We included adults ≥18 years with AUD, defined by ICD codes for alcohol abuse or dependence. Treatment was defined as having any prescription of medications (disulfiram, acamprosate, and naltrexone) or psychotherapy (CPT codes for individual, family or group-based psychotherapy sessions). We compared rates of treatment across racial/ethnic groups, income levels, insurance types, and area deprivation index (ADI) quintiles. In multivariable logistic models, we assessed the association between these factors and receipt of medication or psychotherapy, adjusted for additional sociodemographic characteristics and coexisting substance and other mental health disorders.

RESULTS: We identified 15,315 patients with AUD. Mean age was 55 years; 59% were male. Most patients did not receive any treatment (70%) while 7% received medication only, 18% psychotherapy only, and 5% both medication and psychotherapy. In multivariable models, Black and Hispanic patients, those with Medicare plus Medicaid or Veteran Affairs (VA) plus Medicare, and those living in fourth and fifth ADI quintile areas were less likely to receive medication compared to White patients, those with private insurance and living in the first ADI quintile areas, respectively. In contrast, higher education level and income were associated with larger odds of receiving medication. Black patients (vs. White) were more likely to receive psychotherapy, as were those with VA insurance (vs. private insurance), and higher income. Patients who were homeless, had Medicaid (vs. private insurance), or lived in more economically deprived areas were less likely to have psychotherapy. Finally, patients with other coexisting substance and mental health disorders had larger odds of receiving medication and psychotherapy than those without.

CONCLUSIONS: Disparities in receipt of AUD treatments exist across race/ethnicity and socioeconomic subgroups. Specifically, Blacks and patients with VA insurance were more likely to receive psychotherapy but not medication. Patients living in more economically deprived areas were less likely to receive both medication and psychotherapy. Systematic approaches are required to improve diverse patient access to effective treatments for AUD.

DISPARITIES IN RECOMBINANT ZOSTER VACCINE COVERAGE AMONG ADULTS IN THE UNITED STATES BY RACE AND ETHNICITY AND SOCIAL DETERMINANTS OF HEALTH David Singer¹; Kathryn Evans²; Nikita Stempniewicz¹; Veronique Page³; Samuel Huse²; Ariel Berger²

¹US Health Outcomes and Epidemiology, GSK, Philadelphia, PA; ²Evidera, Bethesda, MD; ³Evidera, Montreal, QC, Canada. (Control ID #3855623)

BACKGROUND: Recombinant zoster vaccine (RZV) has been recommended for prevention of herpes zoster (HZ) in adults aged ≥50 years in the United States (US) since 2018. Despite this, RZV coverage remains suboptimal and previous studies have found disparities in HZ vaccine coverage by race/ethnicity. This study aimed to describe disparities in RZV coverage by race/ethnicity and by social determinants of health (SDOH).

METHODS: The study used data from the 2020 National Health Interview Survey (NHIS), a cross-sectional survey of noninstitutionalized civilian households in the US conducted by the National Center for Health Statistics at the Centers for Disease Control and Prevention. NHIS uses a geographically clustered sampling design and includes survey weights for projecting findings to the noninstitutionalized US population. The survey includes questions on respondent clinical characteristics, SDOHs, race/ethnicity (self-reported) and receipt of vaccines, e.g., RZV (at least one dose). We report RZV coverage for all adults in the US aged ≥50 years, as well as by SDOHs and race/ethnicity. Survey weights were applied to project findings from each analysis to the US population. To assess for differences in vaccine coverage across groups, Rao-Scott Chi- square tests were used. Two-sided CIs were calculated using the Clopper-Pearson method adapted for complex surveys by Korn and Graubard.

RESULTS: Of the 31,568 individuals surveyed in the 2020 NHIS, 18,384 were aged ≥50 years. RZV coverage in the US population aged ≥50 years was estimated to be 14.1% (95% confidence interval [CI]: 13.4-14.8%), with an estimated 15,791,839 adults in the US aged ≥50 years having received at least one dose as of the time of the survey. Significant differences were observed by race/ethnicity, an estimated 16.6% (95%CI: 15.8-17.5%), 13.7% (95%CI: 10.1-18.0%), 6.9% (95%CI: 5.4-8.6%), and 6.3% (95%CI: 5.1-7.7%) of non-Hispanic white, non-Hispanic Asian, non-Hispanic Black, and Hispanic adults were covered, respectively (Rao-Scott Chi-square p<0.0001). Among adults reporting household income ≥$100,000, RZV coverage was 18.8% (95%CI: 17.4-20.3%) versus 12.8% (95%CI: 11.4-14.4%) and 8.8% (95%CI: 7.9-9.7%) among those with self-reported household incomes of $35,000-$49,999 and <$35,000, respectively. Adults with high food security had RZV coverage of 15.2% (95%CI: 14.4-16.1%) while those with marginal, low, or very low food security had lower coverage (range: 5.2-7.3%).

CONCLUSIONS: RZV coverage in the overall US adult population aged ≥50 years is suboptimal and significant disparities exist in coverage by social determinants of health and by race/ethnicity. These findings demonstrate the need to improve vaccination against HZ in older adults and show that it will be important to ensure this recommended preventive care service is provided in a more equitable manner.

DISPARITIES IN VIDEO CARE USE AMONG VETERANS WITH CARDIOVASCULAR DISEASE Rebecca L. Tisdale^1,3; Claudia Der-Martirosian²; Caroline Yoo⁴; Karen Chu²; Donna Zulman^5,1; Lucinda B. Leung^6,2

¹Center for Innovation to Implementation (Ci2i), VA Palo Alto Health Care System, Menlo Park, CA; ²Center for the Study of Healthcare Innovation, Implementation & Policy (CSHIIP), Veterans Affairs Greater Los Angeles Healthcare System, Los Angeles (GLA), US Department of Veterans Affairs, Washington, DC; ³Department of Health Policy, Stanford University School of Medicine, Stanford, CA; ⁴Center for the Study of Healthcare Innovation, Implementation, and Policy, VA Greater Los Angeles Healthcare System, Los Angeles, CA; ⁵Medicine, Stanford University, Stanford, CA; ⁶Division of General Internal Medicine, UCLA David Geffen School of Medicine & West LA VA Medical Center, Los Angeles, CA. (Control ID #3860899)

BACKGROUND: Video care expanded rapidly in the Veterans Health Administration (VA) at the onset of the COVID-19 pandemic and remains a significant proportion of all VA care. Prevalence of cardiovascular disease (CVD) in Veterans is high, and ensuring access to care for CVD will likely require continued virtual care use. However, there remains a lack of evidence regarding which patients with these CVD conditions are more likely to receive video care. We sought to characterize use of video care for Veterans with two common cardiovascular diseases, heart failure and hypertension.

METHODS: This retrospective cohort study included Veterans established in VA primary care with diagnoses of heart failure and/or hypertension between 3/11/2019 and 3/10/2022, i.e., in the calendar year prior to the novel Coronavirus (COVID-19) pandemic and for the first two pandemic years. We identified individual-level predictors of one or more video-based visits, accounting for patient- and site-level clustering with a two-level mixed-effects logistic regression model adjusted for sociodemographic and clinical covariates and time.

RESULTS: Our analytic cohort comprised 3,807,820 Veterans with diagnoses of heart failure, hypertension, or both with 52 million visits. 456,901Veterans had both heart failure and hypertension, 50,753 had heart failure only, and 3,300,166 had hypertension only. Veterans with heart failure and hypertension had an average baseline age of 71.6 years and mean Charlson Comorbidity Index of 3.0. 2.9% were female at birth, and 34.8% lived in a rural or highly rural setting.

In our multi-level logistic regression model, odds of using video care were highest during the first year of the pandemic, then declined in the second (AOR 15.3, 95% CI 15.1-15.4 and 11.5, 95% CI 11.3-11.6, respectively, compared to the pre-pandemic year). Male patients had lower odds of ever using video care than female patients (adjusted odds ratio [AOR] 0.73, 95% confidence interval [CI] 0.72-0.74). Age showed a gradient: patients 75 years or older had an AOR of 0.38 compared to those aged 18-44 years (95% CI 0.38-0.39). Rural-dwelling Veterans had lower odds of using video care than urban-dwellers (AOR 0.71, 95% CI 0.70-0.71). Veterans with heart failure had slightly higher odds of video care use than those with hypertension only (AOR 1.06, 95% CI 1.05-1.07).

CONCLUSIONS: Veterans with CVD had a 15-fold increase in odds of use of video-based care in the first year of the COVID-19 pandemic. Male, older, and rural-dwelling Veterans had lower odds of using video care than their respective reference groups. Video care remains an important proportion of care delivered to Veterans with cardiovascular disease. Given lower odds of video care among certain veteran groups, continued expansion of video care could make CVD services increasingly inequitable. As VA expands virtual care for CVD, these insights can inform equitable and effective triage of patients to virtual versus in-person care.

DISPARITY IN AGE OF PRESENTATION OF ALCOHOL-RELATED CIRRHOSIS AND ALCOHOL-RELATED HEPATITIS DUE TO THE COVID-19 PANDEMIC IN A TERTIARY U.S. CENTER

Corinne P. Camp¹; Elizabeth Esselman¹; Angela Keniston²; Lauren McBeth³; Susan Calcaterra⁴; Reem M. Hanna⁵

¹Internal Medicine, University of Colorado, Denver, CO; ²Division of Hospital Medicine, University of Colorado, Aurora, CO; ³Medicine/Hospital Medicine, University of Colorado, Aurora, CO;⁴Medicine, University of Colorado - Anschutz Medical Campus, Aurora, CO; ⁵Department of Medicine, University of Colorado, Denver, CO. (Control ID #3876794)

BACKGROUND: The average age of patients hospitalized for chronic liver disease (CLD) in the US is 57 years old. However, data suggests that alcohol-related cirrhosis (ARC) presents at a younger age compared to other causes of CLD, with one study demonstrating that 11% of Caucasian and 16% of Hispanic patients diagnosed with ARC were younger than 40 years old. Additionally, data has shown increased hospitalizations for alcohol related liver disease since the start of the COVID-19 pandemic with a possible cause being increased alcohol consumption in the US since 2019. Our study aimed to evaluate if the COVID-19 pandemic contributed to an earlier age of presentation of ARC and/or alcohol-related hepatitis (ARH) in hospitalized patients, specifically among Hispanic and Black populations compared to non-Hispanic white populations.

METHODS: This was a retrospective observational analysis of all hospitalized patients, aged 18 years and older, with an ICD 9/10 code for ARC or ARH to a university hospital in Colorado from January 2018 to December 2019 and January 2021 to April 2022. We limited encounters to the patient’s first hospitalization (index encounter) during each time period and evaluated for temporal trends across time periods using chi- square tests.

RESULTS: A total of 2,366 patients were included. An unadjusted chi-square test was used to evaluate whether the proportion of patients younger than 40 years who were hospitalized with ARC and/or ARH increased relative to those >40 years after the start of the COVID pandemic. We observed a significant increase in the proportion of patients younger than 40 in the time periods after the start of COVID (18.2%) compared to before (12.5%, p= 0.0001). While the comparison limited to Hispanic patients was trending towards significant, there was insufficient data to corroborate this.

CONCLUSIONS: Our preliminary data shows there was a significant increase in the proportion of younger patients hospitalized with ARC and ARH between the two study time periods. Our study includes hospitalizations that occurred before and during the COVID-19 pandemic. Future work should explore temporal trends of ARC and AHR related hospitalizations by race and ethnicity to inform culturally appropriate interventions to address risks related to alcohol use among various populations.

DISTANCE TO CARE BARRIERS FOR FORMERLY HOMELESS INDIVIDUALS IN A LARGE NATIONAL SURVEY

Stefan Kertesz^2,8; Aerin deRussy¹; Sally K. Holmes²; Ann E. Montgomery^2,3; Kevin Riggs^2,8; April Hoge¹; Audrey L. Jones^4,9; Lillian Gelberg^5,6; Joshua S. Richman^2,8; Erika L. Austin⁷; Adam J. Gordon^4,9.

¹Research , Birmingham VA Medical Center, Birmingham, AL; ²Birmingham VAMC, U.S. Department of Veterans Affairs, Birmingham, AL; ³Health Behavior, The University of Alabama at Birmingham School of Public Health, Birmingham, AL; ⁴VA Salt Lake City Health Care System, Salt Lake City, UT; ⁵Family Medicine, University of California Los Angeles, Los Angeles, CA; ⁶VA of Greater Los Angeles, Los Angeles, CA; ⁷Biostatistics, The University of Alabama at Birmingham School of Public Health, Birmingham, AL; ⁸The University of Alabama at Birmingham Heersink School of Medicine, Birmingham, AL; ⁹University of Utah Health, Salt Lake City, UT. (Control ID #3873911)

BACKGROUND: The impact of distance on access to care for formerly homeless (FH) individuals –who are often accommodated in areas remote from city centers– has not been studied. Using a national survey of FH Veterans, we hypothesized that (a) actual distance would be associated with reported barriers to care, and (b) reported barriers would predict lower primary care (PC) utilization.

METHODS: In 2018, we surveyed 4654 FH Veterans who got PC at 26 VA Medical Centers (VAMCs). Dependent variables: Reported distance barrier, based on “does the distance you…travel make it difficult to get care at [VAMC]?” Distance was “as the crow flies” from mailing address to VAMC. PC utilization was 12-month visit count. Covariates:

Predisposing: gender, race/ethnicity, age

Enabling/impeding: difficulty paying for needs, employment, social support, recent criminal justice experience

Need: emotional distress, chronic pain, self-reported health, drug or alcohol problem, chronic homelessness, 24 Elixhauser comorbidities

Analysis: We used multivariable logistic regression to model reported distance barrier in relation to geographic distance and all covariates, with a site random effect. A generalized linear mixed regression modeled PC visit counts in relation to a reported distance barrier, the listed covariates (sans geographic distance) and Elixhauser comorbidities.

RESULTS: Among FH respondents, 40% resided <5 miles, 29% >5 to <10 miles, 19% >10 to <20 miles and 12% >20 miles from their VAMC. Reporting a distance barrier (28% overall) rose with distance (14%, 27%, 39% & 57% across 4 categories, Mantel-Haenszel Χ² p<0.001). Multivariable-adjusted predicted probabilities rose similarly(Fig). Most covariates were significant, with odds ratios of 1-2. In multivariable regression, respondents reporting distance barriers had fewer PC visits (adjusted mean 2.9, 95% CI 2.7-3.2) compared to those not reporting (mean 3.5, 95% CI 3.4-3.7), adjusted for all covariates.

CONCLUSIONS: Reporting a distance barrier to care is common among FH Veterans, and is associated with fewer PC visits, even for those 5 or more miles from clinic, which applied to 60% of respondents. Health systems serving formerly homeless individuals should address these challenges, with transportation (e.g. rideshares), mobile services or telehealth.

DIVERSITY, BIAS, AND DISCRIMINATION IN PRIMARY CARE: FACULTY SELF- ASSESSMENT OF ATTITUDES, OPINIONS, AND EXPERIENCES IN AN ACADEMIC HEALTH SYSTEM

Erica Chan²; Regina Wang³; Megha Shankar¹; Nancy Binkin⁴; Akbar Rahman³

¹Medicine, University of California San Diego, La Jolla, CA; ²School of Medicine, University of Colorado Anschutz Medical Campus School of Medicine, Aurora, CO; ³Family Medicine, University of California San Diego, La Jolla, CA;⁴Public Health, University of California San Diego, La Jolla, CA. (Control ID #3876215)

BACKGROUND: Academic medical centers are looking to advance diversity, equity, and inclusion (DEI) efforts through major structural changes. Efforts include promoting faculty and student diversity, monitoring patient outcomes, and integration of DEI principles into clinical training. While increasing diversity of trainees is crucial, downstream efforts to improve diversity, equity, and inclusion in clinical practice are more challenging. Understanding faculty members’ experiences, opinions, and perceptions around DEI issues and discrimination, is essential to dismantling structural racism, and diversifying the physician workforce. This pilot study of primary care clinical faculty was undertaken to fill this research gap, and guide internal efforts.

METHODS: From 2021-2022, we distributed an online survey to faculty in the Departments of Internal Medicine (IM) and Family Medicine (FM) at an academic medical center in the California-Hawaii Region (N=132, response rate 55%). The survey used a 5-item Likert scale to assess personal observations and actions related to DEI, and perceptions of colleagues’ DEI knowledge and practices; it also collected data on demographic characteristics, personal and family background information, and personal experiences of discrimination. Likert scale items were dichotomized, and then summed to create a composite score. Items were then compared between demographic groups and those who did and did not experience discrimination, using t-test and chi-squared testing.

RESULTS: Our sample included 72 respondents (30 IM, 42 FM), of whom 67% were female, 56% were White, and 31% were Asian/Pacific Islander. A quarter (25%) were born outside of the U.S; mean year of medical school graduation was 2000 (SD=11). Most reported past experiences of discrimination (83%). A total of 83% had witnessed patients experiencing bias or stereotyping and 56% had witnessed such behaviors toward colleagues. Compared to faculty of color, white faculty were significantly more likely to perceive their colleagues favorably around issues of bias, diversity, and equity (p=0.007). Those who experienced discrimination were more likely to witness instances of patients experiencing bias or stereotyping (p=0.043), and to intervene in such instances (p=0.044), than those who had not experienced discrimination.

CONCLUSIONS: Our results show that personal experiences of discrimination, as well as witnessed episodes of bias and stereotyping against patients and colleagues, are common among clinical faculty. They also provide quantitative evidence of the difference in perception of DEI climate, between white faculty and faculty of color. These results also suggest the possibility of creating a virtuous cycle, whereby recruitment of faculty from marginalized backgrounds can improve the experience of patients who currently experience bias and stereotyping, since these faculty are more likely to intervene and potentially change institutional culture.

DOES DISABILITY IDENTITY INFLUENCE PATIENTS’ PERCEPTIONS OF TRUST, RESPECT, AND FAIRNESS?

Maggie R. Salinger¹; Brian Feltz²; Stephanie H. Chan^4,3; Anna Gosline^4,3; Carine Davila⁸; Suzanne Mitchelle⁵; Lisa I. Iezzoni^6,7

¹General Internal Medicine, Massachusetts General Hospital, Boston, MA; ²3D Research Partners LLC, Boston, MA; ³Blue Cross Blue Shield of Massachusetts, Boston, MA;⁴Massachusetts Coalition for Serious Illness Care, Boston, MA; ⁵University of Massachusetts Chan Medical School, Worcester, MA; ⁶Health Policy Research Center, Massachusetts General Hospital, Boston, MA; ⁷Harvard Medical School, Boston, MA; ⁸Internal Medicine, Massachusetts General Hospital, Boston, MA. (Control ID #3858850)

BACKGROUND: People with functional impairments do not necessarily self-identify as disabled. However, most studies of disabled populations use impaired functioning alone to specify these individuals. With this functional status-based definition, studies have shown disparate care quality for people with disability. We examined whether disability identity itself was associated with differing perceptions of healthcare encounters.

METHODS: We analyzed findings from a nationally representative survey of 1854 US adults conducted in Spring 2021. The survey asked seven impairment questions [six on functional status, adopted from the American Community Survey (ACS-6) and one on health decline]; it also asked, “do you have a disability?” We grouped the 816 participants with impairments by disability identity: Yes-DI=“yes” to disability; No-DI=“no” to disability but “yes” to ≥ 1 impairment question. We also dichotomized Likert scale measures of trust, communication, respect, and fairness in healthcare encounters (henceforth “procedural justice”). Across groups, we compared sociodemographic indicators and rates of dichotomized procedural justice responses (X², ANOVA). We performed multivariable logistic regressions adjusting for baseline characteristics to estimate effects of disability identity (Yes-DI vs. No-DI) on perceptions of procedural justice. All analyses applied survey weights to reflect the US Census.

RESULTS: In Yes-DI (disability identity, n= 340), 39% (n=131) indicated no impairment in response to the ACS-6 questions and thus would have been missed by this measure of function alone. In No-DI (n=476), all participants reported impairments but no disability identity. Most functional impairment subtypes were more prevalent in Yes-DI than in No-DI; Yes-DI had more total impairments [Yes-DI vs. No-DI mean(SD): 1.7(1.3) vs. 1.4(0.8), p<0.001] whereas No-DI had a higher prevalence of reported health decline [No-DI: 329/476 (69%) vs. Yes-DI:178/340 (52%)].

Yes-DI was also older and had higher proportions of participants who were Black, single, not-working, publicly-insured, in low income brackets, and/or who had less than a college degree. Compared to No-DI, Yes-DI rated clinician effort and understanding of health goals more favorably but had worse ratings of fairness and respectful communication. In multivariable models, disability identity was independently associated with higher odds of reporting unfair treatment but also of being unafraid to disagree or ask questions.

CONCLUSIONS: Disability is an integral dimension of diversity and identity that is distinct from impairment status. We showed that perceptions of fairness, communication, and respect differ between people with impairments who do vs. do not self-identify as having disability. In addition to functional status, health systems should capture disability identity to better assess and mitigate disparities in care quality and outcomes for people with impairments.

DO SOCIAL NEEDS INTERVENTIONS ADVANCE RACIAL HEALTH EQUITY? FINDINGS FROM A RAPID REVIEW OF SCOPING REVIEW

Crystal W. Cene¹; Meera Viswanathan⁶; Caroline Fichtenberg²; Sara M. Kennedy⁶; Nila A. Sathe⁶; Laura M. Gottlieb³; Yuri N. Cartier⁴; Monica E. Peek⁵

¹Medicine, University of California San Diego, La Jolla, CA;²Center for Health and Community, UCSF, San Francisco, CA; ³Family and Community Medicine, UCSF, San Francisco, CA; ⁴Social Interventions Research and Evaluation Network, University of California, San Francisco, San Francisco, CA; ⁵Medicine, The University of Chicagoa, Chicago, IL; ⁶RTI International, Research Triangle Park, NC. (Control ID #3874636)

BACKGROUND: Social needs interventions aim to improve health outcomes and mitigate inequities by addressing health-related social needs, such as lack of transportation or food insecurity. However, it is unclear whether these interventions are reducing racial or ethnic inequities. We sought to examine whether studies of social needs interventions explain how race and ethnicity were conceptualized and used in analyses of intervention effectiveness.

METHODS: We conducted a rapid review of studies in multiracial or multiethnic populations included in the Patient-Centered Outcomes Research Institute’s (PCORI’s) recent scoping review and evidence map of social needs interventions in health care settings to explore how these studies conceptualized race or ethnicity and analyzed differential intervention effects outcomes by race or ethnicity.

We developed a framework to assess whether studies were “conceptually thoughtful” for understanding root causes of racial health inequities (i.e., noted that race or ethnicity are markers of exposure to racism) and whether analyses were “analytically informative” for advancing racial health equity research (i.e., examined differential intervention impacts by race or ethnicity).

RESULTS: Of the 152 studies conducted in multiracial or multiethnic populations, 44 studies included race or ethnicity in their analyses, but only 4 of those studies (9%) were conceptually thoughtful about what race or ethnicity means. Among the 152 studies, only 21 (14%) were analytically informative.

CONCLUSIONS: Studies of social needs interventions in multiracial or multiethnic populations were rarely conceptually thoughtful for understanding root causes of racial health inequities and infrequently conducted analytically informative analyses on intervention effectiveness by race or ethnicity. Racism at multiple levels (e.g. interpersonal, structural) impacts minoritized groups’ access to and experience of social needs interventions. Studies of social needs interventions need to be conceptually thoughtful and analytically informative to advance racial equity.

DO VOICE ASSISTANTS PROVIDE EQUALLY APPROPRIATE ADVICE REGARDING COLORECTAL CANCER SCREENING IN DIFFERENT LANGUAGES?

Atul Sinha¹; RAJMOHAN RAMMOHAN²; Tulika Saggar¹; Melvin Joy¹; Paul Mustacchia¹

¹Internal Medicine, Nassau University Medical Center, East Meadow, NY; ²INTERNAL MEDICINE, NASSA UNIVERSITY MEDICAL CENTER, East Meadow, NY. (Control ID #3877022)

BACKGROUND: A voice assistant (VA) is a voice enabled artificial intelligence that allows users to communicate with a device to perform a task or seek information. From the year 2000 to 2019 there was a 56.8% and 10.0% increase in Spanish and French spoken at home in the United States, respectively. With the increase in the percent of non-English languages spoken, we were curious if non-English speakers were receiving appropriate medical advice when asking their VAs questions. We did this study to determine if VAs provide the same quality of advice regarding colorectal cancer in English as compared to in other languages.

METHODS: Four voice assistants: Apple Siri, Amazon Alexa, Google Assistant, and Microsoft Cortana were tested. Voice recordings were done for 5 commonly asked questions regarding colorectal cancer screening in 4 different languages. The authors decided if each of the four VAs provided clinically appropriate advice in each of the languages.

RESULTS: We found that clinically appropriate advice was provided 100% by Apple Siri, 60% by Amazon Alexa, 100% by Google Assistant, and 40% of the time by Microsoft Cortana in English (Image 1). Clinically appropriate advice was provided 80% by Apple Siri, 60% by Amazon Alexa, 80% by Google Assistant, and 40% of the time by Microsoft Cortana in Spanish (Image 1). Clinically appropriate advice was provided 40% by Apple Siri, 0% by Amazon Alexa, 60% by Google Assistant, and 100% of the time by Microsoft Cortana in French (Image 1). Clinically appropriate advice was provided 100% by Apple Siri, 60% by Google Assistant, and 0% of the time by Microsoft Cortana in Mandarin (Image 1).

CONCLUSIONS: Overall, the virtual assistants provided clinically appropriate advice regarding colon cancer in English better than in the other 3 languages. Overall, Apple Siri was the best performing virtual assistant at proving clinically appropriate advice regarding colon cancer. None of the VAs instructed users to speak to a healthcare provider, which we believe is a vital aspect to any medically related search result. Most of the VAs performed well in our study but we believe there is a need for improvement, especially with how technology is becoming more ingrained in our everyday lives how much non-English spoken language has grown in the United States.

DYNAMICS OF THE DIGITAL DIVIDE: TELEMEDICINE USE AMONG MEDICAID BENEFICIARIES IN WASHINGTON

Anna M. Morenz^1,4; Lingmei Zhou^2,4; Edwin S. Wong^3,4; Ashok Reddy⁵; Joy S. Lee^1,4; Judy Zerzan- Thul^1,6; Chistopher P. Chen^1,6; Joshua M. Liao^1,4

¹Department of Medicine, University of Washington, Seattle, WA; ²General Internal Medicine , University of Washington, Seattle, WA; ³University of Washington Department of Health Services and Population Health, Seattle, WA; ⁴Value and Systems Science Lab, University of Washington, Seattle, WA; ⁵Medicine, University of Washington System, Seattle, WA; ⁶Washington State Health Care Authority, Olympia, WA. (Control ID #3872559)

BACKGROUND: COVID-19 necessitated expansion of telemedicine as alternatives to in-person care. However, this expansion has been accompanied by concerns about a “digital divide” – a chasm in access and use among older, non-English-speaking, and minoritized patients. Insight about patterns and equity of telemedicine utilization can inform coverage and payment policies that promote equity.

METHODS: In Washington State (WA), Medicaid changed telemedicine policy in response to COVID-19.

We conducted a retrospective analysis of telemedicine use between March and December 2020 among adults in WA enrolled in Medicaid managed care for at least 6 months in 2019 – 2020. Outcomes were (a) any telemedicine use and (b) intensity of use based on per-person counts of telemedicine visits. Telemedicine included audio-only (phone) and audio-video for new or return visits.

Multivariable generalized linear models were used to assess the association between beneficiary characteristics (age, gender, race, ethnicity, primary language, federal poverty level, clinical complexity) and outcomes. We also assessed the association between residence in a high Social Vulnerability Index (SVI) area (defined by overall and four SVI sub-categories) and outcomes.

RESULTS: Among 355,443 Medicaid beneficiaries, 57,342 used audio-only (16.1%) and 62,804 used audio- video (17.7%) telemedicine. Mean age was 37.8 years, with 57.5% female, 8.5% Black, and 16.4% Hispanic. Overall, 7.1% were homeless, 92.8% identified English as their primary spoken language, and 70.1% resided in areas of high social vulnerability.

Individuals of Black race (aOR 1.11, 1.07-1.14), with primary spoken language of English (1.10, 1.04-1.18), or residing in area with greater household type and transportation vulnerability (1.14, 1.12-1.17) were more likely to have any use of telemedicine. Individuals experiencing homelessness (0.90, 0.87-0.93), of Hispanic ethnicity (0.97, 0.95-0.99), or residing in areas with greater household composition and disability vulnerability (0.977, 0.959-0.995) were less likely to use any telemedicine.

Older age, English as primary spoken language, or residence in areas with greater household type and transportation vulnerability were associated with greater intensity of telemedicine use. Lower intensity of use was associated with Black race (aIRR 0.90, 0.89-0.91) or Hispanic ethnicity.

CONCLUSIONS: Amid COVID-19, there were varied patterns of telemedicine use among different groups of Medicaid beneficiaries in WA. Some patterns – such as greater use of any telemedicine coupled with lower intensity use, as observed among Black individuals – may suggest limited acceptability of telemedicine, or barriers to repeated use, rather than outright access barriers. To bridge the digital divide and promote equity through telemedicine, future policy may need to account for these dynamics and emphasize not just access, but also acceptability of and engagement with telemedicine among historically marginalized populations.

EARLY IMPACT OF A BIDIRECTIONAL CLINIC TO COMMUNITY ELECTRONIC REFERRAL SYSTEM ON ADDRESSING HEALTH-RELATED SOCIAL NEEDS AND HEALTH OUTCOMES Shari Bolen^1,2; Jonathan Lever²; Chris Mundorf²; Alvonta Jenkins²; Samantha Smith³; Matthew Finley⁴; Joseph Daprano⁵; Eva Johnson⁶; Marie Masotya⁶; Anandhi Gunder⁶; Melissa E. Lohr⁷; David Bar-Shain⁷; David Kaelber⁸; Tatyana Khaled³; Dieter W. Sumerauer⁶

¹Medicine, MetroHealth/Case Western Reserve University, Cleveland, OH; ²Better Health Partnership, Cleveland, OH; ³Epidemiology, Surveillance, & Informatics, Cuyahoga County Board of Health, Parma, OH; ⁴United Way of Greater Cleveland, Cleveland, OH; ⁵Internal Medicine/Pediatrics, MetroHealth Medical Center, Willoughby Hills, OH; ⁶Pediatrics, University Hospitals Rainbow Babies & Children's Hospital, Cleveland, OH; ⁷Pediatrics, The MetroHealth System, Cleveland, OH; ⁸Internal Medicine, Pediatrics, and Informatics, The MetroHealth System/CWRU, Cleveland, OH. (Control ID #3874734)

BACKGROUND: Addressing social needs of safety-net patients could have a large impact on health, yet effective models to address these needs are limited. We describe the impacts of an electronic health record (EHR) facilitated, clinic-to-community linkage program addressing social drivers of health (SDoH) on health outcomes.

METHODS: We report on an ongoing prospective longitudinal cohort study with enrollment of 5 safety net clinics at 3 health systems between 2018-2022. Patients were eligible if they were: 1) adults with hypertension and uncontrolled blood pressure (BP≥140/90 mmHg) at 1 clinic; or 2) children 2-17 years old with overweight/obesity (i.e., body mass index (BMI) percentile ≥85^th) or with asthma at the 4 remaining clinics. Primary care teams referred consenting eligible patients (or whose parent/legal guardian consented) to United Way 2-1-1 (UW 211) via a point-of-care EHR referral. UW 211, a non-profit organization covering 2.5 million Ohioans, called the patients to assess and refer them to any of over 20,000 community resources from their database. UW 211 provided feedback to clinics (e.g., resources provided) via standardized electronic health information exchange in the EHR. First, we describe SDoH referrals. Next, using a case- control study with 4:1 matching on age, race/ethnicity, neighborhood income, sex, insurance type and comorbidity, we compare selected health outcomes (BMI percentile and asthma exacerbations for kids, and BP for adults) in the 1-year post vs. the 1-year pre-referral between the intervention groups (IG) and control groups (CG).

RESULTS: From the 5 clinics, 1,192 eligible patients were referred (referral ranges by clinic: 6% to 40%). Of the 1,192 patients, 450 (38%) were reached by a UW 211 navigation specialist. All 450 had at least one need identified, and 88% (n=395) had at least one need resolved or a resolution in progress. Reached patients reported an average of 2.9 (SD 1.3) SDoH needs per patient, and were referred to an average of 10.1 community resources (SD 6.1). Patients had the following top 5 needs: food (75%); recreation (68%); health care (32%); housing (32%); and utility assistance (30%). In the matched case-control study, no statistically significant differences were identified in BMI, asthma exacerbation, or BP health outcomes between groups. In children, the BMI percentile increased by 3 kg/m2 in the IG (n=243) and 4 kg/m2 in the CG (n=1,000), and asthma exacerbations did not change in the IG (n=6) and increased by 1% in the CG (n=17). In adults, BP control (<140/90 mmHg) increased 7% in the IG (n=210) and 6% in the CG (n=900)

CONCLUSIONS: A seamless EHR-facilitated closed-loop community resource referral and feedback process is feasible in busy primary care practices to address patients’ SDoH. Early changes in health outcomes were not identified. However, impacts on health outcomes in the outpatient setting from addressing SDoH will take years and should not be the sole criteria used to determine continued program funding.

EFFECT OF PATIENT-PHYSICIAN RACIAL CONCORDANCE ON OUTCOMES OF PATIENTS TREATED BY HOSPITALISTS

Evan Shannon, Mariah B. Blegen, Ruixin LI, Yusuke Tsugawa

Division of General Internal Medicine and Health Services Research, University of California Los Angeles, Los Angeles, CA. (Control ID #3877324)

BACKGROUND: Literature suggests that patient-physician racial concordance may improve care quality and patient satisfaction for Black patients. This could be due to shared cultural values, improved communication and attenuated implicit bias between concordant pairs. However, evidence that such concordance contributes to improved patient outcomes is mixed. The aim of this study is to determine if patient-physician racial concordance is associated with outcomes for patients treated by hospitalists

METHODS: We linked 2016-2019 data on Traditional Medicare beneficiaries from the 100% Medicare inpatient file with Medical Board of California data, which contains physician self-reported race and ethnicity for 24.7% of hospitalists. Physicians were considered hospitalists if their specialty was general medicine, >=90% of their E&M claims were in the inpatient setting and at least 20 claims were filed during the study period. We included all medical primary admission diagnoses and restricted the sample to Black and White patients aged 65-99 years. We then coded patient-physician dyads as racially concordant if patients and their physicians shared the same race. For the primary analysis, we calculated if racial concordance was associated with improved 30-day mortality and 30-day readmission compared to non-concordance using generalized linear models adjusted for patient age, gender, zip code median income, Medicaid status, comorbidities, month/year of admission, weekend admission and physician age, gender and patient volume, with robust standard error estimates.

RESULTS: Among the 8957 patients (7.5% Black patients) treated by 320 hospitalists (11.6% Black hospitalists) in our sample, 114 (16.9%) Black and 7687 (92.6%) White patients were treated by racially- concordant hospitalists. Although the differences were not statistically significant, after adjustment, we found a trend toward lower mortality when Black patients were treated by racially-concordance hospitalists. The risk-adjusted mortality rates were 6.9% (95%CI 2.4% to 11.3%) for Black patients treated by Black hospitalists, 9.1% (95% CI 6.6% to 11.5%) for Black patients treated by White hospitalists, 8.3% (95%CI 7.7% to 8.9%) for White patients treated by White hospitalists, and 8.2% (95% CI 6.0% to 10.3%) for White patients treated by Black hospitalists (p-for-interaction=0.69).

CONCLUSIONS: In a sample of Medicare beneficiaries treated by hospitalists with California medical licenses, we did not find evidence that patient-physician racial concordance was associated with clinical outcomes for Black patients; however, we found a trend toward lower mortality for Black patients treated by racially-concordant hospitalists. Our study was limited by the relatively small number of patients linked to physicians with self-identified race. Further studies with a larger sample size are warranted to determine whether racial concordance leads to better patient outcomes among patients hospitalized for medical conditions.

EQUITABLE PATIENT ENGAGEMENT IN A REMOTE PATIENT MONITORING PROGRAM FOR HYPERTENSION

Nadia Liyanage-Don^3,4; Harry West¹; Kelsey B. Bryant²; Jessica R. Singer⁴; Luis Blanco³; Adina Fraser⁵; Rakhi Kalra⁵; Nadine Dandan⁶; Ian Kronish^3,4

¹IEME, Columbia University, New York, NY; ²Icahn School of Medicine at Mount Sinai Department of Medicine, New York, NY; ³Center for Behavioral Cardiovascular Health, Columbia University Irving Medical Center, New York, NY; ⁴Department of Medicine, Columbia University Irving Medical Center, New York, NY; ⁵IT, NewYork-Presbyterian Healthcare System Inc, New York, NY; ⁶Pharmacy, NewYork-Presbyterian Healthcare System Inc, New York, NY. (Control ID #3871924)

BACKGROUND: Home blood pressure (BP) monitoring using wireless BP devices that transmit data to the electronic health record, also known as remote patient monitoring (RPM), is recommended by national guidelines as an evidence-based approach to improve hypertension (HTN) control. However, there are disparities in patient access to and engagement with such technology, particularly among the elderly, minorities, or those with low health literacy. Barriers to equitable access and uptake must be considered in the design and implementation of RPM and other telemedicine modalities.

METHODS: We used principles of human-centered design to develop a HTN RPM program and implementation strategy. This involved careful mapping of the user experience based on input from key stakeholders and iterative testing in real-world settings. The final program provided patients with a wireless BP cuff, two-way tablet, and virtual nursing support free of charge. Other features aimed at reducing disparities included bilingual program materials, use of interpreter services, option for in-person device installation, and no technology requirements (e.g. Wifi). We retrospectively reviewed medical records of patients referred to the program to assess sociodemographic differences in program engagement. We quantified the proportion of referred patients who subsequently enrolled in the program and the frequency of home BP readings during the first 6 months of enrollment. Regression analysis was used to identify patient characteristics associated with program enrollment and BP measurement.

RESULTS: From June 2019 to August 2021, 443 patients were referred to the HTN RPM program (mean age 67±13 years, 68% female, 24% Black, 62% Hispanic, 60% non-English). Of these, 74% actively enrolled and took at least one home BP measurement. The mean number of home BP measurements was 19±9 in the first month, declining to 11±8 by the sixth month. There were no differences in program enrollment or number of mean monthly BP readings based on patient age, sex, race, ethnicity, or language (Table)

CONCLUSIONS: There were no significant sociodemographic differences in patient enrollment or engagement with our HTN RPM program over six months. Although the mean number of home BP measurements declined from the first month to the sixth month, this occurred irrespective of age, sex, race, ethnicity, or language. These findings suggest that our approach to program design and implementation was equitable, highlighting the importance of equity considerations when developing telehealth solutions for diverse populations.

EVALUATING RACIAL/ ETHNIC EQUITY IN THE PREVENTIVE HEALTH INVENTORY DELIVERY AND OUTCOMES AMONG VETERANS WITH DIABETES AND HYPERTENSION Leah Marcotte¹; Chelle Wheat²; Mayuree Rao³; Edwin S. Wong⁴; Paul Hebert⁴; Karin Nelson¹; Jorge Rojas Jr.⁵; Eric Gunnink⁶; Ashok Reddy⁷

¹Medicine, University of Washington, Seattle, WA; ²Primary Care Analytics Team, Veterans Health Administration, Seattle, WA; ³Veterans Health Administration, Seattle, WA; ⁴Center of Innovation for Veteran-Centered and Value-Driven Care, VA Puget Sound Health Care System Seattle Division, Seattle, WA; ⁵HSR&D, Veterans Health Administration Operations, Washington, DC; ⁶HSR&D Puget Sound, US Department of Veterans Affairs, Washington, DC; ⁷Medicine, University of Washington System, Seattle, WA. (Control ID #3870749)

BACKGROUND: The Preventive Health Inventory (PHI) is a virtual care intervention implemented in March 2021 across the Veterans Health Administration (VA), a national system of over 900 primary care clinics, to address pandemic-related disruptions in primary care. PHI consists of a national quality metrics dashboard and template for individualized outreach by nurses to Veterans via video or telephone to address preventive- and chronic-care gaps. This study evaluated whether PHI was delivered equitably among racial/ethnic groups and if disparities in diabetes and hypertension outcomes changed after PHI receipt.

METHODS: This is a retrospective study among 1,805,658 Veterans with diabetes and/or hypertension and eligible for PHI from 2/28/2021 to 3/31/2022. Two measures - Hemoglobin A1c >9% and blood pressure

<140/90 mmHg - were used to evaluate diabetes and hypertension outcomes. Race/ethnicity was determined using standard groups and a validated algorithm. We defined disparities as worse group outcomes compared to non-Hispanic white (NHW) Veterans. We used logistic regression to evaluate PHI receipt by racial/ethnic group, compared to NHW Veterans. Next, we used logistic regression to separately estimate disparities in outcome probability at the pre-intervention (2/28/2021) and post-intervention (3/31/2022) time points. We then calculated a difference-in-difference (DID) estimate of the change in disparity over time.

RESULTS: Of eligible Veterans, 5.7% NHW (N=68,744), 5.6% non-Hispanic Black (NHB) (N=22,580), 10.2% Hispanic (N=13,313), 6.2% Asian/Pacific Islander/Native Hawaiian (N=1,868), 5.1% American Indian/Native Alaskan (N=744), and 5.6% multiple races or other race (N=1,647) Veterans received PHI. We found no significant population-level or within clinic disparities in PHI receipt. Prior to PHI, NHB (42.2%) and Hispanic (39.5%) Veterans were less likely to have controlled hypertension vs. NHW Veterans (47.5%; p <0.001 and p = 0.048, respectively); NHB Veterans (32.9%) were more likely to have uncontrolled diabetes vs. NHW Veterans (25.1%; p <0.001). Hypertension and diabetes measures improved in all groups post- intervention. There were no new or worsened disparities after PHI, and in DID analysis, among NHB Veterans, the disparity in uncontrolled diabetes improved by 1.9 percentage points (95% confidence interval 0.2, 3.6).

CONCLUSIONS: Organizations should routinely evaluate equity in novel interventions. Our findings support continued use of the PHI as an equitably deployed intervention. Outcomes in all groups improved but most racial/ethnic disparities did not after PHI implementation; findings highlight additional need for tailored interventions to achieve equity in health outcomes.

EXAMINING WITHIN-GROUP HETEROGENEITY IN COVID-19 ASSOCIATED FOOD INSECURITY AND MENTAL HEALTH IMPACTS AMONG LATINO ADULTS IN THE UNITED STATES

Maissa Trabilsy²; Kasim Ortiz³; Marlene Camacho-Rivera¹

¹Community Health Sciences, SUNY Downstate Health Sciences University, New York City, NY; ²SUNY Downstate Health Sciences University College of Medicine, New York, NY; ³Drexel University Dornsife School of Public Health, Philadelphia, PA. (Control ID #3876709)

BACKGROUND: The COVID-19 pandemic has disproportionately burdened the Latino community, exacerbating the negative impacts on mental health and food insecurity. Latinos are more likely than Whites to experience occupational and residential segregation that puts them at a higher risk for COVID-19 infection and mortality. However, the differential impacts of social and material hardships related to COVID-19 within the Latino community, by ethnicity, generational status, and chronic disease history, remains an unexplored area of research.

METHODS: We used publicly available data from the nationally representative Understanding

America Study, administered from April 1, 2020 through July 20, 2021. Our primary outcome of food insecurity was based on the following three questions: “In the past seven days, were you worried you would run out of food because of a lack of money or other resources?”; “Did you eat less than you thought you should because of a lack of money or other resources?”; and “Did you go without eating for a whole day because of a lack of money or other resources?”. Participants were categorized as reporting food insecurity if they responded “Yes” for any of the three questions. Mental health symptoms were assessed using the PHQ9 and GAD7. Covariates included gender identity, age, immigration status, education, insurance status, and smoking status. We conducted generalized linear models, weighted for the complex survey design, stratified by gender to examine associations between food insecurity, demographic characteristics, and mental health symptoms.

RESULTS: In fully adjusted models, Puerto Rican and Mexican men were more than twice as likely and 1.3 times as likely to report food insecurity compared to non-Hispanic white men (aPR 2.16, 95% CI 1.50-3.12 and aPR 1.30, 95% CI 1.08-1.57). Conversely, men of other Latinx background were 50% less likely to report being food insecure compared to non-Hispanic white men (aPR 0.43, 95% CI 0.28-0.66). Mexican and Puerto Rican women were 1.2 times as likely and 3 times as likely to report being food insecure compared to non-Hispanic White women (aPR 1.21, 95% CI 1.12-1.39 and aPR 3.01, 95% CI 2.42-3.75). Examining mental health symptoms, statistically significant interactions between food insecurity and asthma status were observed where asthmatics that were food insecure were more than 4 times as likely to report mental distress compared to asthmatics without food insecurity (aPR 4.43, 95% CI 3.10-6.33).

CONCLUSIONS: Among Latinos, significant heterogeneity in the burden of food insecurity among Latino subgroups emerged, which was masked when examining Latinos as one overall ethnic group. As COVID-19 disparities have disproportionately impacted both Latinos and individuals living with chronic health conditions such as asthma, health professionals must continue to screen for and intervene on social determinants of health, including food insecurity, to improve engagement in care and chronic disease self-management.

EXPERIENCES OF VIOLENCE AMONG PEOPLE EXPERIENCING HOMELESSNESS (PEH) IN CALIFORNIA

Anita S. Hargrave^1,2; Kim H. Nguyen^3,4; Kara Ponder^4,3; Jennafer Birkmeyer^4,3; Margot Kushel^3,4

¹Internal Medicine, University of California San Francisco, San Francisco, CA; ²Internal Medicine, San Francisco VA Health Care System, San Francisco, CA; ³Department of Medicine, ZSFG, UCSF Center for Vulnerable Populations, San Francisco, CA; ⁴UCSF Benioff Homelessness and Housing Initiative, San Francisco, CA. (Control ID #3872336)

BACKGROUND: Sexual and physical assault are critical determinants of health and healthcare costs. In a representative sample of PEH in CA, we assessed the prevalence of violence and who perpetrated it.

METHODS: Using a sampling frame to recruit a representative sample of all adults experiencing homelessness, we recruited adults (18+yrs) in 8 CA counties from 10/2021-12/2022 with venue-based sampling. The questionnaire inquired about types, duration and perpetrators of violence. We performed bivariate analyses with stratification by gender identity, adjusted for survey weight to generate representative statewide estimates, and using tests of significance (chi-square, anova).

RESULTS: Among 3014 PEH, 71% male (cis), 79% female (cis), and 92% of all other genders reported a lifetime experience of violence. While homeless, a third reported physical assault; 8% males, 18% females and 28% of all other genders reported sexual assault. Other-gendered participants reported greater exposure to childhood, emotional and financial abuse as well as police violence. The most common perpetrators of physical, sexual, and financial abuse differed by participants’ gender.

CONCLUSIONS: Violence is highly prevalent among PEH and may impact risk of becoming homeless or ability to exit homelessness.

EXPLAINING REDUCED BLACK-WHITE DISPARITIES IN VIDEO-BASED PRIMARY CARE VISITS IN THE VETERANS HEALTH ADMINISTRATION: A DECOMPOSITION ANALYSIS OF GEOGRAPHIC ACCESS AND PATIENT FACTORS

Lucinda B. Leung¹; Karen Chu²; Nicholas J. Jackson¹; Maia Carter⁴; Leonie Heyworth³; Donna L. Washington¹; Claudia Der-Martirosian⁵

¹Division of General Internal Medicine, UCLA David Geffen School of Medicine & West LA VA Medical Center, Los Angeles, CA; ²Pharmaceutical Economics and Policy, University of Southern California, Los Angeles, CA; ³General Medicine, VA San Diego Healthcare System, La Jolla, CA; ⁴Office of Primary Care/Patient Care Services, Veterans Affairs, Washington, DC; ⁵Center for the Study of Healthcare Innovation, Implementation & Policy (CSHIIP), Veterans Affairs Greater Los Angeles Healthcare System, Los Angeles (GLA), US Department of Veterans Affairs, Washington, DC. (Control ID #3875582)

BACKGROUND: While telehealth is increasingly used as a tool to reduce health care access barriers, Black patients have been found to have fewer video visits than White patients. Differences in geographic access (e.g., rurality) and patient factors (e.g., age) affect this disparity, but their relative contributions are unknown. The Veterans Health Administration (VA) rapidly expanded telehealth use for patients during the COVID-19 pandemic (e.g., distributed tablets for video visits; provided digital health education). We aimed to examine Black-White differences in video visits for VA primary care services over time and then to quantify the influence of various factors on differences.

METHODS: This retrospective study examined 0.9M self-reported Black and 3.6M Non-Hispanic White patients who received primary care services from 138 VA healthcare systems nationally. We determined if each patient ever had a video-based primary care visit in the year before (3/16/2019-3/15/2020) and year after pandemic onset (3/16/2020-3/15/2021). We fit linear probability regression models, adjusting for geographic access (rurality, VA healthcare system, broadband/high-speed internet available in census block [≥25/3Mbps per FCC]) and for patient factors (age, gender, comorbidity, received VA tablet, census-based socioeconomic data [Area Deprivation Index]). We used decomposition methods to quantify the proportion of Black-White differences explained by differences in study covariates, and that which was left unexplained (e.g., discrimination), before and after pandemic onset.

RESULTS: Initially, 1.8% of Black and 2.6% of White patients ever used video visits, compared to 24.4% of Black and 17.8% of White patients after pandemic onset. Geographic access and patient factors accounted for the majority of Black-White differences in video use (60.9% pre and 82.6% post; p’s<.001). Pre-pandemic, racial differences were almost entirely due to geographic access. Of the 0.8 percentage-point (pp) difference between Black and White patients, rurality and VA healthcare system each made up 0.4pp (p’s<.001), both favoring White patients. Post-pandemic onset, differences were not as much from geographical access. Of the 5.6pp difference, contributions from VA healthcare system made up 1.6pp (p<.001; rural, p=.13), newly favoring Black patients. Among patient factors, age was the greatest contributor to Black-White differences (2.4pp, p<.001), while others (e.g., VA tablet) made up <1pp.

CONCLUSIONS: Black-White differences in ever using video-based primary care visits were initially due to geographic access barriers, but largely reversed to favor Black patients with pandemic-related VA telehealth expansion. Only a minority of Black-White differences were due to unmeasured variables (e.g., discrimination). VA telehealth implementation helped reduce health system and contextual reasons for disparities, but outreach to older Veterans is still needed to achieve digital health equity.

EXPLORING RACIAL AND ETHNIC DISPARITIES AND RISK FACTORS FOR NONALCOHOLIC FATTY LIVER DISEASE (NAFLD)

Janet N. Chu¹; Mindy Hebert-Derouen³; Alison Canchola³; Aly Cortella³; Yihe Daida²; Scarlett L. Gomez³; Pushkar P. Inamdar³; Hashem El-Serag⁴; Salma Shariff-marco³

¹Internal Medicine, University of California San Francisco, San Francisco, CA; ²Center for Integrated Health Care Research, Hawaii Permanente Medical Group, Honolulu, HI; ³Department of Epidemiology & Biostatistics, University of California San Francisco, San Francisco, CA; ⁴Medicine-Gastroenterology, Baylor College of Medicine, Houston, TX. (Control ID #3876998)

BACKGROUND: While 25% of U.S. adults has NAFLD, few studies have examined NAFLD risk factors in racially and ethnically diverse populations. Minoritized racial and ethnic groups are at higher risk for diabetes and obesity, which are associated with NAFLD.

METHODS: This electronic medical record (EMR)-based cohort study included adults (18+ years) at a large, integrated healthcare system in Hawaii during 2000-2017. We extracted sociodemographic data (age, sex, race/ethnicity) as well as prospective data on behaviors (tobacco use and care utilization), International Classification of Diseases (ICD)-9 and -10 codes, prescriptions, vital signs, and lab results. NAFLD was defined using ICD codes (571.8, 571.8, K75.81, K76.0, K75.8). Descriptive statistics were used to describe the cohort and those with NAFLD. Univariable logistic regression with an outcome of NAFLD was used to estimate odds ratios (OR) and 95% confidence intervals (CI).

RESULTS: Of 637,436 patients, 10,590 (1.7%) had a diagnosis of NAFLD. There were significant differences between those diagnosed with NAFLD vs. those who were not, based on sociodemographic characteristics and medical comorbidities. Compared to non-Hispanic (NH) White patients, minoritized racial and ethnic groups had higher odds of NAFLD (Table 1). Patients in specific racial and ethnic groups (including American Indian/Alaska Native, Native Hawaiian, Pacific Islander, Chinese, Japanese, Korean, and Hispanic) had higher frequency of obese BMI, limited English proficiency, ever smoked, hypertension, dyslipidemia, and diabetes, than NH White patients, which may contribute to these disparities

CONCLUSIONS: In this EMR-based cohort, only 1.7% had a NAFLD diagnosis, consistent with prior studies that NAFLD is underdiagnosed in the general population. This is one of the first studies to report disparities in NAFLD diagnosis based on detailed race and ethnicity data. Next steps include determining which risk factors are independently associated with NAFLD across detailed racial and ethnic groups.

EXPLORING THE IMPACT OF SOCIOECONOMIC HEALTH DISPARITIES IN THE US HOSPITALS ON PREGNANT PATIENTS ADMITTED WITH SEPSIS

Andrea J. Zapata¹; Falah I. Abu Hassan³; James ". Walker²

¹school of medicine, Texas Tech University Health Sciences Center School of Medicine, Lubbock, TX; ²Internal Medicine, Texas Tech, Amarillo, TX;³Internal Medicine, Texas Tech University Health Sciences Center, Amarillo, TX. (Control ID #3874319)

BACKGROUND: In pregnant women literature shows a large stratification in the rate of complications and negative outcomes between socioeconomic classes, regardless of resource availability. Similarly, mortality due to sepsis has been attributed to socioeconomic variables, but the impacts of socioeconomic status (SES) on sepsis outcomes in pregnant patients specifically is poorly understood. It is in this population that we sought to explore the impact of health disparities by examining rates of spontaneous abortion (SA) and hospital length of stay (LOS) in the United States.

METHODS: The national inpatient sample (NIS) was searched for pregnant women between the years 2016 to 2019 who were admitted with a sepsis diagnosis using the appropriate ICD-10 codes. The median household income (MHI) of the patients’ ZIP Code of residence was used to place patients into MHI determined quartiles. Disparity between patients residing in areas within the lower 25th percentile of MHI was compared to all other patients with higher MHI. Additionally, self-pay patients' outcomes were assessed against patients with insurance. Univariate logistic regression was performed to compare the outcomes of the two groups in terms of the occurrence of SA and LOS. The results were reported as an odds ratio with a 95% confidence interval. Significant results had a p value < 0.05.

RESULTS: The NIS resulted 7,346 patients meeting the selected criteria. Cases with missing ZIP codes or insurance data were excluded. The mean age of patients residing in the lowest 25th percentile of MHI was 27 years compared to 28 years in other patients. Pregnant patients with sepsis from the lower 25th percentile had an odds ratio (OR) of 1.409 (CI 1.152 - 1.723, p = 0.001) of developing spontaneous abortions. Similarly, OR of developing the same complication in uninsured patients was 1.920 (CI 1.334 - 2.763, p < 0.001). OR for the length of stay in the hospital for patients of the lower 25th percentile MHI was 5.259 (CI 5.026 - 5.492, p < 0.001), and 5.495 (CI 5.304 - 5.685, p < 0.001) for uninsured patients.

CONCLUSIONS: Our findings suggest that there persists in the United States a significant impact of social and economic factors on health outcomes. In septic pregnant women fetal mortality was significantly higher in patients who were uninsured and patients from low-income zip codes. LOS was also prolonged for these two populations, however prolonged LOS for low SES patients is not a new phenomenon, as these patients tend to require more resources. Consequently supplementary compensation has been proposed for hospitals that disproportionately serve these populations. Our findings would therefore support a need in the US for greater resources devoted to supporting these patients, and the hospitals that serve them.

FACTORS ASSOCIATED WITH ACUTE HEALTHCARE USE IN A REPRESENTATIVE SAMPLE OF ADULTS EXPERIENCING HOMELESSNESS IN CALIFORNIA

Jessica D. Fields^1,2; Kim H. Nguyen^2,3; Tiana Moore^2,3; Jennafer Birkmeyer^2,3; Eve Perry^2,3; Margot Kushel^2,3

¹University of California Berkeley University of California San Francisco Joint Medical Program, Berkeley, CA; ²University of California San Francisco Benioff Homelessness and Housing Initiative, San Francisco, CA; ³University of California San Francisco Center for Vulnerable Populations, Department of Medicine, Zuckerberg San Francisco General Hospital, San Francisco, CA. (Control ID #3874593)

BACKGROUND: Poor health increases risk for homelessness, and homelessness is detrimental to health. Due to high prevalence of comorbidities and poor access to longitudinal care, people experiencing homelessness (PEH) are at high risk of emergency department (ED) use and hospitalization. In a representative sample of adult PEH in California (CA), we assessed prevalence of and factors associated with acute healthcare use.

METHODS: We recruited a representative sample of adult PEH across 8 counties in CA between Oct. 2021 and Nov. 2022. Our dependent variables were any past 6-month use of ED and physical health hospitalization. Using the Gelberg-Anderson Behavioral Model for Vulnerable Populations, our covariates included predisposing factors (sociodemographics, homeless histories), enabling factors (healthcare access), and need factors (physical and mental health, substance use). We weighted data to generate state-wide estimates and assessed bivariate relationships with chi-square and t-tests. Multivariable regression models were informed by collinearity diagnostics and prior hypotheses.

RESULTS: Among 3065 participants, the weighted median age was 47 (IQR 37-57), 69% male, 36% Latinx/Hispanic, 23% Non-Hispanic (NH) White, 24% NH Black. Three-quarters (73%) met time criteria for chronic homelessness; 74% were unsheltered. In the prior 6 months, 38.4% visited the ED and 22.2% were hospitalized. We present unadjusted bivariate rates and multivariable odds ratios of acute care use (Table 1). A sensitivity analysis removing insurance and regular source of care, out of concern for effect-cause, was similar.

CONCLUSIONS: Adult PEH have rates of ED use and hospitalization higher than previous reports and dramatically higher than housed populations. Need factors, which increase the risk of and are exacerbated by homelessness, are associated with acute care use. While insufficient to prevent acute care, insurance and primary care may facilitate access to needed care. Preventing and ending homelessness may decrease acute care use.

FACTORS ASSOCIATED WITH HIV TESTING AMONG ASIAN-AMERICANS: AN ANALYSIS OF THE NATIONAL HEALTH INTERVIEW SURVEY

Armaan Jamal¹; Malathi Srinivasan²; Andrew Nevins³; Gloria S. Kim⁴; Sanah Vohra²

¹Department of Medicine, Johns Hopkins University, Baltimore, MD; ²Medicine, Stanford University School of Medicine, Palo Alto, CA; ³Stanford University School of Medicine, Stanford, CA; ⁴Medicine, Stanford University, Stanford, CA. (Control ID #3873403)

BACKGROUND: Asian Americans are the fastest growing ethnic group in the United States. Despite universal HIV screening recommendations, Asian Americans have the highest rate of undiagnosed HIV, often due to cultural factors inhibiting HIV screening, prevention, and treatment. Limited research exists about factors associated with HIV testing among Asian Americans.

METHODS: We examined the National Health Interview Survey (NHIS) from 2006-2018 to determine factors associated with self-reported HIV screening (ever tested) in non-Hispanic white (NHW), Chinese, Asian Indian, Filipino, and Other Asian adults. Our sample included 200,349 individuals (NHW [n=185,357], Chinese [n=3,111], Asian Indian [n=3,033], Filipino [n=3,435] and Other Asian [n=5,413]). We used logistic regression to calculate the odds ratios (OR) and 95% confidence intervals (CI) to identify demographic (age, sex, race, marital status, region of residence [Western/Eastern/Southern/Midwestern US], nativity [US born], US citizenship), socioeconomic (education, income, occupation) and health-related (insurance coverage [Medicare/Medicaid/Private/Other], self-reported health status [HRQOL], and number of doctor visits in the past 12 months) factors associated with HIV testing.

RESULTS: In our sample, 36% of NHWs, 34% of Chinese, 36% of Asian Indians, 39% of Filipinos and 35% of Other Asians reported lifetime HIV testing. All Asian subgroups reported lower odds of HIV testing compared to NHWs (Chinese: OR [95% CI] = 0.63 [0.58-0.69]; Asian Indian: 0.60 [0.55-0.65], Filipino: 0.84 [0.77-0.91] and Other Asian: 0.68 [0.63-0.72]). Within aggregate Asians, female gender (vs. male), 27-64 age group (vs. 18-26), higher education level, higher income, health insurance coverage and higher number of doctor visits were positively associated with HIV testing (OR range: 1.16-1.94). On the other hand, the 65+ age group (vs. 18-26), and foreign-born status decreased the odds of HIV testing (OR range: 0.62-0.91). Across the four Asian American subgroups, 27-64 age group, any health insurance coverage and higher number of doctor visits increased odds of HIV testing (OR range: 1.91-2.41). Within Chinese, Filipinos and Other Asians, higher income increased odds of HIV testing (OR range: 1.16-1.34). Additionally, within Asian Indians and Filipinos, foreign-born status decreased odds of HIV testing (OR range: 0.52-0.80)

CONCLUSIONS: Asian Americans report being screened for HIV less often as non-Hispanic whites, after adjusting for sociodemographic features. Creating culturally appropriate interventions to increase HIV screening rates may improve HIV outcomes amongst Asian Americans.

FACTORS ASSOCIATED WITH HIV-TESTING IN THE US POPULATION OVERALL AND BY RACE/ETHNICITY: AN ANALYSIS OF THE 2006-2018 NATIONAL HEALTH INTERVIEW SURVEY

Armaan Jamal³; Malathi Srinivasan¹; Andrew Nevins²; Gloria S. Kim¹; Sanah Vohra⁴

¹Medicine, Stanford University School of Medicine, Palo Alto, CA; ²Stanford University School of Medicine, Stanford, CA; ³Department of Medicine, Johns Hopkins University School of Medicine, Baltimore, MD; ⁴University of California Los Angeles David Geffen School of Medicine, Los Angeles, CA. (Control ID #3869894)

BACKGROUND: The Center for Disease Control and Prevention recommends that individuals between 13 and 64 obtain HIV screening at least once, as part of routine preventive health care. While universal HIV screening for adults has been recommended since 2006, screening has been uneven across racial/ethnic groups. In this study, we examined national HIV screening prevalence among racial/ethnic groups in the US, and the sociodemographic factors associated with testing.

METHODS: Using National Health Interview Survey (NHIS) (2006-2018), we analyzed participant HIV testing status (self-reported, ever tested), analyzed by self-identified race and demographics. Our sample included 282,433 individuals (non-Hispanic white (NHW) [n=183,323], Hispanic [n=45,271], non-Hispanic Black (NHB) [39,104] and Asian [14,753]). Using logistic regression, we calculated odds ratios (OR) and 95% confidence intervals (CI) to identify demographic (age, sex, race, marital status, region of residence [Western/Eastern/Southern/Midwestern US], nativity [US born], US citizenship), socio-economic (education, income, occupation) and health-related (insurance coverage [Medicare/Medicaid/Private/Other], self-reported health status [HRQOL], and doctor visits in the past 12 months) factors associated with HIV testing.

RESULTS: We found that 36% of NHWs, 46% of NHBs, 60% of Hispanics and 36% of Asians reported having an HIV test. Hispanics (OR [95% CI]: 1.30 [1.27-1.34]) and NHBs (2.46 [2.40-2.42]) had higher odds of HIV testing, whereas Asians (0.74 [0.71-0.77]) had lower odds of HIV testing compared to NHWs.

Factors that were positively associated with HIV testing included female gender, 27-64 age group (vs. 18-26), higher education level, higher income, working in service and professional specialties (vs. managerial/administrative), any health insurance coverage, higher number of doctor visits, US citizenship, foreign born status and residing in western and southern US (vs. northeastern) (OR range: 1.04-2.23). Factors that were negatively associated with ever being tested for HIV included age 65+, production/manufacturing jobs, married, better self-reported health status (vs. fair/poor), and living in the midwestern US (OR range: 0.62-0.91). Factors associated with HIV testing varied when stratified by race/ethnicity.

CONCLUSIONS: Race/ethnicity is significantly associated with self-reported HIV testing. We found that non-Hispanic blacks reported testing twice as often, Hispanics a third more often, and Asians three-quarters less often, as non-Hispanic whites, after adjusting for key covariates. Tailoring HIV screening outreach to specific racial/ethnic and sociodemographic groups may reduce health disparities in HIV detection and treatment, improving public health across the nation.

GETTING GRANULAR: DIABETES AND HYPERTENSION DISPARITIES AMONG CHINESE AND SPANISH SPEAKERS IN A LARGE HEALTH SYSTEM

Chai Arnold^1,2; Leonor Fernandez³

¹Diversity, Equity and Inclusion, Beth Israel Lahey Health, Cambridge, MA; ²Harvard Medical School, Boston, MA; ³Medicine, Beth Israel Deaconess Medical Center, Waban, MA. (Control ID #3874436)

BACKGROUND: Patients with Limited English proficiency (LEP) experience many health disparities. As part of a larger effort to reduce racial and ethnic disparities in diabetes and hypertension in a large health system, our analysis aims to provide a preliminary perspective on the relationship between LEP and disparities in chronic disease.

METHODS: This QI study is a cross sectional analysis of data collected from January to July 2022 in a large health system in New England. We used Arcadia software to identify patients with at least one billed diagnosis of hypertension and/or diabetes in the preceding 12 months. Preferred language (PL) was obtained from EMR patient registration data. Our analysis focused on Medicaid and Medicare subgroups. Primary measures included the most recent HbA1c for patients with diabetes and the most recent documented blood pressure for patients with hypertension. Sample T-tests were used to compare LEP and English-speaking subgroups.

RESULTS: Chinese-speaking patients in the Medicare subgroup had comparable mean HbA1c to English speakers (7.25 (N=132) vs. 7.27 (N=5977)), but significantly higher mean blood pressures (140.07 (N=160) vs. 129.58 (N=8337), p-value <0.001) compared to English-speaking patients. In contrast, Spanish speakers in the Medicare subgroup had significantly higher mean HbA1c compared to English speakers (7.68 (N=133) vs. 7.25 (N=5977), p-value 0.0012), with no significant differences in mean blood pressure. Similar findings were present within the Medicaid subgroups though with less variation among PL groups.

CONCLUSIONS: While limited English proficiency is often associated with disparities in both diabetes and hypertension, Chinese and Spanish LEP groups exhibit markedly different patterns in this study, with less glycemic control among Spanish speakers, and less hypertension control among Chinese speakers. This heterogeneity in chronic disease subgroup outcomes is relevant when setting meaningful goals for improving equity. Among many, education, literacy, diet, physical activity, cost of specific medications, and cultural factors specific to each chronic condition may each play important roles in shaping intersectional or compound inequities. Exploring and understanding these granular patterns should guide health system equity strategies and population health interventions.

HEALTH AND CARE NEEDS OF TRANSITION AGE YOUTH EXITING LOS ANGELES JAILS AND SUCCESSFUL ENGAGEMENT DURING REENTRY

Joyce Lee¹; Jessica Jara^2,6; Stefanie D. Vassar³; Laura Abrams⁴; Christine Grella⁵; Mitchell D. Wong⁶; Elizabeth Barnert⁷

¹School of Medicine, University of California Los Angeles David Geffen School of Medicine, Los Angeles, CA; ²Los Angeles County Department of Health Services, Los Angeles, CA; ³Division of Internal Medicine & Health Service Research, UCLA, Los Angeles, CA; ⁴University of California Los Angeles Meyer and Renee Luskin School of Public Affairs, Los Angeles, CA; ⁵Psychiatry and Biobehavioral Sciences, University of California Los Angeles David Geffen School of Medicine, Los Angeles, CA; ⁶Medicine, David Geffen School of Medicine at UCLA, Mitchell Wong, CA; ⁷Pediatrics, University of California Los Angeles David Geffen School of Medicine, Los Angeles, CA. (Control ID #3869432)

BACKGROUND: Reentry from jail to the community is a difficult juncture for transition age youth (TAY; ages 18-24), who simultaneously face the challenges of emerging adulthood. Health-related needs during reentry for TAY may be substantial. However, few empiric data exist to describe the healthcare and social service needs of TAY during reentry. Further, poor engagement with reentry interventions for TAY limitstheir impact, and factors affecting engagement are not well understood. We sought to describe: 1) the healthcare and social service needs of, and 2) factors associated with, TAY engaging with a reentry intervention post-release.

METHODS: We partnered with a community health worker-based intervention that connects adults returning home from jails to health and social services. We examined county electronic health and sheriff’s department data to describe participants’ criminal justice involvement, health diagnoses, reentry service needs, and reentry services utilization from May 2017 to December 2021. Post-release engagement was defined as connection to post-release services. We performed chi-squared tests and t-tests to compare needs and characteristics of TAY who did versus did not engage in post-release services. We then used logistic regressions to identify factors associated with engagement.

RESULTS: Our analytic sample included 2,525 TAY clients in the community program. Most participants were male (72.6%), Hispanic or Black (80.2%), and had a history of being unhoused (60.9%). Among all TAY clients, diagnoses of mental disorders were most common (57.2%), followed by substance use disorders (45.8%), physical health conditions (39.9%), and chronic pain (38.1%). Overall, most participants had good physical health with mean Charlson Comorbidity Index (CCI) score 0.53. Most common reentry service needs by domains were social needs (40.7%) and substance use treatment (21.8%), and by sub-domains were primary care access (19.6%) and housing (14.4%). Older age and history of being unhoused were independently associated with post-release engagement (age: odds ratio [OR]=1.06, 95% CI 1.02 to 1.11; ever unhoused: OR=1.18, 95% CI 1.00 to 1.40).

CONCLUSIONS: TAY exiting jail have high-stakes behavioral health and social needs, with housing a key basic need for linkage. Interventions tailored to TAY, particularly younger TAY, may improve linkages to care during reentry, and can ultimately serve as a lever to address health inequities and address cycles of incarceration at a crucial life stage.

HEALTHCARE ACCESS AND COVID-19 VACCINATION IN THE US: A NATIONAL STUDY Lily Gage¹; Jaclyn Perreault¹; Chloe A. Thomas¹; Charles De Guzman¹; Dier Hu¹; Lynn Wiwanto¹; Jose Henriquez-Rivera¹; Emily Harris¹; Danny McCormick^2,1; Adam Gaffney^2,1

¹Internal Medicine, Cambridge Health Alliance, Cambridge, MA; ²Medicine, Harvard Medical School, Boston, MA. (Control ID #3873460)

BACKGROUND: Healthcare access affects use of many preventive medical services. Although federal legislation made COVID-19 vaccines free, access to medical care may affect vaccine uptake through, for example, facilitating recommendations for vaccination from trusted primary providers. We sought to assess whether medical care access was associated with uptake and timeliness of COVID-19 vaccination in the US.

METHODS: We used a cross-sectional design to examine exposure to four metrics of healthcare access among sample adults in the 2021 National Health Interview Survey (NHIS): health insurance, having an established place for medical care, having a provider visit within the past year, and medical care affordability. We examined two outcomes: receipt of one or more COVID-19 vaccines, and (among those vaccinated) receipt of a first vaccine within 6 months of vaccine availability. We examined the association between each healthcare access metric and each outcome using logistic regression, unadjusted and adjusted for demographic, geographic, and socioeconomic covariates. We also constructed an “access score” with one point assigned for having access to each of these measures (score range 0-4) and then examined the association between each access score and receipt of COVID-19 vaccine using logistic regression.

RESULTS: Our sample included n=21,532 adults. Overall, participants with better access to care were older and had higher socioeconomic status. In unadjusted analyses, each metric of healthcare access was associated with uptake of COVID-19 vaccination, and (among those vaccinated) early vaccination. In adjusted analyses, having health coverage (adjusted odds ratio [AOR] 1.51; 95% CI 1.31, 1.74), a usual place of care (AOR 1.59; 95% CI 1.42, 1.76), and a provider visit within the past year (AOR 1.42, 95% CI 1.27, 1.58) remained associated with COVID-19 vaccination. Only having a usual place of care was associated with early vaccine uptake in adjusted analyses. In adjusted analyses, relative to those with an access score of 0, the odds of vaccination were 1.08-fold higher among those with a score of 1 (95% CI 0.68, 1.71), 1.30-fold higher among those with a score of 2 (95% CI 0.87, 1.94), 1.54-fold higher with a score of 3 (95%CI 1.05, 2.26), and 1.98-fold higher with a score of 4 (95% CI 1.35, 2.90).

CONCLUSIONS: Several individual metrics of healthcare access, as well as higher access scores based on these metrics, are independently associated with uptake of COVID-19 vaccines. Policies that achieve universal coverage, and facilitate long-term relationships with trusted providers, may be an important component of a pandemic response.

HEAT STRESS AND KIDNEY FUNCTION AMONG CONSTRUCTION WORKERS IN TEXAS Yosha Singh¹; Bethany Boggess Alcauter²; George Delclos³; Snehal Patel⁴

¹Dell Medical School, The University of Texas at Austin, Austin, TX; ²National Center for Farmworker Health, Buda, TX; ³The University of Texas Health Science Center at Houston, Houston, TX; ⁴Internal Medicine, The University of Texas at Austin Dell Medical School, Austin, TX. (Control ID #3872309)

BACKGROUND: Construction workers comprise 6% of the U.S. workforce but 36% of workplace heat- related deaths. Yet, there is a lack of peer-reviewed research documenting heat stress and strain among U.S. construction workers. Our study had two aims: quantitatively assess heat strain among construction workers in Texas and measure kidney function changes during a work shift.

METHODS: Twenty construction workers were recruited to assess heat stress, strain, and kidney function during working hours over two weeks. Workers were eligible if they were currently employed in construction; between 18 and 54 years old; and of male sex. Participants were excluded if they reported Type 1 diabetes.

Heat stress was assessed through temperature and humidity data from an iButton sensor on the participant. Core body temperature was also estimated through heart rate recordings from a BioHarness device. Dehydration levels were assessed through urine specific gravity measured pre- and post-shift each day. Participants completed a daily survey to report signs of heat-related illnesses, fluid intake, and use of non- steroidal anti-inflammatory drugs or alcohol, derived from the Second Central American Survey of Working Conditions and Health.

Kidney function changes were assessed through pre- and post-shift urine specimens, done twice a week for two weeks. Urine protein and pH were tested through dipstick analysis. A point-of-care device was used to analyze urine creatinine, albumin, and hemoglobin levels.

RESULTS: 75% of participants endorsed one or more symptoms of heat illness, per the survey. 56% reported fatigue, 12% reported headaches, and other symptoms included dizziness, cramps, blurred vision, and dark urine. BioHarness data was inconclusive.

100% of participants had at least one post-work urine specific gravity > 1.030, reflecting severe dehydration.

12% of participants’ urine revealed proteinuria, and 44% revealed hematuria without RBCs

CONCLUSIONS: Construction workers are a high-risk group to experience heat stress and strain, which affects kidney function.

Three-quarters of workers endorsed one or more symptoms of heat illness, and every worker exhibited severe dehydration at least once throughout the study. The BioHarness was unable to accurately record temperature in half of participants. We thus rely on survey data to reflect heat stress but will consider other methods in further studies.

Nearly half of participants had urine with hematuria without RBCs, concerning for myoglobinuria. Although the sample size and study duration were limited, this study reveals that Texan construction workers may be at risk for rhabdomyolysis.

Further study of heat strain and kidney function in construction workers is needed through serum creatinine and creatinine kinase to assess for rhabdomyolysis and kidney injury. Policy to prevent dehydration among construction workers should be improved. After the study ended, researchers held a workshop to recommend hydration tactics, short breaks, and urine tracking to workers.

HEMOGLOBIN A1C AND BODY MASS INDEX CHANGES FOLLOWING A HEALTHCARE SYSTEM-BASED FOOD INSECURITY INTERVENTION: PRELIMINARY DATA

Alexandra Cardy¹; Colleen Ereditario²; Elizabeth Cuevas¹; Bruce Ling²

¹Internal Medicine, Allegheny Health Network, Pittsburgh, PA; ²Allegheny Health Network, Pittsburgh, PA. (Control ID #3877217)

BACKGROUND: The Healthy Food Center (HFC) in Pittsburgh, Pennsylvania is a unique social determinants of health concept-driven referral-based intervention for patients experiencing food insecurity. The HFC is embedded within Allegheny Health Network, a large healthcare system in Pittsburgh, Pennsylvania, and offers comprehensive services addressing social determinants of health related to food insecurity. Patients are referred by providers based on screening positive for food insecurity, and receive fresh produce and cooking utensils, transportation to the HFC, produce coupons, dietitian counseling, and budgeted produce shopping education. Here we examine preliminary data related to changes in hemoglobin A1c (HgbA1c) and body mass index (BMI) following participation in this food insecurity intervention

METHODS: HgbA1c and BMI data were abstracted from HFC participants using the electronic medical record. BMI data were examined for the year 2021 and HgbA1c data for the years 2018-2021.These two variables were evaluated for change across two time points: prior to the initial HFC encounter and at six months post-initial HFC encounter. Participants were categorized into various baseline levels of BMI and HgbA1c with the data presented as frequencies. This data was compared statistically for changes at six months after baseline in BMI and HgbA1c respectively using Chi-Square test.

RESULTS: BMI data was available for 255 participants (mean baseline BMI 33.4) and HgbA1c data was available for 174 participants (mean baseline HgbA1c 7.9) at both time points. There was a higher frequency of decrease at six months in BMI (45.1%) than those who stayed the same (20.8%) or increased (34.1%). While not statistically significant (p=0.14), there was a trend for a higher percentage of those categorized as obese at baseline (BMI > 30, n=162) (49.4%) to show a decrease in BMI at six months compared to those not obese (BMI < 29.9, n=93) (37.6%).

For HgbA1c at six months, there was a higher frequency of decrease in HgbA1c (32.8%) than increase (25.3%). This finding was most notable (p<.00001) in the highest HgbA1c group (HgbA1c > 9, n=64) where 60.9% had a decrease at six months compared to those with a baseline HgbA1c of 7-8.9 (n=57) (26.3%) or < 7 (n=53) (5.7%).

CONCLUSIONS: Those with higher hemoglobin A1c and higher BMI at baseline had a higher propensity to show reductions at six months, suggesting that the intervention may be more beneficial for patients with poorly controlled chronic conditions. This preliminary data suggests that addressing social determinants of health with a healthcare system-based food insecurity intervention can facilitate chronic disease management at the population level, especially for those with more severe disease.

HISTORIC STRUCTURAL RACISM AND PRESENT-DAY DIABETES PREVALENCE: AN EXAMINATION OF STATE SPECIFIC ASSOCIATIONS

Sebastian Linde¹; Rebekah J. Walker³; Jennifer A. Campbell⁴; Laura Hawks⁵; Leonard E. Egede²

¹Medicine, Medical College of Wisconsin Department of Medicine, Milwaukee, WI; ²Medicine, Medical College of Wisconsin, Milwaukee, WI; ³Medicine, Medical College of Wisconsin, Milwaukee, WI; ⁴General Internal Medicine , Medical College of Wisconsin, Milwaukee, WI; ⁵General Internal Medicine, Medical College of Wisconsin, Wauwautosa, WI. (Control ID #3874548)

BACKGROUND: Recent work has shown that structural racism, as captured by historic residential redlining practices under the Home Owners’ Loan Corporation (HOLC), is associated with present-day diabetes (and all cause) mortality and years of life lost rates, and that these associations have remained consistently strong across decades (Linde, Walker, Campbell and Egede 2022). However, it remains unknown whether associations between historic structural racism and present-day diabetes prevalence exist and whether these associations vary across different US states. To this end, the objective of our study was to examine whether structural racism (proxied using historic residential redlining practices) is associated with present-day diabetes prevalence, and whether these associations vary across US states.

METHODS: In constructing our analytic sample of 11,457 census tracts, we combined census tract level data across multiple sources. First, diabetes prevalence was sourced from the CDC Places database. Second, historic neighborhood redlining grades were constructed using historic HOLC maps from the Mapping Inequality project (Nelson et al. 2022). Third, census tract data on socioeconomic demographics and population counts were sourced from the Opportunity Insights database. For our estimation approach we utilized linear regression models with stepwise additions of covariates; and we also adjusted for state fixed effects to remove bias due to unobserved state-level confounders.

RESULTS: Diabetes prevalence was 8.9% within areas with a historic HOLC grading of A (best); 10.3% within areas graded as B (still desirable); 11.9% within areas graded as C (declining); and 13.1% within areas graded as D (hazardous). Preliminary state fixed effect regression estimates indicate that a unit increase of an area’s historic HOLC grading (e.g., an increase from a grade of B=”still desirable” to C=”declining”) is associated with a 1.47% (95%CI=; p<0.001) increase in diabetes prevalence. Additional estimation results indicate considerable heterogeneity (across different US states) within the association between historic residential redlining and present-day diabetes prevalence. Regression estimates from models with stepwise additions of covariates further suggest that historic redlining is primarily mediated by factors related to poverty, incarceration and education.

CONCLUSIONS: Structural racism, as proxied by historic residential redlining, is found to be associated with present-day diabetes prevalence. These associations are found to be both statistically and clinically meaningful, and they are found to vary across states with the strongest associations noted within southern US states. Policy programs targeted at addressing community level disparities in poverty, incarceration and education may help ameliorate the persistent influence of historic structural racist policies upon present-day population health.

HOW HOUSING STATUS AT BASELINE IMPACTED HOSPITALIZATION IN COMPLEX PATIENTS ENROLLED IN THE SUMMIT A-ICU RANDOMIZED CONTROLLED TRIAL AT A HEALTHCARE FOR HOMELESS CLINIC

Anna Geduldig¹; Meg Devoe¹; Priya Srikanth²; Eileen Vinton³; Richard Bruno³; Samuel T. Edwards^4,1; Brian Chan¹

¹Medicine, Oregon Health & Science University, Portland, OR; ²School of Public Health, Oregon Health & Science University, Portland, OR; ³Central City Concern, Portland, OR; ⁴Portland VA Medical Center, Portland, OR. (Control ID #3876844)

BACKGROUND: People experiencing homelessness are at risk for poor health outcomes and early death. “Ambulatory Intensive Care Units” (A-ICUs) have multi-disciplinary team members and flexibility to tailor care for medically and socially complex patients at risk for high-utilization. It is unclear which patients most benefit from these high-resource interventions. We assessed how housing instability at baseline impacted outcomes of patients enrolled in SUMMIT, a recently completed randomized controlled trial of an A-ICU at an urban healthcare for the homeless federally qualified health center.

METHODS: The SUMMIT A-ICU is made up of physician, nurse, care coordinators, social workers, pharmacist, and team manager with reduced panel size and flexible scheduling. Patients could be referred if they had 1+ hospitalizations in the prior 6-months and had 2 or more chronic medical, substance use disorder, or mental health diagnoses. We collected baseline surveys including unstable housing status (e.g., sleeping outside, place not meant for habitation, shelter) and assessed changes in 6-month rates of hospitalization (primary outcome) and patient reported health related quality of life scores (HRQoL) using the SF-12 survey. To assess whether housing status modified the impact SUMMIT had on outcomes, we fit a linear mixed effects model and tested for the three-way interaction between randomization arm, time point, and housing status, using P-value < .20 as the cutoff for significance. We report mean within-group (WG) change and standard error (SE) from the model.

RESULTS: The trial enrolled 159 patients. Patients with unstable housing (UH) at baseline were younger (UH 53.4 [SD 10.4] vs stably housed 57.0[SD 8.4], P-value=0.02) and had higher rates of cognitive impairment (49.3% vs 31.2%, P-value=0.02). We found evidence that housing status at baseline modified the effect of SUMMIT on hospitalization rates (P-value for interaction=0.19). SUMMIT patients with unstable housing at baseline had increased hospitalizations at 6-months (WG change in hospitalization = 0.40 [SE 0.66]) compared to those who were stably housed at baseline (-1.12 [SE 0.34]). We found similar evidence of interaction for HRQoL domains of physical health composite (interaction P-value = 0.15, UH WG change = 0.06 [SE 1.67] vs stably housed WG change = 4.91 [SE 2.21]) and physical functioning domain (interaction P-value = 0.05, UH WG change = -2.42 [SE 1.93] vs stably housed WG change = 4.18 [SE 2.12]).

CONCLUSIONS: Compared to patients who lacked housing immediately prior to enrolling in SUMMIT, patients who were stably housed experienced decreases in 6-month hospitalization rates and improved physical health and functioning. Having secure housing may promote ability to engage with the SUMMIT team. Stabilizing housing status for medically complex patients at risk of homelessness should be prioritized for these patients in parallel with intensive primary care interventions to improve outcomes.

HOW IMMIGRATION-RELATED FACTORS DIFFER BY LEVELS OF ACCULTURATION AND SOCIOECONOMIC STATUS: RESULTS FROM THE HISPANIC COMMUNITY HEALTH STUDY/STUDY OF LATINOS (HCHS/SOL) SOCIOCULTURAL ANCILLARY STUDY

Chelsea Truong¹; Amanda S. Hinerman¹; Erik Rodriquez¹; Aimee Afable²; Eliseo J. Perez-Stable¹

¹Division of Intramural Research, National Heart, Lung, and Blood Institute, Bethesda, MD; ²SUNY Downstate Health Sciences University School of Public Health, New York, NY. (Control ID #3872419)

BACKGROUND: Although it is generally believed that greater acculturation adversely impacts the health of Hispanics/Latinos and that acculturation level and socioeconomic status (SES) are closely related, segmented assimilation theory would suggest that years lived in the U.S. and age at immigration differ between immigrants who have higher vs. lower acculturation and SES levels. Using data from the Hispanic Community Health Study/Study of Latinos (HCHS/SOL), we tested this hypothesis.

METHODS: The study used data from Visit 1 (2008-2011) and the Sociocultural Ancillary Study of HCHS/SOL, a prospective cohort study in Bronx, Chicago, Miami, and San Diego. Binary and ordinal logistic regressions models were used to evaluate the cross-sectional differences between (1) years lived in the U.S. (0-5; 6-10; 11-15; 16-20; and 21+) and perceived SES (low; moderate; high), (2) years lived in the U.S. and acculturation level (low; high), and (3) age at immigration (<12; 12-18; 19+) and education (<9^th; 9^th to <12^th; high school; >high school).

RESULTS: Adults (n=4,287) who identified as Central American, Cuban, Dominican, Mexican, and Puerto Rican were included. Overall, 52% were female, the mean age was 41, the mean years lived in the U.S. was 21, and 60% had a high school education or less. Individuals who have lived in the U.S. for 11-15 years were more likely to have a higher perceived SES compared to individuals who have lived 0-5 years in the U.S. (OR=1.7, 95% CI=1.2, 2.3). Those who have lived in the U.S. for 0-5 years were more likely to be less acculturated compared to those who have lived 21+ years in the U.S. (OR=11.9, 95% CI=8.1, 17.4). Respondents who immigrated at age <12 were more likely to have higher education compared to individuals who immigrated at age >19 (OR=1.6, 95% CI=1.2, 2.1).

CONCLUSIONS: Different levels of acculturation are associated with diverging paths of SES, leading to unique immigrant experiences. Our results highlight the heterogeneity of Hispanics/Latinos and may provide a link to observed differing health patterns.

IDENTIFYING INDIVIDUALS WITH HIGHEST SOCIAL RISK IN ADULTS WITH TYPE 2

DIABETES: TWO PARAMETER ITEM RESPONSE THEORY METHODOLOGY

Leonard E. Egede²; Rebekah J. Walker¹; Sebastian Linde³; Joni S. Williams⁴

¹Medicine, Medical College of Wisconsin, Milwaukee, WI; ²Medicine, Medical College of Wisconsin, Milwaukee, WI; ³Medicine, Medical College of Wisconsin Department of Medicine, Milwaukee, WI; ⁴Medicine, Medical College of Wisconsin, Milwaukee, WI. (Control ID #3876216)

BACKGROUND: Social determinants of health are widely recognized as a contributor to health outcomes for adults type 2 diabetes. Though multiple social risk factors are identified, their impact on outcomes is not equal. In addition, prior research suggests simply adding the number of social risk factors present does not identify high social risk. The aim of this analysis was to use item response theory to discriminate across social risk factors in adults with type 2 diabetes and identify social risk indicators that reliably distinguish between high social risk and low social risk subgroups.

METHODS: Cross-sectional data of 615 adults with diabetes recruited from two primary care clinics were used. Participants completed assessments including validated scales on economic instability (financial hardship), the neighborhood and built environment (crime, violence, and neighborhood rating), education (highest education and health literacy), food environment (food insecurity and healthy food access), social and community context (social isolation), and psychological risk factors (perceived stress, depression, serious psychological distress, and diabetes distress). Item response theory (IRT) models were used to understand the association between a participant’s underlying level of a particular social risk factor and the probability of that response. A two-parameter logistic IRT model was used with each of the 12 social determinant factors being added as a separate parameter in the model. Higher values in item discrimination indicate better ability of a specific social risk factor in differentiating participants from each other.

RESULTS: The mean age was 61.3 with mean years having been diagnosed with diabetes 12.3. Sixty four percent of the sample was Non-Hispanic Black, 38.4% were women, and 48.8% received care at a VAMC. Rate of crime reported in a neighborhood (discrimination 3.13, SE 0.50; item difficulty -0.68, SE 0.07) and neighborhood rating (discrimination 4.02, SE 0.87; item difficulty -1.04, SE 0.08) had the highest discrimination (see Figure 1 for comparison between factors).

CONCLUSIONS: Based on these findings, crime and neighborhood rating discriminate best between individuals with type 2 diabetes who have high social risk and those with low social risk. These two questions can be used as a parsimonious social risk screening tool for research and clinical purposes.

IF MY MOOD IS LOW, I CAN’T HELP THE PATIENT: A QUALITATIVE STUDY TO UNDERSTAND THE MENTAL HEALTH CHALLENGES OF HOME HEALTH AIDES

Melissa Yanez¹; Elizabeth Kuo¹; Ann Lee²; Aida E. Ramos¹; Joanna B. Ringel¹; Elissa Kozlov³; Susan J. Andreae⁴; Monika M. Safford¹; Ariel Avgar⁵; Megan Shen⁷; Daniel Shalev¹; Catherine Riffin⁶; Faith Wiggins²; Madeline R. Sterling¹

¹Medicine, Weill Cornell Medicine, New York, NY; ²Training and Employment Fund, 1199SEIU United Health Care Workers East, New York, NY; ³School of Public Health, Rutgers The State University of New Jersey, New Brunswick, NJ; ⁴Kinesiology, University of Wisconsin- Madison, Madison, WI; ⁵School of Industrial and Labor Relations, Cornell University, Ithaca, NY;⁶Medicine, Weill Cornell Medicine, New York, NY; ⁷Fred Hutchinson Cancer Research Center, Seattle, WA. (Control ID #3872556)

BACKGROUND: Home Health Aides (HHAs) provide personal and medical care in the home to older adults and individuals with chronic conditions. Despite this, HHAs, themselves, are a vulnerable workforce that experiences high levels of stress and social isolation; these outcomes have worsened since the COVID-19 pandemic. In the present study, we aimed to understand the perspectives of HHAs towards their mental health challenges including how their job impacts their mood and potential interventions that can meet their needs.

METHODS: We conducted focus groups from August to December 2022 in partnership with the 1199SEIU Training and Employment Fund (TEF), a benefit fund of the 1199SEIU United Healthcare Workers East, the largest health care union in the U.S. TEF provides training and benefits to 55,000 agency-employed HHAs in New York, NY. We included English or Spanish speaking HHAs with ≥ 1 risk factor for poor mental health, defined as: 1) depressive symptoms (Personal Health Questionnaire 8-item [PHQ8] scale ≥ 5 points), 2) stress (Cohen’s Perceived Stress 4-item scale [PSS4] ≥5), and 3) loneliness (≥6 on the 3-item UCLA Loneliness scale). Focus groups were audio recorded, professionally transcribed, and analyzed thematically. The topic guide and analysis were informed by the Pender’s Health Promotion Model and the National Institute for Occupational Safety and Health’s Total Worker Health Model.

RESULTS: In total, 19 HHAs employed by 11 unique home care agencies participated. They had a mean age of 54 years (SD: 11.7), 18 (95%) were female, and 11 (56%) were Black; they had a median of 21 years (IQR: 10.0, 25.0) of job experience. Five main themes emerged: (1) HHAs personal lives outside of work, including family caregiving responsibilities, influences their mental health; (2) the day-to-day work associated with being a HHA, and their relationships with patients, impacts their mental health; (3) the COVID-19 pandemic worsened HHAs mood and stress levels; (4) HHAs use a variety of strategies (e.g. breathing exercises, religion, etc.,) to cope; (5) HHAs are eager for interventions that can improve their mental health, particularly peer support programs which can bring them closer to other HHAs and foster a sense of community.

CONCLUSIONS: For many HHAs, mental health is currently a challenge that is influenced by occupational and personal factors. Interventions with peer support features may be warranted for this workforce.

IMPACT OF MEDICALLY-TAILORED MEALS ON CLINICAL OUTCOMES AMONG LOW- INCOME ADULTS WITH DIABETES: A PILOT RANDOMIZED TRIAL

Jeanne M. Clark¹; May Thu Thu Maw¹; Kathy Pettway²; Geetanjali Chander³; Susan Elias⁴; Sam J. Zisow- McClean⁴; Nisa Maruthur¹; Raquel C. Greer¹. ¹Medicine/General Internal Medicine, Johns Hopkins University School of Medicine, Baltimore, MD; ²Priority Partners, Johns Hopkins Medicine, Baltimore, MD; ³University of Washington School of Medicine, Seattle, WA; ⁴Moveable Feast, Baltimore, MD. (Control ID #3875122)

BACKGROUND: Low-income adults with diabetes experience significant barriers to healthful eating, including limited knowledge and skills around diet, and limited access to healthy foods. We conducted a pilot randomized controlled trial to examine the impact of medically-tailored meals (MTM) with Medical Nutrition Therapy (MNT) on clinical outcomes among low-income adults with type 2 diabetes.

METHODS: We recruited English-speaking adults with type 2 diabetes (DM) and A1c >8% from Maryland Medicaid plans using mailed brochures or provider referral and follow-up phone calls. We excluded anyone unable to safely consume the MTM. Eligible individuals were randomized to usual care or an intervention involving home delivery of 12 medically tailored, frozen meals and a fresh produce bag weekly for 3 months, and individual phone calls with a Registered Dietitian monthly for 6 months. Our outcomes were change from baseline to 6 months in A1c (primary), body mass index (BMI), blood pressure and diabetes-related quality of life, knowledge and self-efficacy (secondary).

RESULTS: Of 1537 adults assessed for eligibility from January 2020 to March 2022, we randomized 73 into the study, of whom 57 (78%) had complete data on the outcomes of interest. Their mean age was 48 years, 40% were male, 77% Black, 47% had < college education, and 49% had household income < $30,000. The mean A1c was 10.3%, mean BMI was 37.9 kg/m², and 54% had ≥3 medical conditions at baseline. Eighty- six percent of meals were delivered, and the mean number of nutrition visits completed was 4.8. Overall, at 6 months both the intervention and control groups had improvements in A1c (-0.7 vs. -0.6%), with no significant difference between groups. In exploratory analyses, more favorable changes in diabetes medications in the control group likely explained some of the lack of difference in improvement in A1c between groups. There was a trend towards greater improvement in A1c among those who reported eating ≥9 meals weekly compared to less (-1.8 vs. -0.4%, p=0.1), but no difference by number of nutrition visits. Similar to A1c, changes in systolic blood pressure (-6.3 vs. -3.7 mmHg) and BMI (-0.02 vs. -0.20 kg/m²) at 6 months did not differ between groups. Food security improved significantly from baseline to 3 months in the intervention group (47 to 83%) compared to control (52 to 56%; p<0.01), but this was not sustained at 6 months, when meal deliveries had stopped. Diabetes related quality of life, knowledge and self-efficacy improved modestly, but not differently by group.

CONCLUSIONS: Even during COVID, recruitment and retention of an at-risk group of adults with DM was feasible. MTM intervention uptake was good but did not improve A1c or other outcomes. MTM interventions that are more specifically diabetes directed and clinically integrated are likely needed to achieve significant clinical benefits.

IMPLEMENTATION COMMUNITYRX-CARDIOVASCULAR DISASE, AN E-INTEGRATED COMMUNITY RESOURCE REFERRAL SYSTEM IN RURAL NC FEDERALLY QUALIFIED HEALTHCARE CLINICS

Maura Drewry²; Amelia DeFosset²; Christopher Shea²; Emily Abramsohn³; Leah Frerichs²; Donald Cavellini¹; Gaurav Dave²; Giselle Corbie²; Stacy T. Lindau³

¹Project GRACE, Rocky Mount, NC; ²The University of North Carolina at Chapel Hill, Chapel Hill, NC; ³The University of Chicago Medicine, Chicago, IL. (Control ID #3875957)

BACKGROUND: African Americans in the rural Southeast experience disproportionate cardiovascular disease (CVD) burden due to structural inequalities resulting in unmet disease self-management and social care needs. The COVID-19 pandemic has exacerbated these gaps in self- and social care needs, especially in rural minoritized communities. Federally Qualified Health Centers (FQHC) in rural settings, and the communities they serve, have unique strengths and challenges around implementing clinic-community care models. Understanding and working within this rural FQHC context is key to ensuring successful implementation processes and outcomes. CommunityRx-CVD is an evidence-based intervention that integrates an electronic health record (EHR)-based referral system with clinical workflows to link African American patients with local community resources to address self- and social care needs and support cardiometabolic health. Using Implementation Science frameworks, researchers partnered with clinic-based implementation teams to build, test, and refine custom workflows supported by mixed methods data collection activities. This study describes outcomes of that process.

METHODS: This study triangulates data from 3 sources conducted at 2 clinics within the same FQHC system: 1) an 8-item survey based on the Organizational Readiness for Implementing Change (ORIC) measure (item range: 1-5, 5 indicating complete readiness), self-administered to clinic staff pre- and post- testing cycles, 2) a structured reporting form to capture clinic staff experiences with the implementation workflow during testing cycles, and 3) semi-structured focus groups conducted with Medical Assistants (MAs) and Care Coordinators to further refine the workflow post testing cycles.

RESULTS: Thirty-four staff and providers across the 2 intervention clinics completed ORIC surveys (56.7% response rate). Preliminary data shows that physicians and leadership (n=8) have high change commitment scores and moderate change efficacy scores (mean=4.45 and 3.85). Clinic support staff (e.g., MAs, front desk staff) show high scores for change commitment and change efficacy (mean=4.35 and 4.23)

CONCLUSIONS: Clinical Staff readiness for changes related to the intervention is high as they enter the testing phase of implementation, but physicians and leadership may need additional support to feel ready.

This presentation will describe how changes in ORIC scores over the testing period are interpreted in tandem with data from the structured reporting form and qualitative focus groups, and used to drive implementation planning, such as changes to the implementation workflow or staff training, or provision of targeted supports to clinic leadership. Physician leaders can build off of the results and processes described in this presentation to explore implementing e-integrated community-resource referral systems in rural health clinics, allowing them to apply innovative approaches to address emerging and complex patient care challenges.

IMPLEMENTATION OF A UNIVERSAL FOOD INSECURITY SCREENING TOOL AT AN ACADEMIC FAMILY MEDICINE CLINIC

Katherine E. Barnes²; Chioma Onuoha¹; Kira Furie¹; Kevin Reyes¹; Aria Sikaroudi¹; Hasan Abdulbaki¹; Shwesha Govil²; Sarah GaleWyrick²

¹School of Medicine, University of California San Francisco School of Medicine, San Francisco, CA; ²School of Medicine, University of California San Francisco, San Francisco, CA. (Control ID #3858566)

BACKGROUND: Only a minority of hospitals and physician practices in the United States report regular screening for basic needs, including food insecurity. These low screening levels indicate that most practices lack a systematic screening procedure. This study assesses the feasibility and utility of a universal food insecurity screening tool at an academic family medicine clinic.

METHODS: In this observational study, we distributed a food insecurity screening tool to all patients at our clinic for three months. Our screening tool consisted of the validated 2-item Hunger Vital Sign™ screening

tool and two additional questions that assessed interest in food resources. The screening tool was distributed by front desk staff and abstracted by medical assistants. Patients could request information about local food resources regardless of whether they screened positive or negative. Physicians discussed these resources with patients and included them in the after-visit summary. Demographic information and screening results were summarized descriptively.

RESULTS: 1,005 patients were screened in three months. 122 patients (12.1%) screened positive for food insecurity. Distribution by race/ethnicity is shown in Table 1. Food insecurity was highest among patients residing in zip codes 94165 (66.7%), 94103 (55.6%), 94134 (22.9%), and 94124 (21.1%). It was lowest in zip codes 94080 (6.4%) and 94131 (7.7%). Data on interest in food resources was missing for 128 patients; of the remaining 734 patients, 78 (10.6%) requested resources. 48 of these patients (48.7%) did not report food insecurity.

CONCLUSIONS: Rates of food insecurity are highest in Black/African American, Latinx, and American Indian/Alaska Native populations, and the zip codes with the highest rates of food insecurity all had median incomes below the county average. This distribution mirrors that seen in social determinants of health nationwide, suggesting validity of our data. Our data also suggest that food resources should be offered regardless of whether a patient indicates food insecurity, as many patients who screened negative requested information about resources. Overall, implementation of a short universal food insecurity screening tool is feasible and reveals a substantial need for screening.

IMPLEMENTING LOW BARRIER COVID-19 WALK-UP TESTING CLINICS FOR VULNERABLE POPULATIONS TO INCREASE TESTING ACCESS.

Gloria Sclar²; David Ngandu²; Sumayo Awale²; Ambia Ahmed¹; Kathleen Fairfield^2,1; Swapnika Mallipeddi²; Hina Hashmi²; Grace Price²; Andrew Volkers³; Caroline Fernandes³; Marie Louise Mudasigana⁴; Leslie Nicoll⁵; Elizabeth A. Vinton²; Daisy E. Parker²; Elizabeth Jacobs⁶

¹Medicine, Maine Medical Center, Portland, ME; ²Institute for Research, MaineHealth, Portland, ME; ³Preble Street, Portland, ME; ⁴Greater Portland Health, Portland, ME; ⁵Portland Community Free Clinic, Portland, ME; ⁶Academic Affairs, MaineHealth, Portland, ME. (Control ID #3874470)

BACKGROUND: COVID-19 disproportionately affects people from vulnerable communities, who are also less likely to receive antiviral treatment. To reduce spread of COVID-19 and advance equitable access to antivirals, we must improve accessibility and availability of COVID-19 testing and treatment. Effective public health strategies for delivering care include community input and partnerships.

METHODS: With national RADx-UP initiative funding, we created a partnership between an academic medical center and three community partners to offer low-barrier COVID-19 walk-up testing clinics for vulnerable populations in Portland, Maine. We aimed to serve people who are unhoused, low-income and/or uninsured, and immigrant groups. The clinics were run collaboratively by the study team and community partners: one community-based organization (serving people who are unhoused), the City of Portland’s public health center (serving low-income/uninsured, as well as providing a syringe exchange program and a sexually transmitted disease treatment clinic), and a Federally Qualified Health Center (FQHC; serving a large and diverse immigrant community). We offered COVID-19 rapid antigen tests. Patients completed their own anterior nasal swab while clinic staff processed the test. Language interpretation was provided, as needed, and individuals were asked to complete a brief survey while awaiting results. For positive cases, we reported the case to the Maine CDC and ensured patients were connected to antiviral treatment (Paxlovid) within 24 hours, if desired and eligible.

RESULTS: Between 1/14/2022 and 11/28/2022, we tested 217 people across the three sites, 17 of whom were positive for COVID-19 (8%). Among positive patients, 4 received Paxlovid while the remaining declined treatment. People sought testing for a variety of reasons, including symptoms (n = 128; 59%), close contact exposure (n = 66; 30%), and/or need for a negative test result to access services (i.e. healthcare, shelter) or an activity (i.e. work, travel) (n = 59; 27%). Among patients who completed the survey questions, 38% were unhoused (38/101), 26% uninsured (25/97), 37% made less than $15,000 before taxes in 2019 (33/89), and 49% were non-White (70/143). Only 9% were unvaccinated for COVID-19 (8/88). Lastly, almost half of all patients tested previously had COVID (n = 104, 48%). The collaboration with community partners helped build trust; identify unique COVID support needs for different groups; and co-locate testing with other desired services such as clothing and other medical care.

CONCLUSIONS: Providing low-barrier testing clinics is instrumental to reducing the spread of COVID-19 and reducing disparities in antiviral treatment access. Partnering with trusted community organizations is an approach that may make vulnerable populations more comfortable accessing testing. With the availability of effective antiviral therapy, these partnerships provide a way to reduce COVID-19 treatment disparities.

INEQUALITIES IN COVID-19 VACCINE COVERAGE AMONG U.S. 5-11-YEAR-OLDS VS 12-17- YEAR-OLDS

Israel Agaku¹; Theodore G. Long²

¹NYC Test and Trace Corps, New York City Health and Hospitals Corporation, New York, NY; ²Ambulatory Care, NYC Health + Hospitals, New York, NY. (Control ID #3865771)

BACKGROUND: The Pfizer-BioNTech COVID-19 vaccine has been approved for use among U.S. children aged 5-17 years, but CDC vaccination data shows less vaccine coverage among children aged 5-11 years vs adolescents 12-17 years. This study quantified the disparity in COVID-19 vaccine coverage among U.S. children aged 5-11 vs 12-17 years using population-level data and explored underlying contributing factors.

METHODS: Data were from the Household Pulse Survey conducted during December 1, 2021—February 7, 2022. Target population was parents/guardians of vaccine-age children 5-17 years who reported living with their children in the household (n=52,345). Descriptive and multivariable analyses were performed.

RESULTS: Of all U.S. parents/guardians of children aged 5-17 years, 55.6% reported that at least one child in their household had received a COVID-19 vaccine, ranging from 35.8% in Mississippi to 72.4% in Massachusetts. For each given number of children in U.S. households during December 1, 2021—February 7, 2022, having ≥one child vaccinated was less prevalent among parents/guardians of 5–11-year-olds than those of 12–17-year-olds. For example, of single-child households, 72.5% of parents/guardians of 12-17- year-olds reported that the adolescent had received a COVID-19 vaccine, vs 41.6% of parents/guardians with a 5-11-year-old child. Compared to parents/guardians of unvaccinated 12-17-year-olds, parents/guardians of unvaccinated 5-11-year-olds were more likely to report that their pediatrician had not offered a COVID-19 vaccine to them; they also had higher prevalence of various safety-related concerns.

CONCLUSIONS: Intensified efforts are needed to educate parents/guardians about the safety of COVID-19 vaccines, and to offer the vaccines to those eligible.

Key words: COVID-19, children, vaccine, hesitancy, disparities.

INFLUENCE OF NEIGHBORHOOD SOCIO-ECONOMIC DEPRIVATION ON EFFECTIVENESS OF AN INTENSIVE LIFESTYLE INTERVENTION

Mamadou Sy¹; Scott J. Pilla¹; Kesha Baptiste-Roberts²; Tiffany L. Gary-Webb³; Jessica Yeh¹; Dhanajay Vaidya¹; Jeanne M. Clark¹

¹General Internal Medicine, Johns Hopkins University, Baltimore, MD; ²Department of Behavioral Health Science, Morgan State University, Baltimore, MD; ³Epidemiology, University of Pittsburgh Graduate School of Public Health, Pittsburgh, PA. (Control ID #3875163)

BACKGROUND: Lifestyle interventions, including behavior modification, nutrition, and physical activity, are recommended in diabetes management. While neighborhood socioeconomic status has been shown to be associated with glycemic control, whether it impacts the effectiveness of lifestyle interventions remains unclear. Therefore, we conducted an ancillary study to the Look AHEAD (Action for Health in Diabetes) trial to examine if neighborhood socio-economic deprivation, a composite measure of SDOH (education, income, and occupation), influenced the effectiveness of an intensive lifestyle intervention (ILI). We hypothesized that higher neighborhood deprivation would be associated with less improvement in clinical outcomes from an ILI. METHODS: Look AHEAD randomized patients with overweight/obesity and type 2 diabetes to an ILI aimed at weight loss, or a comparison group, Diabetes Support and Education (DSE). We linked individual- level data of participants from 4 study sites to environmental and community data from the 2000 United States (U.S) Census used to generate a summary score estimating neighborhood socio-economic deprivation. We analyzed the effect of tertile of neighborhood deprivation on weight and A1c changes from baseline across the first 4 years of follow-up using a mixed effects linear model with random intercept and an interaction term between neighborhood, study year, and study arm.

RESULTS: Among 1,213 participants, the mean age was 60 years, 41% were male, and 65%, 26%, and 4% had self-reported race/ethnicity of white, Black, and Hispanic, respectively. 84% had a college degree or greater, and 75% reported an annual income over $40,000. At baseline, the mean A1c was 7.3% and mean BMI 36.1 kg/m², mean deprivation score was -5.33 in the 1^st tertile (min: -12.04; max: -2.61) and 5.68 in the 3^rd tertile (min: 2.01; max:18.69). The intervention effectiveness on weight at year 1 did not differ by tertile of neighborhood deprivation in the ILI group (1^st vs 3^rd tertile: -8.99% vs -9.32%, p=0.6); or the DSE group (1^st vs 3^rd tertile: -0.19% vs -0.62%, p=0.5). There we also no differences at year 1 in effect on A1c in the ILI group (1^st vs 3^rd tertile: -0.71% vs -0.59%, p=0.27) or the DSE group (1^st vs 3^rd tertile: -0.13% vs -0.12%, p=0.96). Finally, there were no statistically significant differences between weight and A1c changes across the tertiles over the 4 years.

CONCLUSIONS: In this clinical trial population, the intensive lifestyle intervention was equally effective across levels of neighborhood socio-economic deprivation. However, these findings may not extend to individuals with the lowest income and educational attainment who are not typically represented in clinical trials and for whom more research is needed.

INSIGHTS INTO PATIENT PERCEPTIONS OF SOCIAL DETERMINANTS OF HEALTH (SDOH) AND THE USE OF CONDITIONAL CASH TRANSFERS (CCTS) TO ADDRESS THEM IN A PRIMARY CARE CLINIC

Marguerite Balasta¹; Kerry Meltzer¹; Anna U. Morgan²; Corinne M. Rhodes³

¹Department of General Internal Medicine, University of Pennsylvania, Philadelphia, PA; ²Medicine, University of Pennsylvania Perelman School of Medicine, Philadelphia, PA; ³Medicine, University of Pennsylvania, Philadelphia, PA. (Control ID #3873490)

BACKGROUND: The Centers for Medicare & Medicaid Services (CMS) has highlighted the importance of social determinants of health (SDOH) and mandated healthcare organizations conduct social needs assessments. Health systems struggle with effective ways to manage SDOH, often referring to community programs and government resources which are helpful but may not provide a solution to unique patient problems. There is little known about patient perception of SDOH and their impact on health. There is an opportunity to learn about how patients experience the health impacts of SDOH and tools for providers to effectively address them. Conditional cash transfers (CCTs) are a well-established intervention in behavioral economics and policy research to address SDOH but have not been studied in primary care.

METHODS: We conducted a randomized controlled trial to explore patient understanding of how SDOH influence health and the impact of CCTs. Thirty patients were recruited from 2 primary care clinics in Philadelphia November 2021 to February 2022. Eligible adults resided in zip codes with annual household income <$50k, had Medicare or Medicaid, and at least 1 chronic medical condition. All screened positive for at least 1 of 3 SDOH factors: financial strain, transportation needs, or food insecurity. We conducted semi- structured interviews at enrollment and 15 participants were randomized to receive $500. Patients were re- interviewed at 6 months.

RESULTS: Participants were 70% female, 90% black, and 54% had Medicare. 83% completed the study. Key themes related to patients’ reflections on the effect of SDOH on health, the role of primary care providers (PCPs) in addressing social needs, and views on CCTs were revealed through rapid qualitative analyses. While patients understood their social needs, they did not readily connect SDOH to health. Most felt PCPs had no role in addressing social needs and instead referenced SDOH as a personal responsibility. All were appreciative of CCTs and a few used it towards specific health-related uses like transportation to appointments, but most addessed a pressing financial need.

CONCLUSIONS: Patients recognize they have social needs, but do not clearly link SDOH to health and do not see addressing social needs as a responsibility of their primary care clinician.

INSURANCE STATUS AND PRIOR KNOWN DIAGNOSIS OF DIABETES AMONG PATIENTS HOSPITALIZED WITH COVID

Anisha Ganguly^2,1; Michael Harms²; Kavita Bhavan^2,1

¹Internal Medicine, The University of Texas Southwestern Medical Center, Dallas, TX; ²Parkland Health, Dallas, TX. (Control ID #3877128)

BACKGROUND: Underdiagnosis of diabetes has been associated with lack of insurance. Texas is a Medicaid non-expansion state and has the largest uninsured population in the country, resulting in decreased access to preventive care. During the pandemic, uninsured patients presented to health systems for the first time due to COVID hospitalization.

METHODS: Using EMR data from our safety-net hospital in Dallas, TX, we conducted a cross-sectional analysis of patients with diabetes hospitalized with COVID infection from June 2020 to December 2021 to examine the relationship of insurance and prior known diagnosis of diabetes. Diagnosis of diabetes was defined by an A1c of ≥6.5 and/or by diagnosis code. Insurance cohorts were uninsured, charity coverage, Medicare, and Medicaid. The primary outcome was known diagnosis of diabetes in the past year. Logistic regression was used to evaluate the association between prior known diagnosis of diabetes and insurance status.

RESULTS: Among patients with diabetes hospitalized with COVID (n=1,254), there were 22.1% uninsured, 20.9% charity coverage, 40.6% Medicare, and 16.4% Medicaid (Table 1). The majority of the population was male, >45, and Hispanic. After adjusting for age and race/ethnicity, the uninsured were less likely to have known diabetes prior to their COVID hospitalization compared to those with Medicaid (OR 0.56 [95% CI 0.39-0.82], p <0.01).

CONCLUSIONS: Uninsured patients with diabetes admitted with COVID were less likely to have a known diagnosis of diabetes before their COVID hospitalization. These findings reflect decreased primary care access and screening for diabetes among the uninsured in Texas. These findings may support universal testing for diabetes in acute settings to screen patients without primary care access and inform future policy efforts to expand healthcare coverage.

INVESTIGATING NAVIGATION TO HELP ADVANCE LUNG EQUITY (INHALE): A PRAGMATIC TRIAL OF PATIENT NAVIGATION FOR LUNG CANCER SCREENING AT A HEALTH CARE FOR THE HOMELESS PROGRAM

Travis P. Baggett^1,2; Nora Sp*rn¹; Joana Barbosa Teixeira³; Elijah C. Rodriguez⁴; Nillani Anandakugan⁵; Natalia Critchley¹; Evangeline Kennedy¹; Katherine Hart¹; Andrea Joyce⁵; Yuchiao Chang¹; Sanja Percac- Lima¹; Elyse Park⁵; Nancy A. Rigotti¹

¹Division of General Internal Medicine, Massachusetts General Hospital, Boston, MA; ²Boston Health Care for the Homeless Program, Boston, MA; ³Ragon Institute, Charlestown, MA; ⁴New York University Grossman School of Medicine, New York, NY; ⁵Massachusetts General Hospital, Boston, MA. (Control ID #3874258)

BACKGROUND: People experiencing homelessness die of lung cancer at over 2-fold higher rates than people in the general population. Lung cancer screening (LCS) with low dose computed tomography (LDCT) reduces lung cancer mortality among those at increased risk, but the circ*mstances of homelessness create multiple barriers to LCS LDCT completion. Patient navigation is a promising but untested strategy for overcoming these barriers in a homeless-experienced patient population.

METHODS: In 2020-2022, we conducted a pragmatic trial of LCS patient navigation at a large urban Health Care for the Homeless (HCH) program. To be eligible, participants were required to have a history of homelessness, have a primary care provider (PCP) at the HCH program, be proficient in English, and meet the 2015-2022 Medicare LCS coverage criteria: 55-77 years old, 30 pack-year smoking history, and smoking within the past 15 years. Enrolled participants completed a baseline survey and were randomized 2:1 to usual care with or without LCS navigation. Following a theory-based, patient-centered protocol, the navigator provided lung cancer education, facilitated LCS shared decision-making visits with PCPs, assisted in making and attending LCS LDCT appointments, arranged follow-up when needed, and offered tobacco cessation support for current smokers. The primary outcome was receipt of LCS LDCT at 6 months, ascertained by medical records review and compared between study arms with the Chi square test. Using a sequential explanatory mixed methods approach, qualitative interviews assessed participants’ perceptions of the navigation intervention.

RESULTS: Two hundred sixty participants were randomized to usual care with (n=173) or without (n=87) LCS navigation. The mean age was 60.5 years, 70.8% were male, 36.9% were Black, 36.9% were white, and

14.6% were Hispanic. One-third of participants (32.3%) were currently homeless, and 67.7% were formerly homeless (23.5% precariously housed, 43.1% stably housed). Half (50.8%) reported fair or poor health, and 85.0% currently smoked. Serious mental illness (28.8%), at-risk drug use (27.7%), and at-risk alcohol use (23.5%) were common. At 6 months, 43.4% of navigation arm participants and 9.2% of usual care arm participants had completed LCS LDCT (p<0.001). The intervention effect did not differ by housing status (p=0.68); however, across both study arms, housed individuals had higher LCS completion rates than currently homeless individuals in unadjusted (37.5% vs. 20.2%) and multivariable adjusted (aOR 3.52, 95% CI 1.57, 7.88) analyses. In qualitative interviews, participants identified multiple types of social support provided by the navigator, including informational, emotional, instrumental, and affirmational support

CONCLUSIONS: Patient navigation produced a nearly 5-fold increase in LCS completion at a large HCH organization. Even with individualized navigation support, housing status remained a major determinant of study outcomes.

INVESTIGATION OF WHETHER PERCEIVED ACCESS MEDIATES RACIAL AND ETHNIC DISPARITIES IN CHRONIC DISEASE CONTROL IN THE VETERANS HEALTH ADMINISTRATION: AN OBSERVATIONAL STUDY

Evan Shannon^2,1; W. N. Steers²; Donna L. Washington^2,1

¹Division of General Internal Medicine and Health Services Research, University of California Los Angeles, Los Angeles, CA; ²VA Center for the Study of Healthcare Innovation Implementation and Policy, Los Angeles, CA. (Control ID #3876060)

BACKGROUND: Literature suggests that minoritized racial and ethnic (e.g., Black and Hispanic) groups report lower perceived access to healthcare compared to non-Hispanic White patients. Prior research has also demonstrated lower rates of hypertension (HTN) and diabetes mellitus (DM) control among minoritized racial and ethnic groups compared to White individuals in the Veterans Health Administration (VA). However, it remains unclear if patient perceived access mediates the association between race or ethnicity and chronic disease control, such that minoritized groups perceive lower levels of access which, in turn, leads to worse disease control.

METHODS: We performed a cross-sectional analysis of linked data from the VA Patient Centered Medical Home (PCMH)-Survey of Healthcare Experience of Patients (SHEP) and the VA External Peer Review Program, a set of standardized clinical performance measures, for fiscal years 2016-2019. The primary outcomes of interest were HTN (<140/90 mmHg) and DM (HgbA1c<9%) control among patients with known disease. Our primary predictor, self-identified race and ethnicity, was categorized as American Indian/Alaska Native (AIAN), Asian, Black, Hispanic, Multi-Race, Native Hawaiian and Other Pacific Islander (NHOPI) and non-Hispanic White (White). The hypothesized mediator was patient perception of access to their PCMH, scored on a scale of 0-100 percentage points. We then tested for disparities between non-White and White patients, and tested whether any observed disparities were attributable to less perceived access among minoritized groups using the Tchetgen Tchetgen inverse odds weighting approach, and adjusting for age, sex, rural/urban status, socioeconomic status, education, self-rated physical and mental health and Gagne comorbidity score.

RESULTS: Our analytic HTN sample had 53,515 patients and our analytic DM sample had 34,004 patients. The race and ethnicity distributions for each sample were comparable, with the distribution for the HTN sample presented here: 0.7% AIAN, 0.6% Asian, 11.6% Black, 4.1% Hispanic, 0.8% Multi-Race, 0.8% NHOPI, 3.9% unknown, 77.4% White. After adjustment for covariates, Black patients had significantly lower rates of HTN control than White patients (75.7% v 79.6%, p<0.001). Both Black (83.7%) and Hispanic (83.1%) patients had significantly lower rates of DM control than White patients (86.9%, p<0.001 for both differences). After adjustment for covariates, there was no evidence that perceived access scores mediated the associations between Black race, Hispanic ethnicity and HTN or DM control.

CONCLUSIONS: We observed disparities in HTN and DM control among minoritized patients in our sample, but did not find evidence that patient’s perception of access to their PCMH mediated the association between race and ethnicity and HTN or DM control for these patients. Reducing racial and ethnic disparities within VA in HTN and DM control requires interventions beyond those focused on improving patient access.

IS SUPPORTIVE HOUSING SUPPORTIVE ENOUGH? QUALITATIVE RESULTS FROM THE PATIENT CENTERED HOUSING OPTIONS, OUTCOMES, SERVICES, AND ENVIRONMENT (PCHOOSE) STUDY

Maria Y. Patanwala^1,4; Roya Ijadi-Maghsoodi^2,3; Bikki Tran Smith⁵; Howard Padwa⁶; Melissa Chinchilla¹; Taylor Harris¹; Lillian Gelberg⁷

¹VA Greater Los Angeles Healthcare System, Los Angeles, CA; ²Division of Population Behavioral Health, University of California Los Angeles Health System, Los Angeles, CA; ³HSR&D Center for the Study of Healthcare Innovation, Implementation & Policy, VA Greater Los Angeles Healthcare System, Los Angeles, CA; ⁴Department of General Internal Medicine, University of California Los Angeles Health System, Los Angeles, CA; ⁵Department of Biomedical & Health Sciences, University of Vermont, Burlington, VT; ⁶Integrated Substance Abuse Programs, University of California Los Angeles David Geffen School of Medicine, Los Angeles, CA; ⁷Family Medicine, University of California Los Angeles, Los Angeles, CA. (Control ID #3876260)

BACKGROUND: The US has over half a million people experiencing homelessness (PEH) who, as a population, exhibit high rates of morbidity and mortality. Social support and housing are key determinants of health, and permanent supportive housing (PSH), by providing both, has the potential to improve health for PEH. Place-based PSH (PB-PSH) are buildings designated for PEH with onsite case managers, and scattered- site PSH (SS-PSH) are apartments throughout the community with mobile case managers. This qualitative study examines how recently housed adults describe their social supports in PSH and how social supports differ across PSH housing models. We aim to report perspectives and experiences of newly housed adults to inform PSH best practices in providing social supports in housing interventions.

METHODS: PCHOOSE is a comparative effectiveness study of different PSH models that recruited over 500 adults who entered PSH between January 2021 and June 2022. We invited a purposive sub-sample to complete 3 semi-structured interviews: at baseline (within the first month of housing) and at 3 and 6-months post-housing. This analysis focuses on baseline interviews (N=27), which explored housing history, health status, social supports, and current housing experience. Multiple researchers iteratively coded interview transcripts and performed thematic analyses related to social relationships and supports. We compared experiences by PB-PSH (N=11) versus SS-PSH (N=16) and interviewee characteristics.

RESULTS: Among the 27 baseline interviewees, the average age was 45 years-old, median time spent homeless was 7 years, 44% were men, 52% reported fair/poor physical health, and 42% reported fair/poor mental health. We identified four themes related to social supports in baseline interviews: 1) housing brought hope for improved relationships – “Now they’re all happy that I have a place, and they’re starting to talk to me again,” 2) housing changed physical or geographic conditions of relationships – “Now, so that I have a place—I have my kids over to visit…” 3) changes in social networks were sometimes desired – “…it’s what I needed, to be away from, one, away from people that I knew,” and 4) participants relied on formal supports like former and current case managers and informal supports like family for different needs or if the other support failed. At baseline, we identified no thematic differences by PSH unit type.

CONCLUSIONS: Our findings elucidate how newly housed adults perceive and navigate formal and informal sources of social support upon moving into PSH. Findings can be used to enhance assessments and services in PSH such as social health screenings that guide placement and connections to community- partnered organizations to fill support gaps. As cities respond to growing rates of homelessness with more PSH development, it is imperative that we also develop best practices within PSH that prioritize the social needs of homeless-experienced individuals to promote their health and retention in housing.

LATINO-WHITE DISPARITIES IN SCREENING FOR UNHEALTHY ALCOHOL USE IN THE OCHIN NETWORK OF COMMUNITY HEALTH CENTERS

Brian Chan^1,2; David Ezekiel-Herrera³; Jennifer A. Lucas³; Miguel Marino³; Steffani Bailey³; Sophia Giebultowicz²; Zoe Larson²; John Heintzman³

¹Medicine, Oregon Health & Science University, Portland, OR; ²Oregon Community Health Information Network, Portland, OR; ³Family Medicine, Oregon Health & Science University, Portland, OR. (Control ID #3876641)

BACKGROUND: Unhealthy alcohol use and the sequelae of alcohol-related diseases continue to be an important health concern; there is evidence that the COVID-19 pandemic was associated with increased alcohol use and related mortality. In 2018, the US Preventative Services Task Force (USPSTF) updated its recommendation to screen for unhealthy alcohol use in primary care settings in adults. It is known that Latino drinking patterns differ from non-Hispanic White (NHW) and other racial groups; while less likely to drink than NHW, Latinos who do drink may consume higher volumes and are at risk for health complications of chronic alcohol use. Given existing cardiovascular health disparities in Latino populations compared to NHW, equitable screening for unhealthy alcohol use and alcohol use disorders is important in community health settings that serve a high proportion of Latino patients.

METHODS: We evaluated the odds of unhealthy alcohol screening in adults by ethnicity and language preference. This was an observational cohort study using OCHIN data, a multi-state electronic health record (EHR) network of community health centers (CHCs). The population was adults seen in primary care between 2012-2020. We used logistic regression via generalized estimating equations to estimate the covariate-adjusted relative odds of receipt of alcohol screening, noted by EHR completion of Screening Brief Intervention and Referral to Treatment (SBIRT). Our predictor of interest was ethnic-language preference groups (NHW, Latino-Spanish language preferred, Latino-English language preferred). For regression analysis, NHW were considered the referent group and robust standard errors were estimated to allow for clustering at the clinic level.

RESULTS: There were 1,781,206 patients in the sample across 26 states with average age of 41.5 years (SD=15.3). Over half were female (57%); 22% were never insured, while 51% reported some public insurance. There were 57% NHW, 27% Latino-Spanish language preferred, and 16% Latino-English language preferred patients. Of the study sample, 41% had received SBIRT screening during this period. Latinos who preferred Spanish had 52% increased odds of alcohol screening compared to NHW (aOR=1.52, 95% CI: 1.26, 1.84). Compared to NHW, Latinos who preferred English had similar odds of alcohol screening completion (aOR=1.04, 95% CI 0.90, 1.22).

CONCLUSIONS: In a multi-state cohort of CHC patients seen in primary care, we found less than half of the patients received USPSTF recommended screening for unhealthy alcohol use, and that Latinos who speak Spanish had 52% higher odds of having been screened compared to non-Hispanic whites. Further investigation on how CHCs implement culturally relevant screening in diverse populations may aid in efforts to address screening and health equity gaps in other health settings.

MEDICAL STUDENTS WITH DISABILITIES: THE CRITICAL IMPORTANCE OF PROGRAM ACCESS TO EFFECTIVE LEARNING ENVIRONMENTS

Christopher Moreland¹; Melissa Plegue³; Karina Pereira-Lima³; Zoie Sheets⁴; Neera Jain⁵; Erene Stergiopoulous⁶; Benjamin Case⁷; Amy Addams⁸; Lisa Meeks²

¹Internal Medicine, Dell Medical School at UT Austin, Austin, TX; ²Learning Health Sciences, University of Michigan Medical School, Ann Arbor, MI;³University of Michigan, Ann Arbor, MI; ⁴University of Illinois Chicago, Chicago, IL; ⁵The University of British Columbia Faculty of Medicine, Vancouver, BC, Canada; ⁶University of Toronto, Toronto, ON, Canada; ⁷University of Wisconsin-Madison, Madison, WI; ⁸Association of American Medical Colleges, Washington, DC. (Control ID #3877317)

BACKGROUND: Disability is underrepresented in the workforce with student-to-physician attrition. Program access facilitates wellbeing for students with disabilities (SWD); learning climates contribute to attrition for many marginalized in medicine. We hypothesized that SWD perceive learning climates as suboptimal (vs nondisabled students) & that program access narrows this gap.

METHODS: The AAMC surveys 2nd-year medical students on agreement with learning environment statements & perceptions of:

1) Schools fostering students’ personal/professional development;

2) Environment via shortened version of validated Medical School Learning Environment Scale (MSLES) inventory;

3) 3 MSLES sub-scales corresponding to emotional climate, student-faculty interactions, student-student interactions.

In two cohorts (2019/2020) totaling 23898 students, we compared 3 groups (students reporting no disabilities, SWD reporting program access, SWD without program access) on learning environment measures (chi-square & ANOVA).

RESULTS: 2438 (10.2%) reported disability. 84% of SWD (n= 2039) reported having program access (i.e. appropriate accommodations or not needing accommodations). 16% (n=399) reported lack of program access (i.e. accommodations denied, under review, or not requested for some reason other than not needing them).

Fostering Development: Students without disabilities reported higher agreement that schools fostered development as a person (72%) & a physician (93%), compared with SWD, both with & without program access (p<0.001). SWD without program access reported lower agreement than SWD with program access (p<0.001). 66% & 86% (respectively) of SWD with program access endorsed development as a person & as a physician, compared to SWD without program access (44% & 79%, respectively).

Medical school environment: 9 statements address multiple domains (e.g., how performance is evaluated, isolation feelings, shared values). For 8 statements, nondisabled students reported higher agreement than SWD, both with & without program access (p<0.001); SWD with program access had higher agreement than those without (p<0.001).

Learning environment scales: address perceptions of student-student interactions (e.g. getting to know one another), student-faculty interactions (e.g. faculty guidance quality), emotional climate (e.g. if educational experience leads to sense of achievement). All groups differed on all scales (all p <0.001). Nondisabled students reported higher agreement than SWD both with & without program access (all p <0.01); SWD with program access reported higher agreement than those without program access (all p <0.001)

CONCLUSIONS: Our study is the first to explore SWD learning environment perceptions & program access. SWD report less favorable learning climates & program access mitigates differences, confirming accessible education’s importance. Medical schools should review structural accommodation barriers & enhance program access pathways.

MEDICATION MANAGEMENT SUPPORT FOR INDIVIDUALS WITH LIMITED ENGLISH PROFICIENCY: AN ONLINE REVIEW OF PATIENT LANGUAGE RESOURCES AND SURVEY OF PHARMACIES

Claire Arnold¹; Nancy Nguyen³; Joseph Herges³; Tyler Benning³; Rozalina G. McCoy². ¹Internal Medicine, Mayo Clinic Department of Internal Medicine, Rochester, MN; ²Division of Community Internal Medicine, Geriatrics, and Palliative Care, Mayo Foundation for Medical Education and Research, Rochester, MN; ³Mayo Clinic Minnesota, Rochester, MN. (Control ID #3876098)

BACKGROUND: Nearly 25 million Americans have limited English proficiency (LEP). Language services are needed in all sectors of the healthcare system, not only in direct clinical care settings. We therefore systematically examined the availability, accessibility, and breadth of language services available in pharmacies in a Midwest city with a high prevalence of LEP.

METHODS: Pharmacies in Rochester, Minnesota, were identified from institutional resources and augmented for completeness using an internet search. Pharmacies were characterized as chain, hospital/clinic-affiliated, and locally owned. We then systematically reviewed language resources offered by each pharmacy as publicized on their public-facing websites to assess what is available and accessible to individuals with LEP, documenting the specific language services offered, quality of patient instructions to access these services, and the number and types of supported languages. Results were cataloged and descriptively analyzed. To complement the information available to consumers online, we then created a RedCap structured interview guide with both closed- and open-ended questions for a telephone survey of pharmacy staff. Questions probed for specific language services available for each of the languages and staff satisfaction with currently available language services.

RESULTS: We identified 29 retail pharmacies, including 15 chain pharmacies, 10 hospital/clinic-affiliated, and four locally owned pharmacies. Ten of the 15 chain pharmacies offered interpreter services, and only one specifically listed the languages supported by interpreter services; others quantified the number of languages offered but did not specify which ones. Ten of the chain pharmacies indicated that they provide information in languages other than English, with 6 specifying what information is provided. Two of the chain pharmacies had a complex multistep process for setting up language services; these steps were outlined only in English with no non-English language support available on the website. Four chain pharmacies offered talking prescription labels that can be translated into other languages to accompany dual-language prescription labels. Locally owned pharmacies did not have any available information about language services but instead publicized different medication packaging options or offered medication consultations. Hospital/clinic- affiliated pharmacies only mentioned that interpreter services and/or medication consultations were available, but did not specify what services were available, the languages supported, or how to access them. Only one chain and one locally owned pharmacy offered the ability to translate their websites into Spanish; remaining pharmacies’ websites did not have a direct method of translating the contents of their websites

CONCLUSIONS: Structured interviews of all pharmacies are ongoing to better understand what may be available to consumers directly.

MENTAL HEALTH OUTCOMES AND HEALTHCARE SYSTEM ENGAGEMENT: AN ANALYSIS OF FORMERLY INCARCERATED TRANSGENDER INDIVIDUALS

Rachel Engelberg^1,2; Maria R. Khan¹

¹Population Health, NYU Langone Health, New York, NY; ²Internal Medicine, NYU Langone Health, New York, NY. (Control ID #3855439)

BACKGROUND: Transgender individuals face distinct challenges compared to cisgender counterparts, particularly in terms of mental health outcomes and healthcare access. Despite an increased risk for transgender individuals to experience incarceration, less is understood about healthcare disparities specific to transgender populations that have intersected with the criminal legal system. We sought to determine if formerly incarcerated transgender individuals experienced mental health disparities and differences in healthcare engagement compared to formerly incarcerated cisgender groups.

METHODS: In a secondary data analysis using the U.S. Transgender Population Health Survey (TransPop), we performed multivariate logistic regression to estimate the relationships between transgender identity and the following outcomes: attempted suicide, psychological distress, self-harming acts, having a doctor, and having health insurance. All analyses were conducted with respondents who had indicated a lifetime history of incarceration. Models were adjusted for key covariates of age, race/ethnicity, and education level.

RESULTS: Of respondents in the analytic sample (N=214), 24.3% identified as transgender. Among people with a history of incarceration, transgender participants (versus cisgender, the referent) were more likely to report attempted suicide (adjusted odds ratio (AOR): 6.3, 95% confidence interval (CI): 2.2-18.5), high psychological distress (AOR: 4.9, CI: 1.8-13.5), and self-harming behaviors (AOR: 3.8, CI: 1.1-12.4). There were no statistically significant differences between transgender status and having a doctor (AOR: 1.2, CI: 0.4-3.5) and having health insurance (AOR: 1.4, CI: 0.3-5.7).

CONCLUSIONS: Despite no differences in healthcare system engagement, formerly incarcerated transgender individuals experience significant mental health disparities when compared to criminal legal- involved cisgender peers. Formerly incarcerated transgender groups represent an at-risk population in need of improved access to mental health treatment.

MINORITIES, ADULTS 45-49, AND THOSE WITH LIMITED ENGLISH PROFICIENCY WERE LESS LIKELY TO REPORT RECEIVING COLORECTAL CANCER SCREENING IN 2019. Jennifer Bayly¹; Mara A. Schonberg²

¹General Internal Medicine, Beth Israel Deaconess Medical Center, Boston, MA; ²Medicine, Beth Israel Deaconess Medical Center, Brookline, MA. (Control ID #3873683)

BACKGROUND: Colorectal cancer (CRC) incidence has risen in adults younger than 50 years old, particularly within minority populations. Therefore, in 2018, the American Cancer Society (ACS) updated its CRC screening guidelines to lower the recommended age to begin CRC screening to age 45. ACS also recommend multiple modalities for screening. We aimed to examine receipt of CRC screening in US adults 45-75 after publication of these guidelines, especially among minorities and those with limited English proficiency (LEP).

METHODS: We analyzed cross-sectional data from the 2019 National Health Interview Survey among adults 45-75 years old without a history of CRC (weighted n=96,709,386). Adults were considered up to date with CRC screening (UTD) if they reported a screening colonoscopy within 10 years, FIT/DNA within 3 years, or any stool test within 1 year. Adults reporting sigmoidoscopy or CT colonography within 5 years were excluded (n=452) since the reason these tests were done was not assessed. Adults not UTD with screening were asked if a doctor had recommended CRC screening, and if so which test. We used multivariable logistic regression to examine receipt of UTD screening by age, race/ethnicity, limited English proficiency (defined by language of interview other than English only), race and other sociodemographic variables.

RESULTS: Participant mean age was 59 years (+/-0.09), 17% were aged 45-49 years, 11% were Black, 14% were Hispanic, and 7% had LEP. Overall, 53% of adults reported being UTD with CRC screening; 10% of adults aged 45-49 compared to 74% of adults 65-75. Adults with LEP were significantly less likely to be UTD with CRC screening than those without LEP (28% vs. 55%, (adjusted odds ratios [aOR] = 0.65; 95% confidence interval [CI] = 0.48 – 0.90) and Hispanic adults were less likely to be screened than non-Hispanic whites (36% vs. 58%, aOR=0.77). Among adults who were not UTD (47%, weighted n = 44,232,365), only 16% reported that their doctor recommended CRC screening (11% of those with LEP); Hispanics were less likely to report receiving a recommendation than non-Hispanic whites (aOR = 0.69 (0.48-0.96). Among

adults who reported being recommended screening, 73% were recommended a colonoscopy and 22% were recommended stool-based tests.

CONCLUSIONS: Despite new ACS guidelines, only 10% of adults 45-49 were UTD with CRC screening in 2019. Regardless of age, CRC screening was especially low in Hispanics and in adults with LEP. Concerningly these adults were also less likely to report a clinician recommendation for CRC screening. Despite guidelines recommending multiple screening modalities, clinicians continue to recommend colonoscopies more often than stool cards which may be easier for patients to obtain. Multi-level interventions are needed to increase CRC screening especially for adults 45-49, minority populations and those with LEP.

MORTALITY TRENDS AMONG HOMELESS-EXPERIENCED ADULTS IN BOSTON, MASSACHUSETTS OVER 16 YEARS

Danielle R. Fine^1,2; Kirsten Dickins³; Logan Adams^1,2; Nora Horick⁴; Jessie Gaeta^5,6; Elizabeth Lewis^5,7; Sarah Looby^8,2; Travis P. Baggett^9,2

¹Division of General Internal Medicine, Massachusetts General Hospital, Brookline, MA; ²Harvard Medical School, Boston, MA; ³Rush University Medical Center, Chicago, IL; ⁴Biostatistics, Massachusetts General Hospital, Boston, MA; ⁵Boston Health Care for the Homeless Program, Boston, MA; ⁶Boston Medical Center, Boston, MA; ⁷Boston University School of Public Health, Boston, MA; ⁸Yvonne L. Munn Center for Nursing Research, Massachusetts General Hospital, Boston, MA; ⁹General Medicine Division, Massachusetts General Hospital, Boston, MA. (Control ID #3847701)

BACKGROUND: Mortality rates among people experiencing homelessness (PEH) are high, but few studies have examined temporal mortality trends in this population. A comprehensive, temporal assessment of mortality among PEH can inform important equity-focused policy decisions. The objective of this study was to assess patterns in all-cause and cause-specific mortality in a large cohort of PEH.

METHODS: We linked a cohort of 60,092 adults seen at Boston Health Care for the Homeless Program (BHCHP) from 2003-2017 to Massachusetts death occurrence files spanning 2003-2018. Using the US urban Northeast population as the reference population, we used indirect standardization to calculate annual age- and sex-standardized all-cause mortality rates in the BHCHP cohort. We used Poisson regression models to compare temporal trends in the standardized BHCHP rates to the rates in the reference population. We used the cohort-by-year interaction term to determine whether the rates of change differed significantly between the BHCHP cohort and reference population. We repeated this process for the six leading causes of death observed in the BHCHP cohort. We also examined age-, sex-, and race/ethnicity-stratified all-cause and cause-specific mortality rates aggregated across time, in comparison to the reference population.

RESULTS: A total of 7,130 deaths occurred during 520,430 person-years of follow-up. From 2003-2018, the standardized all-cause mortality rate in the BHCHP cohort did not significantly change (0.16% annual growth; 95% CI: -0.36%, 0.68%; P=0.54), whereas the all-cause mortality rate in the urban Northeast population decreased 1.21% annually (95% CI: -1.25%, -1.17%; P<0.0001), contributing to a significant widening in the mortality disparity between these populations (P interaction<0.0001). Drug overdose was the leading cause of death (23.7%), with the standardized drug overdose mortality rate increasing on average 9.35% annually (95% CI: 7.90%, 10.79%; P<0.0001), compared to an average annual increase of 7.57% (95% CI: 7.34%, 7.80%; P<0.0001) in the urban Northeast (P_interaction=0.018). Cardiovascular disease, cancer, and liver disease were other leading causes of death in the BHCHP cohort, with rates that declined significantly over the study period. Suicide was a leading cause of death among BHCHP cohort men and women ages 18-34 years; homicide was a leading cause of death among Black and Latinx men <50 years; and HIV was a leading cause of death among Black and Latinx men and women ages 35-64 years.

CONCLUSIONS: In this large cohort of PEH, the mortality gap with the general population widened over a period of 16 years, driven partly by drug overdose deaths. Several demographic subgroups were differentially affected by certain causes of death and frequently demonstrated higher mortality rates compared to the general population. Policies addressing housing access and interventions tailored to meet the distinct needs of PEH are essential to reducing mortality disparities in this population.

NAVIGATING A CLIMATE OF BURDEN: THE EXPERIENCES OF UNDOCUMENTED IMMIGRANTS IN APPLYING FOR COVID-19 DISASTER RELIEF ASSISTANCE FOR IMMIGRANTS (DRAI) IN CALIFORNIA

Irving Ling¹; Hye Young Choi^3,2; May Sudhinaraset²

¹Internal Medicine, Kaiser Permanente Northern California, Oakland, CA; ²Community Health Science, University of California Los Angeles, Los Angeles, CA; ³Yale School of Medicine, New Haven, CT. (Control ID #3859954)

BACKGROUND: Undocumented immigrants experienced some of the highest levels of economic insecurity during the pandemic while being excluded from COVID-19 relief and unemployment benefits. In April 2020, California became the first state to offer financial aid to undocumented immigrants through the Disaster Relief Assistance for Immigrants (DRAI) program in collaboration with community-based organizations (CBOs). However, the application process was marked by many implementation challenges. This study represents one of the first to call attention to the personal experiences of undocumented immigrants in applying for DRAI and examines how administrative burden interacts with structural and personal factors to affect the decision to apply for aid. This study aims to generate recommendations to improve public policy interventions that address the social determinants of health of undocumented immigrants.

METHODS: Twenty participants were recruited through snowball sampling for in depth interviews. Eligible individuals were: 1) API or Latinx; 2) undocumented with or without DACA; 3) 18-39 years old; and 4) residing in California during the study. Audio recordings were transcribed via Otter.ai and coded via DeDoose. Qualitative analysis was used to identify themes through an iterative process between two trained researchers to develop the final coding manual upon saturation of the data.

RESULTS: The COVID-19 pandemic negatively affected the economic conditions of participants and their families. Collective scarcity, self-minimizing need, and notions of low self-efficacy were upstream barriers in applying for DRAI under a climate of burden stemming from factors, such as the fear of immigration enforcement. In the process of applying, the lack of intake capacity across phone lines and language access led to further administrative burdens for individuals already stretched thin. Lastly, trusted CBOs and young adults in undocumented or mixed immigrant households can play important roles in mediating levels of burden in navigating access to needed services and resources.

CONCLUSIONS: Our findings spotlight the complex ecosystems of burden that impact the health of undocumented immigrants. Our work argues for the importance of having their voices at the policy table when developing targeted interventions. If the climate of burden that influences how undocumented immigrants navigate systems and institutions is not addressed, policies intended to provide relief may instead have detrimental impacts that outweigh the benefits. Further research must also be done to improve governmental-CBO collaborations to build up community capacity and implement scalable interventions for undocumented immigrants.

OBESITY MANAGEMENT IN PEOPLE OF COLOR: EXPLORING THE IMPACT OF RACE/ETHNICITY ON PATIENT AND PROVIDER PERSPECTIVES

Michael G. Knight²; Amanda Velazquez³; Zhaoping Li⁴; Goutham Rao⁵; Anthony Fabricatore¹

¹Medical Affairs, Novo Nordisk Inc, Plainsboro, NJ; ²George Washington University Medical Faculty Associates, Washington, DC; ³Cedars-Sinai Center for Weight Management and Metabolic Health, Los Angeles, CA; ⁴University of California Los Angeles, Los Angeles, CA; ⁵University Hospitals, Cleveland, OH. (Control ID #3874408)

BACKGROUND: Over 40% of adults in the United States live with obesity; people of color are disproportionally affected (NHANES, 2021). Race and ethnicity may affect identification and treatment of obesity. This study explored the experiences of people of color with obesity (PoCwO) and health care professionals (HCPs) in obesity management.

METHODS: Patients diagnosed with obesity were recruited for the study via health care panels. In the qualitative phase, Zoom interviews about experiences with obesity management in PoCwO (body mass index >30; n=40) helped develop the larger quantitative survey. Both PoCwO and HCPs (>70% of time in patient care with a mix of patients from various racial/ethnic groups, 63% primary care providers) completed the quantitative phase surveys.

RESULTS: A total of 1001 patients with obesity who self-identified as Black/African American (n=273), Hispanic/Latinx (n=275), Asian American/Pacific Islander (n=249), and Native American/Alaskan Native (n=204) completed the quantitative survey between Nov 27, 2021 and Mar 2, 2022. For the provider perspective, 150 HCPs were included; 77 self-identified as White/European American, 13 as Black/African American, 13 as Hispanic/Latinx, 45 as Asian American/Pacific Islander, and 2 as Native American/Alaskan Native.

Approximately 7 in 10 PoCwO reported a belief that race/ethnicity left them disadvantaged in health care interactions, and ~3 in 10 PoCwO sensed biases about race/ethnicity regarding their weight from HCPs during their first conversation. PoCwO were more likely to feel comfortable initiating conversations about weight management with an HCP of the same race/ethnicity vs a different race/ethnicity (54% vs 47%).

Conversely, PoCwO with an HCP of a different race/ethnicity were more likely to have felt the tone used was condescending (11%) or overly critical (9%) vs those seeing a same-race/ethnic group HCP (≤4% for both terms). Of PoCwO who felt HCPs made assumptions about race/ethnicity in relation to their weight, fewer than half challenged those assumptions (eg, cultural dietary preferences); however, 71% still see the same HCP as was seen at their first conversation.

Approximately 40% of PoCwO desired more tailored advice (eg, racial/ethnic considerations), and 90% of HCPs were willing to receive education on cultural factors for more tailored conversations. Surveyed HCPs suggested PoCwO may have barriers to access to care; further education about which barriers exist and how to overcome them could be beneficial to patient care.

CONCLUSIONS: In this study, PoCwO indicated that perceived HCP biases about their race/ethnicity negatively impacted the care received for obesity management. Also, PoCwO felt more comfortable initiating conversations about weight with HCPs of similar race/ethnicity. Mitigating racial/ethnic assumptions among HCPs through further education and increasing the number of HCPs of color are potential steps toward improving care of PoCwO.

ORGANIZATIONAL ASSETS AND BARRIERS TO IMPLEMENTING EVIDENCE-BASED CARDIOVASCULAR DISEASE PROGRAMMING IN RURAL AFRICAN AMERICAN COMMUNITIES: EARLY FINDINGS OF AN INNOVATIVE COMMUNITY-ENGAGED ASSESSMENT AND PLANNING PROCESS

Breanna Williams¹; Amelia DeFosset²; Veena Reddy⁷; Mysha Wynn³; Katrina Blunt³; Scott Rosas⁶; Giselle Corbie⁴; Gaurav Dave⁵

¹Social Medicine- Center for Health Equity Research, The University of North Carolina at Chapel Hill, Chapel Hill, NC; ²Social Medicine, Center for Health Equity Research, UNC Chapel Hill School of Medicine, Chapel Hill, NC; ³Health Services, Project Momentum, Inc,, Rocky Mount, NC; ⁴Social Medicine, University of North Carolina at Chapel Hill School of Medicine, Chapel Hill, NC; ⁵Medicine, University of North Carolina at Chapel Hill, Chapel Hill, NC; ⁶Concept systems, Inc., Ithaca, NY; ⁷Social Medicine, The University of North Carolina at Chapel Hill, Chapel Hill, NC. (Control ID #3875931).

BACKGROUND: African Americans in the rural southeast US experience significant cardiovascular disease (CVD) disparities. Implementing evidence-based CVD programs (EBPs) within existing local health and social support infrastructure is a strengths-based approach to reducing CVD disparities. We must examine the perspectives of organizations in rural African American communities on delivering and sustaining CVD EBPs. We employ Group Concept Mapping (GCM), a participatory mixed methods approach for synthesizing group perspectives into a detailed conceptual model. We describe early qualitative results from the GCM process and discuss practical implications for action planning.

METHODS: We recruited representatives from clinic-, community-, and faith-based organizations serving African Americans in rural counties in North Carolina using a mixed purposive and convenience sampling approach. Participants engaged in three activities: 1) idea generation in response to an open-ended prompt (i.e., brainstorming); 2) sorting and rating of ideas; and 3) collaborative interpretation of a concept map, a visual and mathematical representation of the interrelationships among their ideas. This presentation outlines results of the first activity, brainstorming, which occurred over six weeks in fall 2022.

RESULTS: Participants (n=31), representing four clinics, 13 community-, and 14 faith-based organizations,contributed 87 ideas during brainstorming. These ideas fell into seven unique categories: CVD education in the community, 2) access to resources to support CVD health, 3) sufficient program funding, 4) sufficient program personnel, 5) access to other program delivery resources, 6) participant motivation, and 7) engagement strategies. Subsequent GCM activities, and related analyses, will occur through Feb. 2023. Participants will sort these ideas into similar categories and rate each idea’s feasibility and importance. The resulting clusters will represent an overall concept map that depicts factors influencing implementation of CVD EBPs in rural communities.

CONCLUSIONS: Preliminary qualitative results highlight education, access to resources, and community support as critical contextual factors influencing the scale of CVD EBPs in rural African American communities. These early findings point to several leverage points for action, including community-engaged CVD education efforts, strengthening networks to promote resource sharing, and utilizing context-grounded marketing approaches to foster community support. Additional activities and analysis generated through this collaborative mixed methods process will provide insight into how emerging ideas should be prioritized in relation to one another to inform our local action planning and implementation effort. The GCM process and results described by this study could serve as a model for other physician leaders interested in using innovative, community-engaged approaches to disseminate and implement best practices to foster health equity.

OUTCOMES OF A PEER NAVIGATION PROGRAM FOR BLACK PATIENTS IN PRIMARY CARE WITH UNCONTROLLED HYPERTENSION

Anna M. Morenz¹; Jessica Bender¹; Babette Hairston²; Helene Starks¹; Sara L. Jackson¹

¹Department of Medicine, University of Washington, Seattle, WA; ²School of Nursing, University of Washington, Newcastle, WA. (Control ID #3874940)

BACKGROUND: Black individuals in the U.S. experience higher rates of hypertension and worse control compared to White individuals, largely driven by interpersonal, institutional, and systemic racism. We conducted focus groups in the local Black community to understand barriers to and facilitators of health and blood pressure control to inform the development of a pilot peer navigation program for Black patients with uncontrolled hypertension in our primary care clinic. Goals were to improve engagement, trust, health care satisfaction, and blood pressure control.

METHODS: Criteria for program enrollment included Black race, being born in the U.S., age over 18 years, and having two systolic readings above 150 mmHg and/or diastolic above 100 mmHg in the past year.

Eligible participants were identified via record review or direct referral from their primary care team. Peer navigators, with experience living and investing in the local Black community, outreached to eligible participants. If the patient was interested, the navigator screened for social determinants of health and connected patients to community resources, supported them for a minimum of six months, and gave a blood pressure cuff if needed. Due to the COVID-19 pandemic, outreach was mostly via phone and later, in person in clinic. Participants’ clinical information was obtained from the medical record, and baseline and exit surveys were conducted.

RESULTS: Among 499 eligible patients in the clinic, 53 (10.6%) were enrolled from 2/1/2020 to 6/5/2022. Compared to a subset of 125 matched controls, participants were more likely to have a higher body mass index (34.6 vs 30.7 kg/m2), higher mean systolic blood pressure (151 vs 140 mmHg), and higher rate of diabetes (58% vs 39%).

Only 14 participants’ post-program blood pressures were available within a month of exit from the program. For these participants, mean systolic blood pressure decreased to 142 mmHg, compared to 139 mmHg in the control group. Average difference in systolic blood pressure pre/post-program for the navigator group was -9.7 mmHg (95% CI -19.3 to -0.1) compared to -1.6 mmHg (95% CI -5.7 to 2.5) for the comparison group (p=0.11). Of the 25 participants who filled out both baseline and exit surveys, 64% (16) of participants reported taking their blood pressure medications as prescribed over the past 7 days compared to 92% (23) post-intervention (p=0.02). In exit surveys, 79% (26/33) rated the overall experience of working with a peer navigator as “very good” or “excellent.”

CONCLUSIONS: In a safety-net primary care clinic, a community-inspired peer navigation program to support Black patients with uncontrolled hypertension led to a 10 mmHg decrease in average systolic blood pressure post-intervention (versus a 1.6 mmHg decrease in matched controls), in a small number of patients in this real-world intervention. Despite challenges brought by the arrival of COVID-19 early in the program, participants reported a high level of satisfaction with peer navigation.

PATIENT PERSPECTIVES ON A HEALTHCARE SYSTEM FOOD INSECURITY INTERVENTION

Alexandra Cardy¹; Colleen Ereditario²; Suzanne Mitchell²; Tori Vallana²; Janelle Craig²; Elizabeth Cuevas^2,1; Bruce Ling²

¹Internal Medicine, Allegheny Health Network, Pittsburgh, PA; ²Allegheny Health Network, Pittsburgh, PA. (Control ID #3877116)

BACKGROUND: Food insecurity (FI) is characterized by limited access to adequate food. The relationship between chronic disease and FI makes it a meaningful investment for healthcare systems. The Healthy Food Center (HFC) in Pittsburgh, Pennsylvania is a healthcare system-based FI intervention. Patients that screen positive for FI during outpatient visits are referred by a physician to one of five HFC locations within Allegheny Health Network. Patients are offered healthy food choices including fresh produce, as well as cooking demos, nutrition counseling with a dietitian, shopping on a budget education, basic cooking utensils, produce coupons, and assistance with transportation to the HFC. This project qualitatively explores patient experiences with the HFC.

METHODS: Focus group discussions were conducted among HFC participants. Semi-structured interview questions guided discussions. Focus groups occurred via secure video due to the COVID-19 pandemic and lasted for approximately one hour. The discussions were recorded and transcribed. Transcriptions were coded for content and organized into themes by five individuals until thematic saturation was achieved.

RESULTS: Three focus groups were conducted with a total of six participants. Five themes emerged: 1) Improvement in overall health, 2) Reduced stress related to FI, 3) Moderating social determinants of health (SDOH) around FI, including financial barriers, 4) Relationships of patients with HFC staff and the healthcare system, 5) Improving the HFC for future participants.

CONCLUSIONS: This qualitative exploration suggests that participant experiences with a healthcare system FI intervention are impactful. The design facilitates meaningful interpersonal relationships and supports chronic disease management. This preliminary data suggests that participants are positively impacted at both the individual and population level, while SDOH related to FI are reduced. This data will help guide HFC delivery of optimal SDOH interventions that address population needs via direct patient engagement.

PATIENTS EXPERIENCING HOMELESSNESS HAVE LONGER HOSPITALIZATIONS AND LESS HOSPITAL FOLLOW-UP—PRELIMINARY RESULTS FROM A NOVEL DASHBOARD

Katarina Leyba¹; Amanda V. Johnson²

¹Internal Medicine, University of Colorado, Denver, CO; ²Division of Hospital Medicine, University of Colorado School of Medicine, Denver, CO. (Control ID #3877335)

BACKGROUND: Individuals experiencing homelessness have significant comorbid physical and mental illness, with a rate of chronic illness and early mortality higher than that of the general population. Individuals experiencing homelessness tend to have longer stays and associated annual costs of hospitalization that are 10% higher than individuals with housing. Overall, individuals experiencing homelessness have poorer health outcomes and increased care utilization, leading to a significant cost and resource burden on the healthcare system and representing a particular challenge for providers and hospitals caring for these patients.

METHODS: In order to track healthcare utilization patterns and identify potential quality improvement targets, we created a novel dashboard to track hospitalizations in patients experiencing homelessness who treated by the Division of Hospital Medicine (DHM) at the University of Colorado. Unhoused patients were identified through documentation across electronic medical record flowsheets. This dashboard captured data from 598 patients experiencing homelessness who were hospitalized at the University of Colorado between November 2021 and November 2022. Metrics of healthcare utilization—including length of stay (LOS), hospital follow-up rates and readmission rates—were compared to metrics of all patients cared for by the DHM.

RESULTS: The average hospital LOS for homeless patients was 5.77 days longer for than that of the general patient population (12.96 days for patients experiencing homelessness versus 7.13 days). Among patients experiencing homelessness, 40.3% had an established primary care physician (PCP), relative to 68.7% among all DHM patients. Similarly, only 76.3% of unhoused patients had a 30-day follow-up appointment scheduled at hospital discharge, whereas the average rate of 30-day follow-up scheduled at hospital discharge was 80.1% amongst all patients. Readmission rates were similar between patients experiencing homelessness and the general patient population.

CONCLUSIONS: Patients experiencing homelessness have a longer hospital LOS, are less likely to be established with a PCP and have fewer follow ups scheduled at the time of discharge relative to the general patient population. Detecting disparities that affect this patient population is essential to identifying targets for interventions to reduce those disparities moving forward.

PATTERNS OF SCREEN DETECTABLE CANCER DIAGNOSIS AMONG INCARCERATED ADULTS IN CONNECTICUT COMPARED TO THE GENERAL POPULATION

Ilana Richman¹; Pamela Soulos¹; Hsiu-Ju Lin²; Jenerius Aminawung¹; Lisa Puglisi¹; Emily Wang¹; Cary Gross¹

¹Department of Medicine, Yale University, New Haven, CT; ²School of Social Work, University of Connecticut, Storrs, CT. (Control ID #3876109)

BACKGROUND: Although incarcerated people have a constitutional right to health care including preventive services, little is known about whether they undergo recommended cancer screenings at rates equivalent to the general population. Patterns of stage at diagnosis can provide insight into whether a population is likely to have been screened. We hypothesized that incarcerated and recently released adults will have a lower incidence of early-stage screen detectable cancers, higher incidence of late-stage screen- detectible cancers, and similar incidence of non-screen detectable cancers compared to community-dwelling adults.

METHODS: We conducted a cross sectional study of cancer incidence in three populations in Connecticut:incarcerated adults, adults recently released from incarceration, and the general adult population between 2005-2015. We categorized cancers as screen-detectable (breast, colorectal, cervical, and prostate) or not. We classified stage at diagnosis as early (stages 1-2) or late (stages 3-4). We calculated the age- and sex- standardized incidence ratio (SIR) for the incarcerated and recently released populations compared to the general population, stratified by screen-detectable category and stage.

RESULTS: Among incarcerated and recently released populations, incidence of early-stage screen- detectable cancers was lower than in the general population (SIR 0.28, 95% CI 0.17-0.43; SIR 0.68, 95% CI 0.51-0.88, respectively). However, compared to the general population, incidence of late-stage screen detectable cancers was lower among incarcerated adults (SIR 0.51, 95% CI 0.27-0.88) and not different among recently released adults (Table). For non-screen-detectable cancers, incarcerated adults had a lower incidence of early-stage cancers (SIR 0.42, 95% CI 0.31-0.56), while recently released adults had a higher incidence of both early and late-stage cancers (SIR 1.27, 95% CI 1.08-1.48; SIR 2.17 95% CI 1.84-2.53, respectively) compared to the general population.

CONCLUSIONS: We found that incidence of early-stage screen detectable cancer was lower among incarcerated and recently released adults compared to the general population, suggesting less access to or use of cancer screening. However, we also observed lower incidence of late-stage screen detectable and early stage non-screen detectable cancers while incarcerated. This pattern suggests that in addition to underscreening, there may be more general under detection of cancer even when symptomatic while incarcerated.

PERSPECTIVES ON BARRIERS AND FACILITATORS OF AMBULATORY CLINICAL ACCESS FOR PATIENTS INSURED WITH MEDICAID IN AN ACADEMIC HEALTH SYSTEM

Jessica Faiz^3,2; Mariah B. Blegen^3,2; Daniel Gonzalez¹; Daniel C. Stokes¹; Kevin Truong¹; Vanessa Nunez¹; Medell Briggs-Malonson⁴; Gery W. Ryan⁵; Katherine L. Kahn¹

¹Internal Medicine, University of California Los Angeles, Los Angeles, CA; ²National Clinician Scholars Program at University of California Los Angeles, Los Angeles, CA; ³Veterans Health Administration, Greater Los Angeles Healthcare System, Los Angeles, CA; ⁴Department of Emergency Medicine, University of California Los Angeles, Los Angeles, CA; ⁵Department of Health Systems Science, Kaiser Permanente Bernard J Tyson School of Medicine, Pasadena, CA. (Control ID #3876567)

BACKGROUND: Substantial variation exists in access to health care for Medicaid-insured patients. This study interviewed staff and providers who work directly with Medicaid-insured patients to explore current barriers to, and facilitators of, access to care for Medicaid-insured individuals. To explore and mitigate access barriers, multi-disciplinary teams with administrative and clinical expertise collaborated to develop a Medicaid Referral Process Map outlining pathways to patient appointments.

METHODS: We used purposive and snowball sampling to identify ambulatory appointment-schedulers, nurses, social workers, and physicians with frequent exposure to patient-specific access problems in a large academic health system. We completed 16 semi-structured, qualitative 30–40-minute interviews via Zoom from March to July 2022. Interviewers prompted informants to describe specific challenges that patients had encountered from initiation to completion of clinical encounters. Using a coding scheme developed from the Medicaid Referral Process Map, detailed interview notes and transcripts were analyzed for barriers and facilitators at the system or individual level.

RESULTS: Informants held a breadth of patient-facing roles including patient navigator, care coordinator, clinical advisor, clinic director, case manager, practice manager, financial counselor, patient service representative, primary care physician, and subspecialty surgeon. Interviewees described challenges and potential solutions for the existing administrative processes including scheduling, authorizing referrals, contracting, and attending the clinical encounter. Three challenges to access were mentioned most frequently: (1) lack of awareness of formal protocols (2) restrictive policies by managed care Medicaid networks (3) patient-level communication issues. Key informants identified potential solutions for these barriers, including the need for (1) system-level policies to specifically address referral authorization barriers; (2) increased provider-level knowledge of insurance policies; and (3) consistent strategies for communicating scheduling information and declined authorizations to patients.

CONCLUSIONS: Key gaps remain in accessing care for patients in some clinics. Although this work focused on Medicaid-insured patients, our study identified barriers that are also relevant for the larger ambulatory care patient population, especially vulnerable groups such as those with low socioeconomic status or limited English language proficiency. Next steps include soliciting patient perspectives and generating consensus on improvement priorities and strategies to enhance access to ambulatory clinics by patients with Medicaid.

PHYSICAL AND MENTAL HEALTH CONDITIONS, ACCESS TO CARE, AND FINANCIAL BARRIERS TO CARE AMONG PEOPLE INCARCERATED IN US PRISONS

Emily L. Lupez^1,2; Laura Hawks³; Stephanie Woolhandler^8,5; David Himmelstein^8,5; Samuel Dickman⁷; Adam Gaffney^2,1; David Bor^2,1; Elizabeth Schrier⁶; Christopher Cai⁴; Danny McCormick^2,1

¹Harvard Medical School, Boston, MA; ²Internal Medicine, Cambridge Health Alliance, Cambridge, MA; ³General Internal Medicine, Medical College of Wisconsin, Wauwautosa, WI; ⁴Internal medicine, Brigham and Women's Hospital, Boston, MA; ⁵Medicine, Cambridge Health Alliance / Harvard Medical School, Cambridge, MA; ⁶University of California San Francisco School of Medicine, San Francisco, CA; ⁷Planned Parenthood of Montana Inc, Billings, MT; ⁸City University of New York at Hunter College, New York, NY. (Control ID #3867482)

BACKGROUND: A 2004 survey in US prisons found high rates of physical and mental health conditions and poor access to care among people incarcerated, prompting calls to improve the quality and availability of medical care for this population. Since then, little has been published on current health conditions, access to care, or the effects of copayments on such access.

METHODS: We analyzed the recently released 2016 Survey of Prison Inmates (SPI), the only nationally representative health survey of people incarcerated in US state and federal prisons. We calculated prevalence of self-reported chronic physical conditions (CC), mental health conditions (MHC), and current severe psychological distress (SPD, score >13/24 on the Kessler Psychological Distress Scale). We assessed several self-reported measures of access to care among persons with CC, MHC, or SPD. Using facility jurisdiction, state of incarceration, and published data on federal and state prison wages and copayments for prison healthcare visits, we assessed the association between copayments and receipt of a medical visit using risk ratios. Analyses were conducted for the overall US prison population and by facility jurisdiction using SPI- provided weights to generate national estimates and statistical procedures appropriate for the complex survey design.

RESULTS: Of 24,848 individuals surveyed (weighted population 1,421,700; 1,248,300 in state prisons, 173,400 in federal) 62.4% of state (56.3% of federal) incarcerated people had >1 CC (weighted estimates). Among those with >1 CC incarcerated within the last year, 22.3% in state (14.9% in federal) prisons had not seen a healthcare provider since admission. Of people with >1 CC, 91% had a medical copay and 47.4% had a medical copay >1 week’s prison wage. Compared to those with no copay, those with any copay (relative risk [95% CI]: 1.47 [1.4, 1.5]), and those with a copay >1 week’s prison wage (1.72 [1.7, 1.8]) were more likely to have not seen a healthcare provider since incarceration. The prevalence of having >1 MHC was 42.5% in state (22.9% in federal) prisons, and the prevalence of SPD was 14.1% in state (7.8% in federal) prisons. Of those with SPD, 40.5% in state (57.7% in federal) prisons had not received any mental health treatment since admission. Among those on a prescription medication for a MHC prior to incarceration, 33.6% in state (42.8% in federal) prisons were not receiving medication in prison.

CONCLUSIONS: CC and MHC remain highly prevalent among those in US prisons. Despite their unique legal right to healthcare, many incarcerated people fail to receive adequate care, and medical copayments appear to magnify this risk. In addition to improving poor carceral environmental conditions that foster CC and MHC, US prison healthcare system planning should account for incarcerated people’s high burden of CC and MHC, and ensure access to care. Elimination of copayments in US prisons could improve access to care and help fulfill incarcerated people’s legal right to healthcare.

PHYSICIAN PERSPECTIVES ON INCARCERATED INPATIENTS

Justin Vincent²; Xiaofan Huang³; Kristen Staggers³; Marc D. Robinson¹

¹Medicine, Baylor College of Medicine, Houston, TX; ²Washington University in St Louis School of Medicine, St Louis, MO; ³Institute for Clinical and Translational Medicine, Baylor College of Medicine, Houston, TX. (Control ID #3876299)

BACKGROUND: On any given day, the United States incarcerates greater than 2 million individuals, more than any other country by total numbers or on a per capita basis. Incarcerated patients are at risk for interrupted chronic care and report higher rates of some conditions such as diabetes, hypertension, asthma, and psychiatric illness. Despite data existing for patient risk factors and worsened outcomes, little is known about physician attitudes towards patients who are incarcerated.

METHODS: An electronic survey was created and distributed to a total of 70 internal medicine residents (PGY2 - PGY4) and 26 academic hospitalists who rotate through a large county hospital that cares for patients from the local jail. Likert scale answers were dichotomized as either 1) Agree or 2) Disagree/Neutral. Summary statistics were stratified by training level and close association with the criminal justice system were compared using Wilcoxon rank-sum test or Fisher’s exact test. A significance level of 0.05 was used.

RESULTS: Overall, responses between hospitalists and residents were similar, except for a few key areas. For the most part, both groups disagreed with the statement “Given the choice, I prefer not to take care of incarcerated patients” (n = 85, 89%). Both also did not believe that jailed patients fake illness more than non- jailed patients (n = 72, 75%) and did not believe their exams for these patients were briefer (n = 65, 68%). Residents and hospitalists disagreed that shackling (use of leg or handcuffs to the bed) was necessary all of the time (n = 79, 82%). Still, both rarely remove shackles while examining incarcerated patients (median percentage of time removed 15% vs. 10%, p = 0.44).

Hospitalists were more likely to know when their patients were likely to be out of custody (19% vs. 3%, p = 0.02), be aware of the services the jail has to offer (31% vs. 10%, p = 0.02), and less likely to report difficulties obtaining follow-up (58% vs. 84%, p = 0.01). Residents reported a greater propensity to consciously or subconsciously treat an incarcerated patient differently (70% vs. 38%, p = 0.01).

Those with a close association with the criminal justice system more often agreed that patients from jail were more likely fake illness (47% vs. 21%, p = 0.03) and were more likely to correctly estimate the number of days incarcerated people typically stay in county jail (47% vs. 18%, p = 0.02).

CONCLUSIONS: Overall, we find that physicians desire to care for this vulnerable population but admit potential biases and gaps in knowledge regarding the nuances of their care. Training level often As millions of adults are currently in jails and prisons, these results speak to the need for further education and training in caring for this vulnerable population.

PRACTICE VARIATIONS IN FOOD INSECURITY SCREENING: ONE SIZE DOES NOT FIT ALL Carter Baughman, Gianna Aliberti, Son Quyen H. Dinh, Elizabeth Targan, Ruth-Alma N. Turkson-Ocran, Katherine Wrenn, Kelly Graham

Internal Medicine, Beth Israel Deaconess Medical Center, Boston, MA. (Control ID #3871969)

BACKGROUND: Screening for and addressing the social determinants of health (SDOH), including food insecurity (FI), is essential to the provision of high-quality medical care. FI impacts outcomes for a variety of chronic disease states, including diabetes, hypertension, and cardiovascular disease risk, with disproportionate effects on minoritized and low-income populations. To address these healthcare inequities, the Centers for Medicare and Medicaid Services encouraged states to implement a comprehensive assessment of the SDOH. Massachusetts, where our practice is located, is now implementing this screening protocol, however little is known about how best to incorporate this into clinical practice.

METHODS: We conducted semi-structured qualitative interviews with a convenience sample of adult primary care providers and community health workers (CHWs) in Boston, MA to better understand their screening practices for FI. We gathered information on practice demographics including provider and patient characteristics. We asked about the process of screening, including who screens, where, when, and how often, and asked practices to specify what form they use. Finally, we inquired about the perceived benefits and challenges to the screening protocol they developed and their recommendations for screening to inform practices beginning this implementation.

RESULTS: We interviewed leadership at 6 unique practices, including both academic and community health centers. Five sites (83%) screened for FI within the context of SDOH screening. The most common time to screen was at rooming (50%). The most common method of screening was a paper form using the PRAPARE tool (50%). The screen was most often input into the electronic health record (EHR) via medical assistants (66%). The most common frequency of screening was at annual visits (100%), with two sites (33%) having additional outreach for overdue screens. The most common method of CHW notification was an EHR notification (50%) followed by a physician referral (33%). All screening models ended with the patient being connected to a designated staff member(s) with expertise in food and nutrition resources, with varying numbers of steps to get there. The most common benefits of screening expressed were the ability to reach all patients, having adequate resources available for patients, and having streamlined and automated processes. The most common challenges were related to staffing and CHW task load, survey limitations (e.g., survey fatigue, lack of language options), and patients not accessing resources due to complex processes or not wanting to report FI.

CONCLUSIONS: SDOH and FI screening is essential for all primary care practices and is becoming a requirement for many states. However, it can be difficult to incorporate this screening protocol due to high workload burdens and lack of standardized processes. We evaluated different models that can be applied to various practices and found an overall benefit to streamlined processes.

PREDICTORS OF PERCEIVED HEALTH DISPARITIES IN CHRONIC ILLNESS AMONG PATIENTS OF AFRICAN ANCESTRY

David Christian³; Kelly Z. Wang³; Nihal Mohamed³; Mimsie Robinson^5,6; Dinushika Mohottige¹; Thomas A. LaVeist⁴; Carol Horowitz²

¹Institute of Health Equity Research; DOM and the Barbara T Murphy Division of Nephrology, Icahn School of Medicine at Mount Sinai Department of Population Health Science and Policy, New York, NY; ²General Internal Medicine, Icahn School of Medicine at Mount Sinai, New York, NY; ³Icahn School of Medicine at Mount Sinai, New York, NY; ⁴School of Public Health and Tropical Medicine, Tulane University, New Orleans, LA; ⁵Beth Gospel Assembly, New York, NY; ⁶BETH-HARK CHRISTIAN COUNSELING CENTER INC, New York, NY. (Control ID #3876838)

BACKGROUND: Racial disparities in health are ubiquitous and result from racism-related inequities in patients’ socio-environmental contexts and health care. Genetic risks including APOl1 high-risk alleles, which have been associated with self-reported race/ancestry also explain some racial disparities. Yet, little is known regarding patients’ beliefs about of the existence of racial health disparities and their key causes.

METHODS: We conducted a cross-sectional analysis of baseline survey data from 1850 adults with hypertension who self-identified as Black/having African ancestry upon enrollment in the GUARD trial designed to determine effects of testing and disclosing renal genetic risks. The study was conducted in 2014-2018 across 15 ambulatory practices and federally qualified health centers in New York City. Participants were asked whether they believe Black/African American people are more likely to have chronic disease than White people (yes/no), and whether the racial disparities are caused by different genes, behaviors/culture, stress, and/or the environment. We describe the association between perceived racism in health care and belief in racial disparities, and we quantified the association between perceived causes of disparities and belief in racial disparities using logistic regression models adjusted for age, gender, income, and education.

RESULTS: Most (74%) participants perceived Black people are more likely to have chronic diseases than White people. They attribute this difference to genetics (50%), behaviors/culture (51%), stress (59%) and their environment (66%) (where they live, pollution, access to care). Adjusted odds of perceiving racial health disparities (versus not) were highest for those who believed differences are due to genes [OR=3.87, 95% CI=2.97 – 5.04]. There was no significant association between perceived racism in healthcare and a belief in racial disparities. A significant proportion (25%) of respondents believed doctors would view them more positively if their disease was caused by genetic factors.

CONCLUSIONS: A majority of Black adults perceived multi-factorial causes of racial disparities in chronic diseases including genetic and environmental factors. Perceived racism was not associated with perceived racial disparities in this study. However, nearly a quarter of participants believed their physicians would view them more favorably if their disease was due to genetic causes. Future studies are needed to further examine patients’ perceptions of the relative role of genetics vs. socio-contextual factors and racism in causing racial disparities. Studies designed to examine patient and physician perceptions and knowledge about of genetic vs. socio-contextual contributions to disease risk are also warranted.

PREFERRED COMMUNICATION STRATEGIES FOR PEOPLE WITH COMMUNICATION DISORDERS IN HEALTH CARE ENCOUNTERS

Erin Hickey¹; Bernice Man²; Kaila Helm⁴; Steven Lockhart³; Megan Morris³

¹Internal Medicine - Pediatrics (Med-Peds), University of Illinois Chicago College of Medicine, Chicago, IL; ²Academic Internal Medicine, University of Illinois Hospital & Health Sciences System, Chicago, IL; ³University of Colorado Anschutz Medical Campus, Aurora, CO; ⁴Icahn School of Medicine at Mount Sinai, New York, NY. (Control ID #3874090)

BACKGROUND: In the United States, 10% of all adults identify as having a Communication Disorder (CD) which includes disabilities in speech, language, voice and/or hearing. Previous research has demonstrated that people with CDs have worse health outcomes, including more difficulties accessing high quality, equitable care compared to those who do not have CDs. Part of this difficulty is due to health care providers not being aware of patients’ CDs and not knowing what communication

strategies people with CDs prefer. We set out to answer the following research question: What communication strategies do people with CDs, their caregivers, and their providers prefer, and how do they want these strategies communicated bidirectionally?

METHODS: We performed a qualitative study utilizing focus groups and interviews with patients who have CDs, their caregivers and healthcare providers over video-conference. Each patient focus group represented a different CD, including stuttering, aphasia, hearing loss, and patients who are nonverbal. Participants were asked about how CDs affect their experiences in healthcare and what strategies are helpful in these encounters. Each focus group or interview was recorded and transcribed. The transcript was then independently coded by two researchers with inductive analysis using ATLAS. After a coding dictionary was created multiple rounds of coding occurred to reach calibration, and the entire team reconciled any discrepancies. After reviewing query summaries, themes were developed and honed over multiple team meetings.

RESULTS: Identified themes are (1) Both patients and providers preferred to have the communication disability and communication preferences disclosed prior to the appointment, (2) Preferred communication strategies should be individualized, but when this is not possible, there are some best practices for communicating with people with CDs that can be implemented and (3) There are systems level and provider level changes that can help improve the care of people with CDs

CONCLUSIONS: Patients with CDs prefer that providers are aware of their CD and know what strategies they prefer to optimize communication. We hope that the implementation of best practices for patient-provider communication will lead to better health outcomes for patients with CDs.

PREVALENCE OF HOMELESSNESS AMONG HOSPITALIZED PATIENTS: A POINT-IN-TIME SURVEY

Neelam Mistry²; Julie Knoeckel¹; Amanda V. Johnson³; Erin Bredenberg⁴; Katie Raffel³; John Cunningham¹; Ellen Sarcone¹; Gregory Misky³; Lauren McBeth⁵; Sarah A. Stella⁶

¹Hospital Medicine, Denver Health, Denver, CO; ²Department of Medicine, University of Colorado, Denver, CO; ³Division of Hospital Medicine, University of Colorado School of Medicine, Denver, CO; ⁴Hospital Medicine, University of Colorado Denver School of Medicine, Aurora, CO; ⁵Medicine/Hospital Medicine, University of Colorado, Aurora, CO; ⁶Department of Medicine, Denver Health, Denver, CO. (Control ID #3874665)

BACKGROUND: Lack of stable housing adversely affects health outcomes, increases healthcare utilization, and increases mortality. Few studies have assessed the prevalence of homelessness and housing insecurity in hospitalized patients. The United States Department of Housing and Urban Development requires an annual point-in-time count of people experiencing homelessness (PEH) on a single night in January. This count excludes PEH who are hospitalized. To address these gaps, we conducted a hospital point-in-time study to determine the prevalence of housing insecurity and homelessness among hospitalized patients at two urban academic centers.

METHODS: Medical-surgical patients admitted to two academic hospitals on January 24, 2022 were eligible to participate. Patients were included if they were > 18 years old, available in their hospital room and able to consent. Researchers administered a survey to determine patients' housing status: stably housed, housing insecure (concerned about losing housing within 2 months or spending > 50% of income on housing), or frankly homeless (living on the street, in a shelter, or other place not intended for housing). Additional medical and sociodemographic characteristics were assessed for patients reporting housing insecurity or homelessness.

RESULTS: A total of 456 patients were approached across the two study sites and 271 consented to participate. 29.2% of these participants were found to be experiencing housing insecurity (17.3%) or frank homelessness (11.8%). 65.8% of the patients with significant housing concerns were male, and 60.8% were 55 years or older. Among surveyed patients, 69.6% reported chronic health conditions, 55.7% reported multiple hospitalizations in the preceding year and 49.4% reported a physical disability. More than a third (35.4%) had previously been in jail or prison. 38% reported a mental health concern and 39.2% reported drug or alcohol use.

Of the 47 patients experiencing housing insecurity, 57.4% had experienced homelessness in the past. The most frequently identified factors that contributed to housing insecurity were cost of living (72.3%), health conditions (70.2%), and loss of income (46.8%).

Of the 32 patients experiencing homelessness, 28.1% reporting using a shelter directly prior to admission. Of those not using a shelter the top reasons cited were safety concerns (34.4%), cleanliness (18.8%), and regulations at the shelter (12.5%).

CONCLUSIONS: Almost a third of hospitalized patients in this study were experiencing homelessness or housing insecurity. Those experiencing homelessness were not accounted for in annual community point-in- time counts, contributing to an underestimate of the true housing crisis. There is a critical need to accurately capture housing status during hospitalization and develop tailored support systems for this population characterized by high rates of physical disability, legal involvement, mental health concerns, and substance use, and low rates of shelter usage.

PREVALENCE OF STIGMATIZING LANGUAGE REFLECTING SPECIFIC TYPES OF BIAS IN MEDICAL RECORDS ACROSS 4 SPECIALTIES

Mary Catherine Beach, Keith Harrigian, Mark Dredze, Dingfen Han, Brant Chee, Ayah Zirikly, Alya Ahmad, Anne Links, Nae-Yuh Wang, Shari Lawson, Somnath Saha

Johns Hopkins University, Baltimore, MD. (Control ID #3877145)

BACKGROUND: Previous studies have documented racial bias in the use of stigmatizing language in medical records. We aimed to build on this literature by evaluating racial differences in physicians’ use of positive adjectives to characterize patients, and by distinguishing racial differences in patterns of negative language reflecting distinct types of bias.

METHODS: Based on previous literature, we developed a taxonomy of positive and negative language used to describe patients within medical records: positive descriptors (e.g. pleasant, charming, delightful), negative descriptors (e.g. belligerent, aggressive, uncooperative), and disbelief/dismissal of patient’s testimony (e.g. claims, insists, poor historian). We searched ambulatory and inpatient notes within our health system’s electronic medical record system across 4 clinical disciplines – internal medicine, emergency medicine, OB/GYN, and surgery – to identify use of terms. Two reviewers annotated a sample of 2440 notes containing these terms to differentiate references that were in-context versus out-of-context (e.g. insurance claims, aggressive tumor). These annotated notes were then used to develop machine learning algorithms that classified term usage with 96% accuracy. We then used these algorithms to classify notes within a large corpus, based on appearance of the relevant terms. We tested associations of patient race with each category of language, using logistic regression to adjust for age and gender, and generalized estimating equations to account for clustering of notes within patients and clinicians.

RESULTS: Our analysis included 109,633 notes in the records of 19,681 patients who had a mean age of 50, and of whom 56% were Black, 37% white and 64% women. Within notes, 13.0% had a positive descriptor, 4.3% had a negative descriptor, and 1.6% had a marker of disbelief. Compared with non-Black patients, notes written in the records of Black patients had a lower odds of containing a positive descriptor (OR 0.87, 95% CI 0.83-0.92), similar odds of a negative descriptor (OR 1.03, 95% CI 0.95-1.12), and a higher odds of indicating disbelief (OR 1.50, 95% CI 1.33-1.70). These patterns were similar across the 4 specialties.

CONCLUSIONS: Our study demonstrates that bias in medical record documentation across specialties may manifest more in the use of positive descriptors and markers of disbelief than in the use of overt negative descriptors. Reducing use of negative descriptors is important to improve the quality of care for all patients, and efforts to promote racial equity in healthcare should also focus on raising awareness of specific biases in who gets positively appraised and whose testimony gets dismissed.

PROVIDER PERSPECTIVES ON HOW SOCIAL DETERMINANTS INFLUENCE HEPATITIS C CARE

Christopher J. Gonzalez¹; Shashi N. Kapadia²; Stephane Labossiere²; Jeff Niederdeppe³; Andrew Talal⁴; Martin F. Shapiro²; Elaine Wethington³

¹Internal Medicine, Weill Cornell Medicine, New York, NY; ²Medicine, Weill Cornell Medicine, New York, NY; ³Cornell University, Ithaca, NY; ⁴Medicine, University at Buffalo, Buffalo, NY. (Control ID #3876125)

BACKGROUND: Hepatitis C virus infection (HCV) continues to disproportionately affect racially minoritized and socioeconomically disadvantaged communities. Disparities in curative treatment uptake may be driven by social determinants, including the behavior of healthcare providers. Our objective was to describe provider perspectives on how social determinants influence HCV treatment uptake and provision.

METHODS: Using purposive sampling between 9/2021 and 9/2022, we recruited healthcare providers in two states with divergent health care systems and policies, Alabama and New York. We sampled for diversity in provider specialty, training, and region of practice. We conducted interviews remotely using a semi- structured interview guide informed by Andersen’s model of health services utilization, prior literature, and expert experience. Interviews were professionally transcribed, and transcripts were analyzed using NVivo software. Each transcript was independently coded by two researchers, then reconciled using a summative approach. Similar codes were combined into categories, which were then consolidated into broader themes.

RESULTS: Nearly half of the 34 providers practiced Primary Care (47%); the remaining practicing providers specialized in Infectious Diseases (26%) and Gastroenterology (12%). The majority (67%) held an MD or DO, whereas 33% were advanced practice professionals (e.g. RN, NP, PA). Four major themes emerged: (1) Numerous sociodemographic factors influence patients’ capacity to initiate HCV treatment- “One of the most important barriers to treatment is lack of shelter, stable homes, or support systems, friends, because of substance abuse.” (2) Providers face their own challenges with providing treatment- “Some people don't recognize that we should treat hepatitis C early and defer until they have very advanced liver disease.” (3) Providers advocate for patient-centered approaches, feeling provider perceptions could create additional barriers to HCV treatment- “Often times, there is a certain negative connotation that these are difficult patients to treat and some providers won't treat them.” (4) Organizational, insurance, and public policies continue to influence access to treatment- “With Medicaid, if your drug screening is not cleared, including cannabinoids, then they won't pay for it.”

CONCLUSIONS: HCV treatment initiation is influenced by the social determinants of health at several levels. Patient sociodemographic factors influence the capacity to initiate treatment, and provider factors create challenges to providing treatment. Many providers advocate for accessible care and treatment, acknowledging that provider practices and larger overarching policies often create additional barriers for patients living with HCV. Overcoming these complex barriers may require multi-faceted interventions targeting multiple levels. Further work will examine how the influence of social determinants vary across specific patient groups and geographic regions.

QUANTIFYING THE IMPACT OF NIGHTTIME GUNSHOTS ON COMMUNITIES IN MAJOR US CITIES, 2015 TO 2021

Rebecca Robbins^1,6; Mahmoud Affouf³; Peter Masiakos^2,5; Jay Iyer⁴; Cornelia Griggs^2,5; Elizabeth B. Klerman^7,1; Chana A. Sacks^8,5

¹Division of Sleep Medicine, Harvard Medical School, Boston, MA; ²Pediatric Surgery, MGH, Boston, MA; ³Mathematics, Kean University, Union, NJ; ⁴Departments of Molecular and Cellular Biology and Statistics, Harvard College, Cambridge, MA; ⁵Harvard Medical School, Boston, MA; ⁶Division of Sleep and Circadian Disorders, Brigham and Women's Hospital, Boston, MA; ⁷Department of Neurology, Massachusetts General Hospital, Boston, MA; ⁸Medicine, Massachusetts General Hospital, Boston, MA. (Control ID #3876933)

BACKGROUND: Gunshots affect not only those directly involved in the incident but also those in the surrounding community exposed to the noise of gunshots and the subsequent emergency response. Nighttime gunshots may interrupt and/or delay sleep of nearby community members.

METHODS: We used publicly available cross-sectional datasets detailing timing and location of gunshots in 6 US cities (Baltimore MD; Boston MA; Washington DC; New York NY; Philadelphia PA; and Portland OR) between 2015 and 2021. Datasets from Baltimore, Philadelphia, New York City and Portland reported only gunshots with victims; datasets from Boston reported gunshots both with and without victims; Washington DC provided data from gunshot detection surveillance technology (victim status unknown). We computed Rate Ratios (RR) to compare nighttime (6:00pm-5:59am) to daytime (6:00am- 5:59pm) gunshots. We use geospatial mapping to estimate the number of person-nights affected by the noise of the gunshots each year and the relationship between nighttime gunshots and median household income at the location of each gunshot.

RESULTS: We analyzed 72,236 gunshots. Gunshots were more common during the nighttime than daytime (RR=2.5). Geospatial analysis demonstrated that the average number of annual person-nights impacted by nighttime gunshots were 0.35 million in Baltimore, 1.62 million in Boston, 5.91 million in Washington, 2.9 million in New York City, 1.29 million in Philadelphia, and 0.40 in Portland. There was an inverse relationship between nighttime gunshots and median household income in each city.

CONCLUSIONS: Nighttime gunshots are prevalent and predominate in locations characterized by low median household income.

RACE-BASED MEDICINE IN MEDICAL TRAINING: ELICITING JUSTIFICATIONS PROVIDED IN THE FORMAL AND INFORMAL CURRICULUM

Jorge A. Caceda¹; Kessy Joseph²; Khalil Savary¹; Ping-Hsin Chen¹; Rachel Rosenberg¹

¹NJMS, Rutgers Biomedical and Health Sciences, Newark, NJ; ²Newark Beth Israel Medical Center, Newark, NJ. (Control ID #3874060)

BACKGROUND: Race-based medicine and race-based correction factors pervade many areas of medicine despite clear evidence that race is a social, not biological, construct. It is unknown what the impacts of race- based medicine are on learners, or how race-based medicine may play into learners’ implicit and explicit biases. It is also unclear what learners may be learning about race-based medicine through the implicit or hidden curriculum. This study seeks to elucidate what learners are being taught in medical school about race- based medicine, particularly through the informal curriculum.

METHODS: We utilized a cross-sectional survey to ask medical residents open-ended questions about what they were taught about the reasoning behind race-based medicine during medical school. For example, one question asked participants to describe the explanations they received in medical school to explain race-based calculations of glomerular filtration rate and identify the resource from where they remember obtaining this information. Participants’ answers to open-ended questions were coded using qualitative data analysis methods. The survey was anonymous and open to residents from all specialties. Residents were recruited through emails and social media. The study was approved by the Rutgers Institutional Review Board. RESULTS: Fifty-four resident physicians participated in the study. Participants were asked to self-identify their race/ethnicity; 16 unique self-descriptors were used. Four main themes emerged regarding the teaching that residents received about race-based medicine in medical school: (1) Race as biology (2) Medical heuristics (3) Othering Black people (4) Left to their own conclusions.

CONCLUSIONS: The explanations that learners are given for race-based medicine and race-based correction factors reinforce the incorrect concept of race as biology. Furthermore, learners are encouraged to use race to make short-cuts in decision-making. Our study suggests that exposure to race-based medicine, and the justifications provided for it through the informal curriculum, could deepen biases among medical students and worsen disparities in medical care.

RACIAL AND ETHNIC DISPARITIES IN OPIOID PRESCRIBING ON HOSPITAL DISCHARGE AMONG OLDER ADULTS WITH CANCER

Koushik Kasanagottu^1,2; Kenneth J. Mukamal¹; Bruce E. Landon³

¹General Internal Medicine, Beth Israel Deaconess Medical Center, Boston, MA; ²Harvard Medical School, Boston, MA; ³Health Care Policy, Harvard Medical School, Boston, MA. (Control ID #3874920)

BACKGROUND: Disparities in opioid prescribing among racial and ethnic groups have been observed in outpatient and emergency department settings. However, it is unknown whether similar disparities exist at discharge among older adults hospitalized for cancer, for whom opioids are recognized as an effective treatment for pain.

METHODS: We performed a retrospective cohort study of Medicare beneficiaries age 65 years or older who were hospitalized in 2016 with a primary discharge diagnosis of cancer. We excluded beneficiaries with hospice claims, and those admitted from or discharged to a skilled nursing facility (SNF), owing to inability to capture medication claims in these patients. Our primary outcome was presence of an opioid claim within 2 days of hospital discharge. Secondary outcomes among patients discharged with an opioid included total morphine milligram equivalents (MME) and quantity of opioids prescribed. To determine the adjusted relative risk of each outcome by race/ethnicity, grouped as Black, White, Asian, Hispanic, and other (native Hawaiian/American Indian/Alaska Native/other), we fitted multivariable generalized estimating equation models accounting for demographics, disability status, Medicaid dual-eligibility, presence of an opioid claim in the 90 days prior to hospitalization, comorbidities, area deprivation index, hospitalization characteristics, primary diagnosis/procedure codes, and hospital characteristics.

RESULTS: There were 13,095 hospitalizations with cancer as the primary discharge diagnosis. Among these encounters, 82.5% were White, 7.8% were Black, 4.4% were Hispanic, 2.8% were Asian, and 2.6% were Other. Overall, 5,343 (40.8%) had a claim for an opioid within 2 days of hospital discharge: 41.4% of White, 37.5% of Black, 34.6% of Asian, 38.1% of Hispanic, and 46.4% of other beneficiaries (p=<0.0001). After adjustment, there were no significant differences in filled opioids by race/ethnicity. However, among beneficiaries who filled an opioid prescription, Hispanic beneficiaries filled lower total MMEs when compared to White beneficiaries (mean adjusted relative difference [RD] -10%, 95% CI -17% to -2%).

CONCLUSIONS: In a national cohort of Medicare beneficiaries hospitalized with cancer, we did not identify racial/ethnic disparities in the proportion of patients with a filled opioid prescription after discharge. However, when prescribed opioids, Hispanic beneficiaries received less total MMEs when compared to White beneficiaries. Given demonstrated disparities in other settings and non-cancer patient populations, these findings suggest that future research should examine whether perceived strength of the indication for opioid prescribing may attenuate opioid prescribing disparities.

RACIAL BIAS IN DOCUMENTATION OF PATIENT ADHERENCE IN MEDICAL RECORDS Somnath Saha³; Keith Harrigian²; Mark Dredze²; Dingfen Han³; Brant Chee²; Ayah Zirikly¹; Anne Links²; Alya Ahmad⁴; Shari Lawson⁵; Nae-Yuh Wang³; Mary Catherine Beach³

¹Medicine, Johns Hopkins Medicine, Baltimore, MD; ²Johns Hopkins University, Baltimore, MD; ³General Internal Medicine , Johns Hopkins University, Baltimore, MD; ⁴General Internal Medicine, Johns Hopkins Medicine, Baltimore, MD; ⁵Johns Hopkins Medicine, Baltimore, MD. (Control ID #3877121)

BACKGROUND: Prior studies examining language in medical records have found that clinicians more commonly describe Black patients as nonadherent or not accepting recommended treatment compared to White patients. The degree to which this finding reflects racial differences in patient adherence vs. racial bias among clinicians is unclear.

METHODS: We searched ambulatory and inpatient notes within our health system’s electronic medical record system across 4 clinical disciplines – internal medicine, emergency medicine, OB/GYN, and surgery – to identify use of terms referring to patient adherence to recommended care, including all forms of the words (non)adherence, (non)compliance, refuse, and decline. Two reviewers annotated a sample of 2177 notes containing these terms to differentiate references to patient adherence as positive (e.g., adherent) vs. negative (e.g., not adherent, refuses), and to exclude uses of these terms with other meanings (e.g., adherent bandage). These annotated notes were used to develop machine learning algorithms that classified term usage with 96% accuracy. We then used these algorithms to classify notes within a large corpus, based on presence or absence of the relevant terms. We tested associations of patient race with positive and negative references to adherence, using logistic regression to adjust for age and gender, and generalized estimating equations to account for clustering of notes within patients and clinicians. We also tested for racial differences in the use of what we considered more vs less paternalistic terms (compliant vs. adherent, refuse vs. decline).

RESULTS: Our analysis included 109,633 notes in the records of 19,681 patients, of whom 56% were Black and 64% were women, with a mean age of 50. Positive and negative references to patient adherence appeared in 3.7% and 14.8% of notes, respectively. The notes of Black patients were more likely than those of non- Black (predominantly White) patients to include both positive (OR 1.98, 95% CI 1.83-2.16) and negative (OR 1.46, 95% CI 1.39-1.53) references to patient adherence. This disparity was greater when examining the use of more paternalistic (OR 1.74, 95% CI 1.64-1.85) vs. less paternalistic language (OR 1.21, 95% CI 1.14-1.28).

CONCLUSIONS: We found that clinicians comment on both adherence and nonadherence more commonly among Black vs. non-Black patients. Our findings indicate that clinicians are more focused on ascertaining adherence with Black patients, and use more paternalistic language when doing so, suggesting racial bias in clinicians’ assumptions about and attitudes towards patient adherence. The embedding of this bias in patients’ medical records contributes to structural racism within health systems and may negatively impact the care of Black patients.

RACIAL DIFFERENCES IN LOW-VALUE CARE AMONG MEDICARE BENEFICIARIES IN HEALTH SYSTEMS

Ishani Ganguli¹; Matthew B. Mackwood²; Ching-Wen W. Yang²; Maia Crawford²; Kathleen L. Mulligan¹; A. James O'Malley²; Elliott S. Fisher²; Nancy E. Morden²

¹Division of General Medicine and Primary Care, Brigham and Women's Hospital, Chestnut Hill, MA; ²Dartmouth Institute for Health Policy and Clinical Practice, Lebanon, NH. (Control ID #3876398)

BACKGROUND: Black patients are less likely than white patients to receive high-value care, but evidence on racial differences in low-value care is scant and conflicting; the role of health care systems in these differences is unknown. We characterized racial differences in receipt of low-value services overall and within health systems among system-attributed Medicare beneficiaries.

METHODS: This was a retrospective cohort study using 100% Medicare Fee-For-Service administrative data (2016-2018) for Black and white Medicare beneficiaries aged 65 or older as of 2016 and attributed to 595 US health systems. We used established, claims-based definitions to measure receipt of 40 low-value services among Black and white beneficiaries. For each service, we built linear probability models to estimate Black-white differences in receipt with and without adjustment for patient age, sex, and prior healthcare use. In additional models, we included 1) health system fixed effects to assess racial differences within health systems, and separately, 2) racial composition of the health system’s population to assess the relative contributions of individual race and health system racial composition in low-value care receipt.

RESULTS: Our cohort included 9,833,304 beneficiaries (6.8% Black; 57.9% female). Compared to white beneficiaries, Black beneficiaries had higher adjusted receipt of 11 of 40 low-value services and lower receipt of 20 services. Specifically, Black beneficiaries were more likely to receive low-value acute diagnostic tests, including imaging for headache (6.9% of eligible beneficiaries vs 3.2%) and head CT scans for dizziness (3.1% vs 1.9%). Black beneficiaries had lower rates of low-value screening tests including preoperative laboratory tests (6.5% vs 10.3%), screening electrocardiograms (1.7% vs. 5.1%), and prostate specific antigen tests in men >70-years-old (25.6% vs. 31.0%) (all p<.001). The direction and magnitude of these differences were similar in models including system fixed effects or racial composition.

CONCLUSIONS: Black beneficiaries were more likely to receive low-value acute diagnostic tests and less likely to receive low-value screening tests. Differences were largely due to differential care within health systems rather than to Black and white beneficiaries receiving care in different systems. These patterns suggest potential individual, interpersonal, and structural factors that researchers, policymakers, and health system leaders might further investigate and address to improve care quality and equity.

RACIAL DISPARITIES IN HOSPITAL-ASSOCIATED DISABILITY AND PHYSICAL THERAPY REFERRAL

Maylyn Martinez¹; Matthew Cerasale¹; Mahnoor Baig¹; Marla Robinson²; Andrew W. Schram¹; Monica E. Peek¹; Vineet M. Arora¹; David O. Meltzer²

¹Medicine, University of Chicago Division of the Biological Sciences, Chicago, IL; ²Medicine, The University of Chicago, Chicago, IL. (Control ID #3876936)

BACKGROUND: Hospital-associated disability (HAD) is the new loss of ability to complete one or more activities of daily living (ADLs) without assistance after hospital discharge and occurs in 31% of hospitalized patients. HAD and low mobility during hospitalization are associated with higher risk of readmission, permanent disability, new institutionalization, and death. Black patients face disadvantages with respect to rehabilitation services and mobility outcomes in outpatient and surgical settings. However, little is known about HAD and PT referrals for Black patients hospitalized for acute medical illness. We aimed to assess racial differences in HAD and PT referral in patients hospitalized for medical illness.

METHODS: This was a cross-sectional study of patients admitted to medicine services 10/2018 – 3/2021.

HAD was defined as admission AMPAC score of > 18 with discharge AMPAC score ≤ 18. Change in mobility (HAD surrogate) was calculated as discharge AM-PAC score – admission AM-PAC score. Differences in HAD, functional impairment, and PT referral by race were analyzed by chi squared test. Differences in mobility change by race was assessed by two-sample T test. Logistic regression was used to analyze association of Black race with referral to PT, controlling for age, sex, baseline mobility, mobility change, and length-of-stay. Linear regression was used to analyze the association of Black race with mobility change controlling for age, gender, baseline mobility, and length-of-stay. Subgroup analyses were stratified by age (<65 and 65+) and baseline mobility (AM-PAC >18 vs. ≤18).

RESULTS: Of the 20,810 patients included in the study, 55% were female, 83% were Black, 42% were > 65 years old, and 55% received referral for PT. Black patients had more functional impairment on admission [48.1% vs 42.0% X² (2, N = 20,810), = 34.1, p < 0.001]. Compared to White patients, Black patients with functional mobility impairment on admission had less mobility gain during hospitalization [-0.69 points t(9.349) = 2.5, p < 0.05]. Fewer Black patients with functional impairment on admission received PT referral [78.3% vs 81.4% X² (2, N = 20,810), = 5.5, p < 0.05]. Black race was associated with a negative effect on mobility change during hospitalization [F(20,810) = 471.8, p < 0.001, R² = 0.15]. The association of Black race with mobility change was stronger for seniors and those with functional impairment on admission. There was no difference in PT referrals or HAD overall by race.

CONCLUSIONS: Our study suggests that Black seniors and those with functional impairments have lower rates of PT referral and less mobility gain during hospitalization. Future studies will assess a broader patient population and control for social determinants of health. Hospitalists may consider Black race a risk factor when prescribing physical therapy.

RACIAL DISPARITIES IN OPIOID PRESCRIBING, PATIENT SATISFACTION AND DOWNSTREAM OPIOID MISUSE RISK AFTER E.D. VISIT FOR ACUTE KIDNEY STONE OR BACK PAIN: SECONDARY ANALYSIS OF A MULTISITE NATIONAL RCT

Max Jordan Nguemeni Tiako¹; Abby Dolan²; Frances S. Shofer²; Eugenia South²; Marilyn M. Schapira³; Zachary Meisel²

¹Department of medicine, Brigham and Women's Hospital, Boston, MA; ²emergency medicine, University of Pennsylvania, Philadelphia, PA; ³Internal Medicine, University of Pennsylvania, Philadelphia, PA. (Control ID #3876802)

BACKGROUND: Black patients receive opioids for pain less often than similar White patients. Some suggest this disparity may have initially protected them from opioid overdoses, but overdose rates among Black people have risen. We hypothesized that underprescribing opioids to Black patients may unwittingly contribute to downstream opioid misuse. In a multicenter, national longitudinal opioid risk communication RCT, we identified disparities in opioid prescribing: Black patients had greater negative preference mismatch & were less likely to receive opioids vs White patients, regardless of expressed preference communicated to providers. We examined the association between negative preference mismatch (henceforth referred to as underprescribing), patient satisfaction & downstream opioid misuse risk.

METHODS: This is a secondary analysis of participants with complete data in an RCT of 1301 patients with no recent opioid use or misuse who presented to the emergency department (ED) for acute kidney stone or back pain. The outcome for this secondary analysis was current opioid misuse measure (COMM), a self- report 17-item measure of opioid misuse risk among people prescribed opioids for pain, measured 90 days after the index ED visit. A COMM score ≥ 9 suggests opioid misuse. We used descriptive statistics & logistic regressions to determine whether opioid underprescribing & patient satisfaction (0-10) predicted opioid misuse risk, by race/ethnicity.

RESULTS: Our sample included 735 (56.5%) participants: 46.4% (n=341) White, 36.9% (n=271) Black, & 16.7% (n=123) Non-Black people of color (NBPOC). Undertreatment was rare (17.4%,n=128) but more common among Black vs (21.8%,n=59) vs White (15%,n=51) & NBPOC (14.6%, n=18) participants. Black participants had higher mean (SD) COMM scores [8.2(10.2)] vs White [4.3(5.7)] participants, & lower satisfaction scores [6.5(6.3)] vs White participants [7.1(3.1)].

Black (vs White) participants had greater opioid misuse odds (OR 2.87,95%CI 1.84-4.46,P<0.001).

Underprescribing was associated with misuse (aOR 1.85,95%CI 1.58-2.18,P<0.001). Greater satisfaction was associated with lower misuse odds (aOR 0.88 per point increase, 95%CI 0.85-0.90,P<0.001). In a model with interactions terms between race, patient satisfaction & underprescribing, race & undertreatment weren't independently associated with misuse, unlike satisfaction (aOR 0.86, 95% CI 0.77-0.97, P=0.02). The interaction between underprescribing & Black (vs White) race was associated with misuse (aOR 3.07,95% CI 1.74-5.41, P<0.001), as was the interaction between greater satisfaction, underprescribing & Black race (aOR 0.86, 95%CI 0.78-0.95,P=0.003).

CONCLUSIONS: Opioid underprescribing was associated with increased downstream opioid misuse risk 90 days post ED visit for acute back/kidney stone pain; conversely, greater patient satisfaction was associated with lower opioid misuse risk, both with a larger effect on Black patients. Opioid underprescribing may thus create racial disparities in opioid misuse.

READMISSIONS AND POST-DISCHARGE MORTALITY BY RACE AND ETHNICITY AMONG MEDICARE BENEFICIARIES WITH MULTIMORBIDITY

Melissa Y. Wei^1,2; Jinmyoung Cho^3,4

¹Medicine, University of California Los Angeles David Geffen School of Medicine, Los Angeles, CA; ²Medicine, VA Greater Los Angeles Healthcare System, Los Angeles, CA; ³Texas A&M School of Public Health, College Station, TX; ⁴Baylor Scott & White Health, Temple, TX. (Control ID #3853306)

BACKGROUND: Medically and socially complex patients, including those with multimorbidity, disability, and unmet social needs, have among the highest rates of 30-day readmissions. Racial disparities in readmission risk have also been documented but have yielded conflicting results. We sought to examine the risk of readmissions and post-discharge mortality by race and ethnicity after rigorous adjustment for multimorbidity, physical functioning, and sociodemographic and lifestyle characteristics.

METHODS: We used Medicare Parts A and B (inpatient, outpatient, carrier) files between 1991-2015 to obtain ICD-9-CM diagnostic codes to compute the ICD-coded multimorbidity-weighted index (MWI-ICD).

Participants must have had at least one hospitalization between January 1, 2000 and September 30, 2015 and continuous enrollment in fee-for-service Medicare Part A 1-year prior to hospitalization. We used multivariable logistic regression to assess the association of MWI-ICD with 30-day readmissions and mortality 1-year post-discharge. Using Health and Retirement Study data linked to Medicare, we adjusted for age, sex, BMI, smoking, physical activity, education, household net worth, and living arrangement/marital status, and examined for effect modification by race and ethnicity.

RESULTS: The final sample of 10,737 participants had mean±SD age 75.9±8.7 years. Hispanic adults had the highest mean MWI-ICD (16.4±10.1), followed by similar values for White (mean 14.8±8.9) and Black (14.7±8.9) adults. A 1-point increase in MWI-ICD was associated with a 2% statistically significant higher odds of readmission (OR=1.02, 95% CI: 1.02-1.03), and there was no significant effect modification by race and ethnicity (p=0.27). For post-discharge mortality, a 1-point increase in MWI-ICD was associated with a 3% higher odds of mortality (OR=1.03, 95% CI: 1.03-1.04), which did not significantly differ by race and ethnicity (p=0.89).

CONCLUSIONS: Multimorbidity was associated with a monotonic increased odds of 30-day readmission and 1-year post-discharge mortality across all race and ethnicity groups. There was no significant difference in readmission or mortality risk by race and ethnicity after robust adjustment. To help reduce health disparities, interventions and policies must be implemented and targeted earlier in multimorbidity onset and progression, much prior to less modifiable sequelae such as functional decline, hospitalization, and premature mortality.

RELATIONSHIP BETWEEN SOCIAL RISK INDICATORS AND CLINICAL OUTCOMES FOR ADULTS WITH TYPE 2 DIABETES: A LATENT CLASS ANALYSIS

Joni S. Williams¹; Sebastian Linde²; Rebekah J. Walker³; Leonard E. Egede⁴

¹Medicine, Medical College of Wisconsin, Milwaukee, WI; ²Medicine, Medical College of Wisconsin Department of Medicine, Milwaukee, WI; ³Medicine, Medical College of Wisconsin, Milwaukee, WI; ⁴Medicine, Medical College of Wisconsin, Milwaukee, WI. (Control ID #3875585)

BACKGROUND: Social risk factors are the adverse social conditions associated with poor health such as housing instability, food insecurity, transportation challenges, and social isolation. Despite studies showing social risk is associated with poor diabetes outcomes, evidence is lacking on how to classify social risk indicators in diabetes. Therefore, the objective of this study was to identify subgroups of social risk among adults with diabetes.

METHODS: Latent class analysis (LCA) is a statistical methodology used to identify different subgroups within populations. Cross-sectional data of 615 adults with diabetes recruited from two primary care clinics were used for the analyses. In this study, we used responses to 19 social risk indicators: employment, financial stability, financial hardship based on income, neighborhood safety, violence in neighborhood, neighborhood rating, neighborhood problems, highest education, health literacy, food insecurity, healthy food access, social isolation, racial discrimination, perceived stress, depression, serious psychological distress, diabetes fatalism, diabetes distress, smoking and alcohol/drug use. Each social risk was dichotomized with high risk as 1 and low risk as 0. A three-step process was then used to identify the best fitting model that captured patterns of social risk indicators across individuals and investigate the relationship of those latent classes with clinical outcomes. First, 2, 3, 4, and 5-class LCA models were tested and compared to identify the best fit model based on AIC and BIC scores. Then, posterior probability was used to classify individuals into their true latent class, with accuracy of classification determined based on entropy score. Finally, using the best fit classification, the relationship with HbA1c, blood pressure and quality of life was investigated. Stata v16 was used for analyses with p<0.05 indicating statistical significance.

RESULTS: Mean age of the sample was 61.3 years with 64.9% Non-Hispanic Black, 38.4% women, and 20.2% having an income less than $10,000. The best fit model identified 2 classes, used an income cut-point for financial hardship of <$10,000 and included covariates of age, sex, race/ethnicity, marital status, duration of diabetes, and self-reported health status. Using the 19 social risk indicators, adults with diabetes were classified into 2 classes: low social risk (64.1% of the sample) and high social risk (35.9% of the sample). The entropy score of 0.96 indicated high accuracy of classification. The high social risk classification was associated with higher HbA1c and lower mental health related quality of life.

CONCLUSIONS: These findings suggest that latent class analyses can be used to classify individuals into high and low social risk categories. In addition, these categories are significantly associated with clinical outcomes for adults with type 2 diabetes.

REVIEWING INSTITUTIONAL ANTI-BIAS INTERVENTIONS- CONTRIBUTIONS FROM THE HEALTH EQUITY COMMITTEE

Sydni I. Williams¹; Elizabeth H. Zeichner¹; Tracey Henry²; Marshall Fleurant²

¹School of Medicine, Emory University School of Medicine, Atlanta, GA; ²General Internal Medicine, Emory University School of Medicine, Atlanta, GA. (Control ID #3874010)

BACKGROUND: Anti-racist or anti-bias interventions are of interest to the academic medical community. The literature base is expansive. To undertake a high-yield review of these interventions, we chose to focus on the body of work produced by the Society of General Internal Medicine’s Health Equity Commission (HEC). HEC is a committed group of health equity experts, researchers, and educators whose work focuses on social justice, eliminating health disparities & promoting equity. We conducted a scoping review of articles produced by HEC members detailing anti-racism interventions within academic medicine initiated at the institutional level as an attempt to understand the scope of interventions within academic medicine. METHODS: We identified members of the Health Equity Commission who were active as of April 2022. Using SCOPUS (an abstract and citation database), we tagged HEC members using specific author IDs; we used broad search terms focusing on anti-racism, bias, and ethnicity. Overall we identified 2831 publications. After duplicates were removed, we were left with 2276. We excluded articles with settings outside of an academic medical center or a primary focus on COVID-19. This left us with 221 publications for Title and Abstract review. Only 32 publications were subject to the full review. In full review- we discussed these articles and removed articles we determined only focused on increasing diversity but lacked initiatives to address bias or racism by consensus. In total, we were left with eight pieces.

RESULTS: We found four perspective pieces, one book chapter, one qualitative study, and two observational studies. We identified 35 different interventions across these eight publications.. The most common interventions were developing pipeline programs, implicit bias training, establishing DEI (Diversity Equity & Inclusion) leadership, or strategic planning centered on equity. Most interventions were aimed at trainees, whereas those aimed at institutions were the fewest. In each paper, interventions were multiple, stressing the need for numerous strategies to push forward anti-bias/ anti-racist agendas.

CONCLUSIONS: Within this group of papers, we have identified a broad range of strategies to reduce bias and racism within academic institutions. These strategies tend to be multi-modal and most frequently target trainees rather than faculty or specific programs. Institutional-level initiatives tended to focus on creating a DEI leadership structure or strategy. A DEI leadership structure appears to be an essential component to anti- bias or anti-racist initiatives. We suggest that institutions take a more significant role in intervening in anti- racist/ anti-bias interventions while supporting personnel in changing the environment.

SKILLED NURSING FACILITIES ARE LESS LIKELY TO ACCEPT PATIENTS WITH OPIOID USE DISORDER

Ifedayo O. Kuye¹; Laura Prichett²; Rosalyn W. Stewart³; Scott A. Berkowitz⁴; Megan Buresh⁵

¹Hospital Medicine, Johns Hopkins Medicine, Baltimore, MD; ²Pediatrics, Johns Hopkins Medicine, Baltimore, MD; ³General Internal Medicine, Johns Hopkins, Baltimore, MD; ⁴Medicine, Cardiology, Johns Hopkins Medicine, Clarksville, MD; ⁵Division of Chemical Dependence, Johns Hopkins School of Medicine, Baltimore, MD. (Control ID #3854194)

BACKGROUND: An increasing number of patients with Opioids Use Disorder (OUD) are hospitalized and referred to Skilled Nursing Facilities (SNFs). The Department of Justice has stated that the American Disabilities Act protects patients with OUD who are not currently engaged in illicit drug use from discrimination by SNFs. However, prior single center studies have identified that patients with OUD encounter more barriers to SNF placement when compared to patients without OUD. In this study, we analyzed SNF referral data from two academic hospitals Maryland. We hypothesized that patients with OUD would have more SNF referrals than those without OUD and referrals sent for patients with OUD would be less likely to be accepted.

METHODS: We examined electronic referrals from Johns Hopkins Hospital and Johns Hopkins Bayview Medical Centers, two academic medical centers in Baltimore, Maryland from January 1 to December 31, 2019. We linked referral data to the corresponding hospitalizations and downloaded clinical, demographic, disposition, and insurance data from the Epic Clarity Warehouse. The main outcomes of interest were the number of SNF referrals sent and the proportion of referrals accepted. OUD status was determined by receipt of medications for OUD(MOUD) during admission, upon discharge or the presence of a diagnosis code for OUD. Multivariate multilevel mixed regression models were constructed using a modified forward stepwise approach, adjusting for age, race, gender, non-English as primary language, type of insurance, psychiatric consult, Elixhauser score, contact precautions during admission, and alcohol use disorder for all outcomes.

RESULTS: Cohort included 6,047 hospitalizations (5422 without OUD and 625 with OUD). Patients with OUD had more SNF referrals sent (8.8 vs 5.6, p<0.001), had a lower proportion of SNF referrals accepted (31.3% vs 46.9%, p<0.001) and were less likely to be discharged to a SNF (65.6% vs 70.3% p<0.001). The effect of OUD status on the number of SNF referrals and the proportion of referrals accepted remained significant in the multivariate analyses. Our sub-analysis showed that reduced acceptances were driven by patients who were discharged without MOUD and on methadone. Those discharged on buprenorphine were accepted at the same rates as patients without OUD.

CONCLUSIONS: This multicenter retrospective cohort study found that patients with OUD had more SNF referrals, fewer referrals accepted and were less likely to be discharged to a SNF compared to those without OUD. Patients with buprenorphine had similar acceptance rates to patients without OUD, while acceptance rates were lower for patients receiving methadone and those not receiving MOUD. Our study highlights the need for more work to address disparities in the access to and quality of post-acute care for patients with OUD. This will require innovative interventions and policies to decrease provider stigma, improve the quality of SNF care and lift barriers to MOUD access, specifically methadone, at SNFs.

SOCIAL SUPPORT ASSOCIATION WITH DIABETES OUTCOMES FOR ADULTS WITH AND WITHOUT FOOD INSECURITY AT AN URBAN FEDERALLY QUALIFIED HEALTH CENTER Nicole Luche¹; Ann-Marie Rosland^2,3; Edith C. Kieffer⁶; Stephanie Perez⁴; Michele Heisler^7,8; Denise Deverts²; Gretchen Piatt^5,7; Felix M. Valbuena⁴; Jonathan Yabes²; Christina M. Lalama²; Margaret Zupa¹

¹Internal Medicine, UPMC, Pittsburgh, PA; ²Internal Medicine, University of Pittsburgh School of Medicine, Pittsburgh, PA; ³Center for Health Equity Research and Promotion, VA Pittsburgh Healthcare System, Pittsburgh, PA; ⁴Community Health and Social Services Center, Inc., Detroit, MI; ⁵School of Public Health, University of Michigan, Ann Arbor, MI; ⁶School of Social Work, University of Michigan, Ann Arbor, MI; ⁷School of Medicine, University of Michigan, Ann Arbor, MI; ⁸Center for Clinical Management Research, VA Ann Arbor Healthcare System, Ann Arbor, MI. (Control ID #3873005)

BACKGROUND: Adults with type 2 diabetes (T2D) and food insecurity have worse glycemic control and more diabetes distress than those without food insecurity. However, high social support is associated with better glycemic control and less diabetes distress in adults with T2D. Social support may also protect against the development of food insecurity and its associated negative effects through direct provision of food, financial support, and increased awareness of community resources. It is unknown whether increased social support buffers the negative impacts of food insecurity on diabetes distress and HbA1c for adults with T2D. We examined the impact of social support on these outcomes in adults with T2D with and without food insecurity.

METHODS: Data were collected from baseline assessments of 204 primarily Latino/a adults with T2D participating in a diabetes self-management support study at an urban federally qualified health center. All patients were enrolled with a family or friend ‘support person.’ Assessments included HbA1c, as well as validated survey instruments for food insecurity (Hunger Vital Sign, y/n), perceived social support (ENRICHD Social Support Instrument, 4-point scale), and diabetes distress (Problem Areas in Diabetes-5, 20-point scale). We used multiple linear regression models to examine the independent associations of food insecurity and social support with diabetes distress and HbA1c. An interaction term (food insecurity*social support) was then added to assess for moderation effects. Models were adjusted for patient age, sex, and insulin use.

RESULTS: Almost half (45%, n=92) of patients reported food insecurity. In bivariate analysis, a 1-point increase in social support was significantly associated with reduced odds of food insecurity (OR 0.75, 95% confidence interval [0.57, 0.99]). In adjusted models, food insecurity was significantly associated with a 3.38 (1.77, 4.99) point increase in diabetes distress score and 0.66% (0.15, 1.16) increase in HbA1c. In the same models, a 1-point increase in social support was independently and significantly associated with less diabetes distress (-0.79 [-1.59, 0.01] points), but not with HbA1c (0.08% [-0.15%, 0.31%]). Social support did not moderate the association of food insecurity with HbA1c or diabetes distress.

CONCLUSIONS: In this primarily Latino/a population of adults in a low-resource community with at least one support person, higher social support was associated with lower diabetes distress among both food secure and food insecure participants. However, higher social support did not mitigate the negative impacts of food insecurity on diabetes distress and HbA1c. This study provides evidence that enhancing social support may be a viable approach to improving diabetes outcomes for adults with food insecurity, but efforts to directly address food insecurity may also be necessary to achieve the greatest impact.

SOCIOECONOMIC AND RACIAL DISPARITIES IN COVID-19 ASSOCIATED HOSPITALIZATIONS AMONG VETERANS IN FIVE NEW ENGLAND STATES

Ya Haddy Sallah^1,2; Kate Ferencsik^3,2; Shaili Gupta²

¹Medicine, Massachusetts General Hospital, Boston, MA; ²Internal Medicine, Yale University School of Medicine, West Haven, CT; ³Augusta University, Augusta, GA. (Control ID #3875376)

BACKGROUND: Race-based COVID-19 outcomes have been reported. Few studies have explored the impact of neighborhood socioeconomic disadvantage (nSED), a key determinant of many health conditions, on COVID-19 outcomes and its interplay with race. We aim to 1) describe the distribution of nSED among Veterans diagnosed with SARS-CoV-2 in the New England area using Area Deprivation Index (ADI) 2) determine the association between ADI and hospitalization, and 3) investigate the interplay between ADI, race and hospitalization.

METHODS: A prospective cohort study was performed using Veteran’s Affairs electronic medical records in five New England states (CT, MA, ME, NH, RI) between 3/1/2020 and 6/30/2022. Patient’s addresses were geocoded and linked with ADI, a validated composite measure of nSED (Higher ADI reflects more disadvantage). ADI terciles were used for statistical analyses. The primary outcome was hospitalization with a positive COVID-19 test 7 days prior to 48 hours after admission. Logistic regressions were performed using STATA 15.

RESULTS: Of 8,538 veterans with COVID-19 (45% in MA, 29% in CT, 12% in RI, 7.3% in NH, 7% in ME), 80.3% identified as White, 10.8% as Black, 0.8% as Asian, 0.7% as American-Indian/Alaskan Native, 0.45% as Native Hawaiian/ other-Pacific-Islander, and 6.9% were unknown. Ethnicity was Hispanic in 6.5%. The mean age was 58.6 years (SD 17.4). Only 9.2% were female. ADI scores ranged from 1 to 98 with a mean of 38.5 (SD 20). ME had the highest average ADI score (58.3; SD 17.8) and MA had the lowest (27.5; SD 15.3). Hispanic patients had a higher mean ADI score (44.1 vs 38.3; p<0.05), but no association with hospitalization. Black patients had higher mean ADI scores compared to White patients (43.6 vs 37.8, p<0.05), and were 42% more likely to be hospitalized (OR 1.42, 95% C.I:1.15-1.78, p<0.05). ADI score did not differ by gender and age. Compared with patients living in the least disadvantaged neighborhoods, those living in the most disadvantaged neighborhoods were 27% more likely to be hospitalized (OR: 1.27, 95% C.I: 1.06-1.51, p<0.05), after adjusting for medical comorbidities. Older age, diabetes, COPD, CHF and cirrhosis were associated with being hospitalized. Among Black patients alone, ADI tercile was not significantly associated with hospitalization. However, among white patients, those living in more disadvantaged neighborhoods had increased odds of hospitalization (OR: 1.34, 95%: 1.11-1.62, p<0.05).

CONCLUSIONS: Hospitalization from COVID-19 was significantly associated with nSED. Black patients had higher odds of hospitalization independent of nSED. Interestingly, the relationship between hospitalization and nSED was only seen among White patients. These findings highlight the importance of assessing the interaction between race and nSED when investigating disparities in health outcomes.

SPANISH ADVANCEMENT FOR LEARNERS UNDERSTANDING DISPARITIES (SALUD): PILOT STUDY OF AN INSTRUCTOR-LED MEDICAL SPANISH CURRICULUM PROMOTING LANGUAGE CONCORDANT CARE

Miguel Tusa Lavieri, George S. Corpuz, Chimsom Orakwue, Andrea Cabassa Miskimen, Christopher J. Gonzalez, Shweta R. Iyer, Maria Lame, Joy D. Howell

Medical College, Weill Cornell Medicine, New York, NY. (Control ID #3877202)

BACKGROUND: The US population is growing progressively more diverse and people from LatinX cultures are the fastest growing segment of the population. Limited English Proficiency (LEP) is a barrier to health equity. Given the impact of language-concordant care on patient safety, high quality language education should be accessible to medical students. Herein, we evaluate the impact and feasibility of SALUD, a five-week intensive Medical Spanish (MS) pilot course implemented at an academic medical center in New York City.

METHODS: A five-week virtual course was led by an experienced medical Spanish instructor. Students met with the instructor thrice weekly and participated in six standardized patient encounters throughout the course. Student self-assessment as well as an informal assessment of baseline conversational competency was performed by the instructor prior to the course. Upon completion of the course, the instructor administered formal evaluations of MS core competencies, including: self-introduction, culturally-appropriate history- taking, effective information transfer, and student-patient dynamic. Focus groups were used to elicit student feedback and reflections immediately after the course.

RESULTS: Fifteen students completed the course. Eighty percent (12 of 15) of students self-reported their overall Spanish competency and comprehension abilities as “intermediate” prior to the course. Eighty-seven percent (13 of 15) of students received a post-course instructor rating of “outstanding,” determined by an average score across all four core competencies. The main themes identified by student focus groups include: (1) enhanced understanding of how to appropriately use medical Spanish versus a certified interpreter and (2) cultural nuances relevant to working with a variety of Hispanic/Latino patients with Limited English Proficiency (LEP) in a clinical setting.

CONCLUSIONS: The implementation of a rigorous medical Spanish curriculum is feasible within the UME curriculum with robust faculty and institutional support. Medical Spanish education is perceived by students as beneficial to their ability to care for Spanish-speaking patients with LEP. Future iterations of SALUD will include formal pre and post course medical Spanish competency assessed by students, instructor and standardized patients. Rigorous educational outcomes can be used demonstrate the impact of future/ similar courses on learners and patients.

SPATIAL EXPLORATION OF NEIGHBORHOOD-LEVEL SOCIAL DETERMINANTS OF HEALTH AND DIABETES MORTALITY IN CHICAGO

Andrew S. Palmer¹; Marynia Kolak²

¹Biological Sciences Division, University of Chicago Pritzker School of Medicine, Chicago, IL; ²Geography and Geographic Information Science, University of Illinois Urbana- Champaign, Urbana, IL. (Control ID #3871928)

BACKGROUND: Within Chicago, the prevalence of diabetes is higher in neighborhoods with larger minority populations and lower socioeconomic status. While the relationship between various Social Determinants of Health (SDOH) and diabetes mortality are well established, relatively little work has been done to examine the spatial heterogeneity of these effects.

METHODS: A conceptual model was developed following a survey of the literature that identified twenty publicly available neighborhood-level variables to capture dimensions of pathways by which SDOH impact disease burden. Datasets represented in this study range from 2007 to 2021. In this cross-sectional study of Chicago, 77 community areas were used for the creation of multidimensional indices of SDOH using Principal Component Analysis (PCA) which retained 5 axes using the Kaiser criterion. The relative contribution of each SDOH factor into a principal component was used to characterize each component: economic/neighborhood stressors, health literacy/language proficiency, neighborhood built resources and healthcare deprivation, healthcare utilization, and air pollution (particulate matter 2.5 microns in size) and community belonging. Six regional typologies were created through K-means clustering with neighborhood PCA index scores. ANOVAs were used to identify which clusters of neighborhoods had significantly different diabetes mortality rates from the rest of Chicago.

RESULTS: Regional typologies with both higher and lower-than-average diabetes mortality did not vary along the principal components in a uniform way. For example, one cluster comprised mainly of neighborhoods with large Black populations, including west Chicago and south-central Chicago with higher diabetes mortality (36.19 vs 24.61 per 100,000; p <0.001), had higher economic/neighborhood stressors, health literacy/English proficiency, neighborhood physical resources and healthcare deprivation, healthcare utilization, and air pollution/community belonging. Another cluster of neighborhoods in far south Chicago (comprised mainly of neighborhoods with large Hispanic/Latino populations) with higher diabetes mortality (39.18 vs 24.61 per 100,000; p= 0.022) had higher economic/neighborhood stressors, higher health literacy/English proficiency, lower neighborhood physical resources and healthcare deprivation, lower healthcare utilization, and lower air pollution/community belonging.

CONCLUSIONS: A deeper understanding of the spatial heterogeneity of the factors that impact diabetes mortality may allow for more spatially targeted allocation of limited public health monies. By elucidating typologies of SDOH and their association with diabetes mortality, neighborhood-specific interventions may improve outcomes more than interventions that treat populations as hom*ogenous.

START: A COLLEGIATE PROGRAM TO BRIDGE THE DIGITAL TELEHEALTH DIVIDE

Zakaria N. Doueiri¹; Rika Bajra²; Erika Schillinger³; Malathi Srinivasan⁴; Nancy L. Cuan²

¹Human Biology, Stanford University, Palo Alto, CA; ²Internal Medicine, Stanford Medicine, Palo Alto, CA; ³Family Medicine, Stanford Medicine, Palo Alto, CA; ⁴Medicine, Stanford University School of Medicine, Palo Alto, CA. (Control ID #3875599)

BACKGROUND: Technology-enabled patients have improved access to care through telehealth. However, telehealth is not available to many vulnerable patients due to issues with technology literacy or access. We partnered with Primary Care and a Federally Qualified Health Center (FQHC) to develop START, a collegiate program to bridge the patient digital divide, while providing a valuable student learning experience.

METHODS: Our Primary Care group partnered with our academic medical center’s Digital Solutions and Technology and Compliance teams to create a program that trained students to help vulnerable patients access video visits during the COVID-19 pandemic. Undergraduate and graduate students learned about social determinants of health, design thinking, and working with community partners and developed their communication skills. Students contacted patients from electronic medical system lists, generated from searches for patients without an electronic medical portal account or who never utilized video visits. Six patients were referred directly to START by clinicians. To extend our services to the community, we partnered with an FQHC to connect community patients with telehealth. During the course, students submitted weekly reflections on their experiences with patients or guest speakers. Student reflections were coded for thematic content, using a rapid cycle methodology.

RESULTS: Over 6 academic quarters, 57 students outreached to 1185 patients to assist with telehealth technology. Sixty percent of the patients were at our institution and 40% were from an FQHC. Of the 229 patients reached, 141 patients wanted access to telehealth. Seventy-nine percent (n=111) of patients successfully established an electronic medical portal account and/or set up a video visit. Reasons for inability to establish access included lack of access to Wi-Fi or a device (n=11), absence of an interpreter (n=4), and a disability that precluded the patient’s use of video visits (n=2). Students reported (n=134 reflections) impact on future career goals and an improved awareness of health equity disparities, differential privilege in technology access, and the impact of the pandemic on health care.

CONCLUSIONS: Amongst patients whom we reached who desired access to telehealth, the START program improved access to telehealth for 79% of patients. Undergraduate and graduate students found that START broadened their understanding of health disparities and social determinants of health and influenced their future career goals.

STORIES FOR CHANGE PROTOCOL: A RANDOMIZED CONTROLLED TRIAL OF A DIGITAL STORYTELLING INTERVENTION FOR HISPANIC/LATINO INDIVIDUALS WITH TYPE 2

DIABETES

Lohr M. Abby^1,2; Katherine Vickery⁸; Valentina Hernadez⁷; Becky R. Ford⁸; Crystal Gonzalez⁷; Silvio Kavistan⁸; Christi A. Patten^4,2; Jane Njeru¹; Paul Novotny⁵; Linda K. Larkey⁶; Davinder Singh⁷; Mark L. Wieland¹; Irene G. Sia³

¹Division of Community Internal Medicine, Geriatrics, and Palliative Care, Mayo Clinic Minnesota, Rochester, MN; ²Center for Clinical and Translational Science, Mayo Clinic Minnesota, Rochester, MN; ³Division of Public Health, Infectious Diseases, and Occupational Medicine, Mayo Clinic Minnesota, Rochester, MN; ⁴Department of Psychiatry and Psychology, Mayo Clinic Minnesota, Rochester, MN; ⁵Division of Biomedical Statistics and Informatics, Mayo Clinic Minnesota, Rochester, MN; ⁶Edson College of Nursing and Health Innovation, Arizona Board of Regents, Phoenix, AZ; ⁷Mountain Park Health Center, Phoenix, AZ; ⁸Health, Homelessness & Criminal Justice Lab, Hennepin Healthcare Research Institute, Minneapolis, MN. (Control ID #3874301)

BACKGROUND: Hispanic/Latino adults are disproportionately impacted by type 2 diabetes mellitus (T2D). Innovative approaches that arise from affected communities are needed. Community-based participatory research (CBPR) has been successful in addressing health issues among Hispanic/Latino and immigrant populations as an engaging and effective approach for addressing health behaviors in a sociocultural and socio-ecological context. The Stories for Change (S4C) Diabetes digital storytelling intervention, developed with and for Hispanic/Latino adults, promotes T2D self-management. We describe the S4C protocol and participant baseline characteristics of participants.

METHODS: Study eligibility criteria included being Hispanic or Latino, age 18 - 70 years, with ≥1 office visit within the previous year at a participating clinic (in the midwest and southwest US), with T2D diagnosis for ≥6 months, HbA1c≥8%, and intention to continue care at the recruitment clinic. We used a two-group, parallel randomized controlled trial design and a community-based participatory research approach. All participants received usual diabetes care and two cards describing how to engage healthcare teams and access diabetes-related resources. At baseline, the intervention group additionally viewed the 12-minute, intervention video (four stories about diabetes self-management). To encourage subsequent video viewing, participants received five monthly text messages. The messages prompted them to self-rate their motivation and self-efficacy for managing their T2D. The control group received no additional intervention. Bilingual (English/Spanish) staff collected data at baseline, six weeks, three months, and six months including biometric measurements and a survey on diabetes self-management outcomes, theory-based measures, and the number of video views. We reviewed the number of diabetes-related appointments attended using electronic medical record data.

RESULTS: Participants (n=451; 70% women, mean age=53 years) had an average HbA1C ≥9%. There were no significant sociodemographic or health-related differences between the intervention and control groups, apart from the use of insulin and other diabetes medications, with more of the intervention participants (61%) compared to the control participants (52%) on insulin. While on average participants had a BMI in the obese range (>30 kg/m2), their low-density lipoprotein (LDL) cholesterol was optimal (<130 mg/dL). On average, most study participants reported excellent diabetes medication use but suboptimal diet, exercise, and blood glucose testing. Participants in both groups described similar levels of self-efficacy, social support for diabetes, and diabetes-related quality of life. The intervention group reported high story identification and transportation after viewing the S4C digital story.

CONCLUSIONS: We present a digital storytelling intervention protocol that will provide a template for future health promotion interventions prioritizing healthy disparity populations.

STRESS EXPOSURE AS A MEDIATOR BETWEEN RACIAL IDENTITY AND HEALTH AMONG LATINO INDIVIDUALS

Jordan J. Juarez¹; Rachel A. Zajdel¹; Erik Rodriquez¹; Eliseo J. Perez-Stable²

¹Division of Intramural Research, National Heart, Lung, and Blood Institute, Bethesda, MD; ²NIMHD, National Institutes of Health, Bethesda, MD. (Control ID #3874146)

BACKGROUND: The objective was to identify disparities in mental and physical health for Latinos who are differentially racialized, and to test stress as a mechanism linking racialization to health.

METHODS: We used data from Visit 1 (2008-2011), Visit 2 (2014-2017), and the Sociocultural Ancillary Study of the Hispanic Community Health Study/Study of Latinos, a multi-center cohort study of Latino adults. Participants (n=3907) were: White (n=1645), Multiracial (n=664), Black (n=135), American Indian/Alaska Native (AI/AN) (n=125), and Unknown/Not Reported (n=1338). Generalized structural equation modeling used racial identity (main predictor), discrimination stress (mediator), and depressive symptoms, anxiety symptoms, diabetes, and hypertension incidence (outcomes).

RESULTS: Mean age was 42, 52% were female, 39.9% had greater than a high school education, 76.5% were first generation immigrants, 51.8% were uninsured, and 40.1% were obese (BMI ≥ 30). Mean depressive and anxiety symptoms were 6.46 and -0.02, respectively. Identifying as AI/AN (β=0.39; CI=0.17-0.82), Multiracial (β=0.07; CI=0.01-0.16), or not racially identifying (β=0.08; CI=0.02-0.18) was associated with greater increases in depressive symptoms due to discrimination (ref=White, Table 1).

Individuals who identified as AI/AN (β=0.05; CI=0.02-0.10), Multiracial (β=0.01; CI=0.0-0.02), or no racial group (β=0.01; CI=0.0-0.03) reported greater increases in anxiety symptoms due to discrimination. Diabetes and hypertension incidence were 16.2 and 11.5 new cases per 100 respondents, respectively. Identifying as AI/AN or not racially identifying was associated with greater diabetes (β=0.12; CI=0.02-0.33; β=0.02;

CI=0.0-0.07, respectively) and hypertension (β=0.16; CI=0.03-0.43; β=0.03; CI=0.0-0.10, respectively; Table 1) incidence due to discrimination.

CONCLUSIONS: Findings suggest discrimination stress is a potential mediator linking racialization to mental and physical health, especially among AI/AN, Multiracial, and racially unspecified Latinos.

TEMPORARILY DE-CENTERING THE PATIENT TO REIMAGINE PATIENT-CENTERED CARE: ENGAGING COMMUNITY EXPERTS AND HEALTH SYSTEM PERSONNEL TO INFORM IMPLEMENTATION STRATEGIES FOR REDUCING HYPERTENSION DISPARITIES IN A MULTI-ETHNIC SAFETY-NET HEALTH SYSTEM

Utpal Sandesara¹; Savanna L. Carson²; Alex Dopp³; Lilian Perez³; Atkia Sadia²; Soma Wali⁵; Nina Park⁶; Alejandra Casillas⁴; Gloria Kim²; Maria Morales²; Ejiro Ntekume²; Priya Gandhi³; Anthony Wafford⁸; Arleen F. Brown⁷

¹Internal Medicine, University of California Los Angeles, Los Angeles, CA; ²University of California Los Angeles, Los Angeles, CA; ³RAND Corporation, Santa Monica, CA; ⁴Medicine, David Geffen School of Medicine @ UCLA, Los Angeles, CA; ⁵Medicine, UCLA Medical Center Olive View, Sylmar, CA; ⁶Los Angeles County Department of Health Services, Los Angeles, CA; ⁷Medicine, University of California Los Angeles, Los Angeles, CA; ⁸I Choose Life Health and Wellness Center, Los Angeles, CA. (Control ID #3875878)

BACKGROUND: Despite advances in evidence-based practices (EBPs) for managing hypertension (HTN), few implementation efforts systematically incorporate community and healthcare engagement strategies for reducing HTN disparities. In parallel, research on HTN has historically overemphasized patient-level factors (such as knowledge, beliefs, and individual behaviors) while underemphasizing community, social structural, and health system factors. Gathering perspectives from community and health system experts can take us beyond individual patients to confront the historical and sociocultural contexts that affect implementation of HTN EBPs in real-world settings.

METHODS: We conducted a multi-modal qualitative study in partnership with a large U.S. municipal health system. We convened 5 race/ethnicity-specific community advisory boards (CABs), each of which met 5 times (25 total meetings) to discuss HTN disparities, treatments, and cultural considerations. We also conducted semi-structured qualitative interviews (SSQIs) with 41 health system personnel about patient, community, and health system characteristics influencing HTN management. We solicited implementation recommendations from both groups. We summarized findings using rapid content analysis, followed by a pile-sort to identify higher-level content themes.

RESULTS: CABs and SSQIs identified historically and socially constituted conditions that act as barriers/facilitators for evidence-based HTN management in the current state. They also identified considerations for culturally/contextually tailoring EBPs. The conditions and considerations identified spanned 7 themes distributed across the 4 domains of the EPIS framework for implementation science: 1) social needs and 2) medical/health engagement needs (Outer Context); 3) dynamics of patient interactions and 4) health system operational factors (Inner Context); 5) standardization/flexibility in EBP implementation and 6) education materials (Innovation Factors); and 7) partnerships and funding (Bridging Factors).

CONCLUSIONS: The conditions and considerations identified will inform a randomized trial to test multi- level implementation strategies for cultural/contextual tailoring of HTN EBPs in our partnered municipal health system. More broadly, this study offers a model for community-engaged, health system-partnered research that can inform multi-level efforts to reduce disparities in HTN and other chronic diseases. By temporarily de-centering the patient as the primary agent (and hence locus of responsibility, blame, etc.) in HTN control, engaging community and health system experts puts the focus on complex processes that span multiple social levels. Amid significant sociocultural and medical challenges to patients achieving transformations in their health-related knowledge, attitudes, behaviors, and outcomes, this epistemological de-centering offers a method for tailoring implementation of EBPs to reshape health systems and community contexts in a way that is truly patient-centered.

THE ASSOCIATION BETWEEN AGE AND FUNCTIONAL DISABILITY IN US ADULTS WITH LIFETIME EXPOSURE TO THE CRIMINAL LEGAL SYSTEM, 2015-2019

Laura Hawks¹; Sanjay Bhandari²; Rebekah J. Walker³; Leonard E. Egede⁴

¹General Internal Medicine, Medical College of Wisconsin, Wauwautosa, WI; ²Internal Medicine, Medical College of Wisconsin, Milwaukee, WI; ³Medicine, Medical College of Wisconsin, Milwaukee, WI; ⁴Medicine, Medical College of Wisconsin, Milwaukee, WI. (Control ID #3874619)

BACKGROUND: Due to the long legacy of mass incarceration, a burgeoning number of older adults have a history criminal legal involvement (CLI). Prior CLI is associated with worse chronic disease control and poorer outcomes, though the exact nature of this relationship is not well understood. Functional disability is an important predictor of disease control, and while CLI and older age are both associated with disability, the nature of this relationship is understudied. Therefore, we examined the association between age and functional status in adults with CLI, hypothesizing that older adults in this group would have greater disability.

METHODS: Using data from the National Survey of Drug Use and Health (2015–2019), we included 37,279 US adults who reported lifetime history CLI (ever been arrested, or on probation or parole). We categorized respondents by age: 18-49 years (younger adults); 50-64 years (middle-aged adults); and 65+ years (older adults). Functional status was measured with the World Health Organization Disability Assessment Schedule (WHO-DAS 2.0), for which higher scores equal greater functional disability. Using multivariate linear regression, we tested the association between age category and functional disability score controlling for relevant sociodemographic and clinical covariates.

RESULTS: In our sample of US adults with lifetime CLI, there were 59.9% younger adults, 28.5% middle- aged adults and 11.5% older adults. The youngest group was more likely to be female, was more racially and ethnically diverse, more likely to be working but also poorer than the middle-aged and older groups. The older group reported higher rates of most comorbid conditions, but fewer substance use disorders or mental health conditions. The unadjusted association between WHO-DAS 2.0 score and age found that functional disability decreased as age increased: with younger adults as the referent, the coefficient for middle aged adults was -0.7 (95% CI -1.0 - -0.5) and for older adults was -1.8 (95% CI, -2.1 - -1.5). After adjustment for all confounders, this association persisted: with younger adults as the referent, the coefficient for middle-aged adults was -0.4 (95% CI -0.5 - -0.3); for older adults with was -1.2 (95% CI, -1.5 - -0.9).

CONCLUSIONS: The pattern of functional decline with age for the general population does not hold true for Americans with lifetime history of CLI, in which younger adults have greater functional disability than middle-aged adults, who have greater functional disability than older adults. These findings persisted after controlling for sociodemographic and clinical covariates and contradicted our hypothesis. They suggest that greater attention should be paid to functional status after exposure to CLI. Interventions which aim to identify and treat causes for functional disability, particularly in younger populations with CLI, may be an important component to reducing health disparities for this highly marginalized group.

THE BURDEN OF PREMATURE CORONARY HEART DISEASE AMONG ADULTS WITH LOW SOCIOECONOMIC STATUS IN ARGENTINA.

Maria V. Salgado¹; Joanne Penko³; Alicia Fernandez³; Francine Rios-Fetchko³; Raul Mejia^1,2

¹Health, Economy and Society, CEDES, Buenos Aires, Argentina; ²Ambulatory Medicine, Hospital de Clinicas Jose de San Martin, Buenos Aires, Argentina; ³University of California San Francisco, San Francisco, CA. (Control ID #3872384)

BACKGROUND: There is an inverse relationship between socioeconomic status (SES) and risk of developing coronary heart disease (CHD). This association cannot be explained solely by differences in the prevalence of traditional cardiovascular risk factors. The objective of this study is to model the role SES plays in the burden of premature coronary heart disease (CHD) in Argentina.

METHODS: We used the CVD Policy Model – Argentina, a Markov computer simulation model, to project incident CHD events and mortality in low-SES adults 35 to 64 years of age from 2015 to 2024. We used data from the 2018 National Risk Factor Survey (NRFS). We defined low SES as not finishing high school and/or reporting a household income in quintiles 1 or 2. We designed simulations to apportion CHD outcomes in low SES adults to: (1) differences in traditional risk factors (hypertension, smoking, diabetes and obesity) observed in adults with low compared with higher SES; (2) risk observed for low SES independent of traditional risk factors; (3) events that could be prevented if risk factors were improved to ideal levels; and (4) underlying age- and sex-based risk not explained by traditional risk factors or SES (unmodifiable risk)

RESULTS: Over half (56%) of adults aged 35 to 64 years have low SES according to the NRFS. We observe that risk factor control was far from ideal among both low and higher SES populations, and that this inadequate management explains more than half of the preventable CHD events in Argentina s population. Simulations 1 and 2 estimate that low SES adults compared to those of higher SES have an overall excess of 28.7 incident CHD events per 10000 person-years (men: 37.9; women: 20.4), and 2.7 excess CHD deaths per 10000 person-years (men: 4; women: 1.4) from 2015-2024. The independent risk associated with Low SES (simulation 2) accounts for 73.5% and 70.4% of this event rate gap between SES levels for incident CHD and CHD mortality rates respectively; this finding represents the disparity that would remain even if levels of traditional risk factors were similar among low and higher SES populations

CONCLUSIONS: Adults with low SES have rates of incident CHD and CHD deaths that are nearly two- fold higher than those of higher SES adults. Most of this disparity is explained by the independent effect of low SES and not by differences in traditional risk factors. Nevertheless, both low and higher-SES Argentinian adults have poor management of cardiovascular risk factors. CHD prevention policies should explore how to address contextual aspects linked to SES, as well as aim to improve preventable risk factor control in all socioeconomic levels.

THE CUMULATIVE BURDEN OF SOCIAL DETERMINANTS OF HEALTH AND THE 10-YEAR CHANGE IN QUALITY OF LIFE.

Ro-Jay Reid²; Monika M. Safford²; W. Marcus Lambert¹; Joanna B. Ringel²; Laura Pinheiro²; Madeline R. Sterling²; Barrett Bowling⁶; Emily B. Levitan³; Samprit Banerjee⁴; Raegan Durant⁵; Michael Kim²; Jennifer D. Lau²; Parag Goyal²

¹Epidemiology and Biostatistics, SUNY Downstate Health Sciences University, New York City, NY; ²Medicine, Weill Cornell Medicine, New York, NY; ³Epidemiology, University of Alabama at Birmingham, Birmingham, AL; ⁴Population Health Sciences, Weill Cornell Medicine, New York, NY; ⁵Department of Medicine, The University of Alabama at Birmingham, Birmingham, AL; ⁶Medicine, Duke University, Durham, NC. (Control ID #3875951)

BACKGROUND: Americans are not necessarily enjoying a higher quality of life (QoL). Social determinants of health (SDOH) have shown to have a significant association on QoL. Yet it is unknown how the cumulative effect of SDOH affects QoL. This study will determine the impact of the cumulative effect of SDOH on patient reported mental and physical health status over 10 years

METHODS: Data was obtained from Black and White participants of age ≥ 45 years old from REasons for Geographic and Racial Differences in Stroke (REGARDS) – a population-based observational cohort study. Exposure: 1) Annual income <$35,000; 2) education < high school; 3) zip code with >25% residents living below Federal poverty line; 4) no health insurance; 5) living in a state with poor public health infra- structure); 6) living in rural areas; 7) living in a health professional shortage area; 8) seeing fewer than 2 friends or family in the past month; 9) having no one to care for you if you fell ill.

Outcome: Decline > 5 points in physical component summary (PCS) and mental component summary (MCS) scores from the Short Form-12 over 10 years.

Statistical analysis: The sample was stratified by race. SDOH associated with decline in PCS and separately MCS (RR > 1.0) were examined and then used to create a SDOH count of (0, 1, ≥2). Poisson regression analyses were used to estimate RR and 95% CI for the age-adjusted and fully adjusted associations between the count of SDOH and each outcome, adjusting for demographics, health behaviors and medical conditions.

RESULTS: At baseline, the mean age (SD) was 62.9 (8.34) years and 56.2% were female. A total of 9085 participants self-identified as White while 5316 participants self-identified as Black. Among White participants 36.4% had a decline in PCS while 15.8% had a decline in MCS. Among Black participants 35.4% had a decline in PCS while 19.2% had a decline in MCS.

White participants.

Compared to those without SDOH in the age-adjusted model for PCS, those with ≧2 SDOH had RR of 1.26 (95% CI 1.17-1.35). In the fully adjusted model, the RR was 1.19 (95% CI 1.09-1.29) when compared to those without SDOH.

Compared to those without SDOH in the age-adjusted model for MCS, those with ≥2 SDOH had RR of 1.48 (95% CI 1.31-1.68). In the fully adjusted model, the RR was 1.32 (95% CI 1.15-1.51) when compared to those without SDOH.

Black participants

Compared to those without SDOH in the age-adjusted model for PCS, those with ≥2 SDOH had RR 1.18 (95% CI 1.06-1.30). In the fully adjusted model, the RR was 1.15 (95% CI 1.04-1.28) when compared to those without SDOH.

Compared to those without SDOH in the age-adjusted model for MCS, those with ≧2 SDOH had RR of 1.32 (95% CI 1.10-1.59). In the fully adjusted model, the RR was 1.26 (95% CI 1.05-1.53) when compared to those without SDOH.

CONCLUSIONS: Exposure to a greater number of SDOH may accelerate impairments in quality of life over time.

THE EXPERIENCE OF VISITING INCARCERATED FAMILY MEMBERS

Louisa Holaday^2,3; Cerella Craig³; Kimberly Stone²; Emily Wang^1,3

¹Internal Medicine, Yale University School of Medicine, New York, NY; ²Icahn School of Medicine at Mount Sinai, New York, NY; ³Yale University SEICHE Center for Health and Justice, New Haven, CT. (Control ID #3876631)

BACKGROUND: The United States currently incarcerates approximately 2 million people. Over 40% of Americans have had an immediate family member incarcerated, including over 60% of Black and Indigenous Americans. Having had an incarcerated family member is associated with lower well-being, and living in a community with high incarceration rates is associated with a higher risk of anxiety and depression. Non- incarcerated people with incarcerated loved ones may visit. Visitation has benefits for incarcerated people, including better mental health, and fewer behavioral infractions in the weeks leading up to visitation. Moreover, after release, formerly incarcerated individuals who experienced visitation are less likely to re- offend or be re-incarcerated. Yet, few studies have examined the experience of visiting for the non- incarcerated person.

METHODS: We performed a cross-sectional analysis of mixed methods data from the nationally- representative FamHIS survey, conducted by NORC in 2018. We examined rates of visitation overall, and odds of visitation by age, gender, income, financial impact of incarceration, race/ethnicity, region, and duration of incarceration using logistic regression. We used thematic analysis to evaluate responses to an open-ended question about the experience of visitation.

RESULTS: Our sample included 1,991 people with family incarceration experience, 22% (n=446) of whom visited. 100% of visitors provided a qualitative response about the experience. In the fully-adjusted model, older age (ORs 2.39-5.10 for 55-75+ years old vs <25), non-Hispanic Black race/ethnicity (OR 1.64 vs NH White), those financially affected by the incarceration (ORs 2.54-2.88 for those negatively affected & 4.83-5.10 for those positively affected vs. not affected), and those whose family member was incarcerated a longer time (ORs 6.96-18.19 for duration >1 month-10+ years vs <1 month) had significantly greater odds of visitation. Qualitative analysis revealed 5 major themes: “Negative emotional reaction”, “Outrage at unjust treatment of incarcerated person”, “Negative logistics”, “Positive experience”, and “Negative feelings about crime/incarcerated person”. Only 8.1% of responses included a code in the “Positive experience” theme, whereas the vast majority consisted of a combination of negative themes.

CONCLUSIONS: Despite evidence of the benefits of visitation for incarcerated people and society as a whole, our analysis reveals harms for non-incarcerated family members doing the visiting. These harms are not equally distributed, given the substantial racial and socioeconomic patterning of incarceration. Further, only about one in five people with an incarcerated immediate family member visit at all. Given the negative health effects of incarceration itself; having an incarcerated family member; and living in a community with high rates of incarceration, further research should be done on how to improve the experience of visitation, and policy makers should act to improve visitation logistics.

THE IMPACT OF THE COVID-19 PANDEMIC ON COLORECTAL CANCER SCREENING PREVALENCE IN BOSTON NEIGHBORHOODS

Roopa Bhat¹; Suzanne Brodney¹; Yuchiao Chang¹; Meghan Rieu-Werden¹; Folasade May³; Jennifer Haas²

¹General Internal Medicine, Massachusetts General Hospital, Boston, MA; ²Medicine, Massachusetts General Hospital, Boston, MA; ³Medicine, University of California Los Angeles, Los Angeles, CA. (Control ID #3851585)

BACKGROUND: Colorectal cancer (CRC) is the third leading cause of cancer-related death in the United States. The prevalence of CRC screening in a neighborhood can reflect unique challenges in accessing screening tests. The COVID-19 pandemic has inhibited CRC screening due to colonoscopy cancelations, one of the primary CRC screening methods, and the limited screening capacity of health systems. The study’s objective was to identify how neighborhood-based disparities in CRC screening prevalence changed during the COVID-19 pandemic.

METHODS: We abstracted CRC screening completion from Massachusetts General Brigham electronic health records for adults due for CRC screening in 27 Boston neighborhoods on 3/1/2020 and 3/1/2022. We calculated whether individuals ages 50-75 were up to date with CRC screening, defined by completion of a fecal immunochemical test (FIT) within 1 year, FIT-DNA within 3 years, sigmoidoscopy within 5 years, or colonoscopy within 10 years. Neighborhoods were characterized by their Center for Disease Control Social Vulnerability Index (SVI), an aggregate measure of socioeconomic status, household characteristics, and race/ethnicity. We examined disparities in CRC screening prevalence using multivariable logistic regression models adjusted for SVI.

RESULTS: Of the 15,762 eligible patients in 2020, 10,253 (65.0%) were up to date. Of the 18,464 eligible patients in 2022, 10,646 (57.7%) were up to date. Results showed that screening prevalence dropped to a similar range in all neighborhoods regardless of pre-pandemic prevalence. There were no statistically significant differences in CRC screening prevalence by SVI: a decrease of 8.3% among neighborhoods with SVI ≤80 compared to a decrease of 6.0% among the neighborhoods with SVI >80 (p=0.56), indicating that neighborhoods with higher vulnerability were not impacted differently than neighborhoods with lower vulnerability. Overall, our results suggest that the COVID-19 pandemic uniformly depressed CRC screening prevalence in all neighborhoods.

A complete map of CRC screening prevalence by neighborhood will be displayed in the presentation.

CONCLUSIONS: Our findings reveal that the COVID-19 pandemic equalized the prevalence of CRC screening across Boston neighborhoods despite pre-existing disparities. As we recover from the pandemic, we must “build back equal” by developing effective strategies to ensure equitable participation in screening.

THE IMPACT OF THE HEALTH TECHNOLOGY NAVIGATORS ON PATIENT PORTAL UPTAKE- A NOVEL HEALTH WORKFORCE ADDRESSING THE DIGITAL DIVIDE IN A SAFETY-NET HEALTH SYSTEM

Cristina Valdovinos^1,2; Jihan Awad³; Veronica Penate³; Anshu Abhat^3,4; Alejandra Casillas⁵

¹University of California Los Angeles David Geffen School of Medicine, Los Angeles, CA; ²Charles R Drew University of Medicine and Science College of Medicine, Los Angeles, CA; ³Office of Patient Access, Los Angeles County Department of Health Services, Los Angeles, CA; ⁴Internal Medicine, Harbor-UCLA Medical Center, Torrance, CA; ⁵Medicine, Division of General Internal Medicine and Health Services Research, University of California Los Angeles David Geffen School of Medicine, Los Angeles, CA. (Control ID #3875299)

BACKGROUND: Significant disparities in the uptake of digital health tools have been previously described. Studies have found that low perceived confidence to engage with digital health resources among medically underserved patients partially account for this. To address barriers to digital health engagement (specifically around patient portal enrollment and use), the Los Angeles County Department of Health Services (LAC DHS) introduced the Health Technology Navigators program. To our knowledge, this is the first formal digital health workforce implemented in a safety net system. The study objectives are to describe the core functions of the health technology navigators and to evaluate their early effects on patient portal uptake.

METHODS: The Health Technology Navigators program was introduced at various sites in the LAC DHS in November 2021. The program is based on the community health worker model, with many navigators being bilingual (Spanish) members from the local communities home to LAC DHS sites. Navigator essential roles include helping patients enroll in the patient portal and download the portal application, and teaching patients how to navigate patient portal functions.

From January 2022-October 2022, 11 navigators worked at 11 LAC DHS sites (hospital and ambulatory care settings). We tracked unique portal registration invitations initiated by navigators, as well as portal account activations during this period. We compared the percent of invitations converted into successful account activations by a navigator versus conversion rates by all other staff types combined (e.g. medical assistants, clerks, nurses, doctors, volunteers). We examined age and language trends of patients invited into the patient portal by navigators.

RESULTS: A total of 72,474 invitations to enroll in the patient portal were sent during the study period. Navigators sent 8,053 unique invitations and 69% resulted in successful account activations. Only 49% of invitations sent by other staff resulted in successful account activations (p<0.05). Among patients enrolled by navigators, there were no noted trends by age: 67% of invitees who were 18-35 years activated an account versus 70% and 65 % for those 36-55 and 56-75 years old, respectively. Of note, Spanish-speaking patients had a higher rate of successful account activations (75%) than English-speaking patients (62%) (p<0.05)

CONCLUSIONS: The Health Technology Navigators Program is an innovative, patient-centered intervention designed to increase digital health engagement among patients in a safety-net health system. The program has been effective in increasing patient portal enrollment across different age groups and especially effective with Spanish-speaking patients. More research will be needed to examine how health technology navigators may affect sustained enrollment and portal use.

THE INFLUENCE OF GENDER, RACE, AND THE COVID PANDEMIC ON THE EXPERIENCES OF ASIAN AMERICAN AND PACIFIC ISLANDER WOMEN HEALTHCARE WORKERS: A QUALITATIVE STUDY

Qian Leng¹; Stephanie Maeda³; Megan Chang³; Monica Pan³; Samuel T. Edwards²; Megan Lafferty²; Belinda McCully³

¹CIVIC, Portland VA Medical Center, Portland, OR; ²Center to Improve Veteran Involvement in Care (CIVIC), VA, Portland, OR; ³Western University of Health Sciences College of Osteopathic Medicine of the Pacific, Lebanon, OR. (Control ID #3870988)

BACKGROUND: In the US, there are 1.17 million Asian American & Pacific Islander (AAPI) women who work in health care. Studies show that microaggressions and discrimination impact workers’ sense of safety, wellness, and burnout. Despite AAPIs making up a large portion of healthcare workers, with 1 in 5 physicians and 1 in 11 nurses identifying as AAPI, there has been little research into AAPI women’s experiences working in health care. Our study investigates the experiences of AAPI women healthcare workers with attention to how gender, race and the Covid pandemic have shaped their day-to-day interactions with other healthcare workers and patients.

METHODS: We conducted semi-structured interviews with 22 participants who identified as AAPI women, resided in WA or OR, and were employed as a CNA, MA, RN, NP, or physician. The interviews were held via videoconferencing and lasted an average of 65 minutes. Interviews were recorded, transcribed verbatim and uploaded to Taguette software for coding and analysis. Each transcript was coded for content and analyzed for themes by a two-member dyad of the research team. The dyad met on a bimonthly basis to discuss the content of each transcript and identify emerging themes, which were then refined through discussion with the rest of the research team.

RESULTS: Our study revealed that both microaggressions and overt discrimination are commonly experienced by AAPI women in health care. Microaggressions centered around the following themes: Patient perceptions of lower competence and respect for AAPI women providers, stereotyping as the perpetual foreigner, sexualization of participant’s race, and expectations that AAPI women be willing to work harder without additional compensation or recognition. Experiences of overt discrimination included patient refusal of care, harassment regarding national origin or accent, exclusion from leadership positions, and threats of violence.

Individuals who experienced these negative encounters reported anxiety, decreased work satisfaction, and about half of participants felt it impacted their decision of whether to stay with or to leave their current employer. Interestingly, the Covid pandemic had negative effects, such as the scapegoating of AAPI health workers, but participants also reported some positive changes, such as more opportunities for remote work and the perception that anti-Asian racism has been brought to light as an issue of national importance.

CONCLUSIONS: Microaggressions and discrimination based on gender and race are often experienced by AAPI women healthcare workers. These frequent, negative interactions affect workers’ morale and employee retention. We need evidence-informed efforts by organizations to decrease the frequency of microaggressions and boost inclusivity of AAPI women in health care.

THE LOW-INCOME HOUSING TAX CREDIT (LIHTC) MODERATES ASSOCIATIONS BETWEEN RACE/ETHNICITY AND HYPERTENSION IN CHICAGO

Tiffany H. Xie, Monica E. Peek, Corey Tabit, Elizabeth L. Tung

Medicine, University of Chicago Division of the Biological Sciences, Chicago, IL. (Control ID #3872165)

BACKGROUND: Housing instability is a social risk factor for racial inequities in hypertension. Literature has focused on pathophysiological mechanisms between housing instability and hypertension, emphasizing stress pathways. Few studies have examined effects of housing policy on health. This study investigates the Low-Income Housing Tax Credit (LIHTC), the largest U.S. program incentivizing affordable housing development, as a moderator between race/ethnicity and hypertension.

METHODS: We conducted a retrospective cohort study of 15,340 patients and 161,451 outpatient encounters at an academic medical center. We paired electronic health records (2018-2019) with LIHTC data from the U.S. Department of Housing and Urban Development (2009–2019). We geocoded patient addresses to census tracts and created a binary variable based on the presence of LIHTC in a tract. We used mixed- effects logistic regression models to examine hypertension (BP≥140/90) as a function of the interaction between race/ethnicity and LIHTC availability, adjusting for age, sex, insurance status, and neighborhood poverty. We also examined free/low-cost health resources, such as health education classes, gyms, and grocery stores, as a function of LIHTC.

RESULTS: In our population, the mean age was 50 years and 63.2% were female; 57.6% were non-Hispanic Black, 23.6% were non-Hispanic White, and 4.3% were Hispanic/Latinx. Only 9.9% lived in a census tract with LIHTC units and 54.8% lived in a poor neighborhood (>20% households in poverty). 29.3% of patients had hypertension. Black (OR = 2.6, 95% CI = 2.3–2.9) and Latinx (OR = 1.3, 95% CI = 1.0–1.5) patients had higher hypertension odds compared to White patients. We observed a significant interaction effect for Black patients: living in neighborhoods with LITHC reduced the odds of hypertension by 31% (aOR = 0.7, 95% CI = 0.5–1.0). However, the odds of hypertension were higher for Black patients compared to White patients (aOR = 1.9, 95% CI, 1.3–2.8) in LIHTC neighborhoods. Neighborhoods with LIHTC had 2.2 times higher odds of having free/low-cost health resources versus those without (95% CI = 1.3–3.5).

CONCLUSIONS: LIHTC may moderate hypertension inequities in Black neighborhoods, perhaps by promoting housing stability and health resources. Aligning healthcare goals of addressing individual social risks (e.g. housing instability) with policy that mitigates historical effects of discrimination (e.g. residential redlining that led to segregated, under-resourced neighborhoods) will be critical to achieving health equity for racially minoritized communities, particularly for pervasive, pernicious diseases like hypertension.

THE POWER INITIATIVE: IMPLEMENTING AN ANTI-RACISM CURRICULUM FOR MEDICAL RESIDENTS

Nisha Sunku¹; Daniel Georgie²

¹Medicine, UCLA Medical Center Olive View, Sylmar, CA; ²Internal Medicine, UCLA Medical Center Olive View, Sylmar, CA. (Control ID #3875483)

BACKGROUND: When examining the social determinants of health, racism has only been recently incorporated into the discussion. While there is still more research to be done at the intersection of racism and healthcare, it is already well established that racism leads to poorer health outcomes and perpetuates health inequities. Although racism has been declared a public health crisis, the medical education system is lacking on how to train physicians to be antiracist. Health professionals are calling for antiracism to be integrated into medical training. Some programs have begun to implement such curricula and have found that longitudinal teaching can be effective. At our institution, members of the Internal Medicine faculty implemented a novel anti-racism curriculum, with aspirations to educate ourselves and the internal medicine residency program. Through this initiative, we strove to shed light on racist practices and biases in medicine as a means of providing more equitable healthcare for our patients.

METHODS: A comprehensive anti-racism curriculum was implemented for the Internal Medicine Residency Program at the start of the 2021-22 academic year. The curriculum consisted of ten, 45-minute sessions covering a breadth of topics including privilege and disadvantage, biases in medical documentation, medical mistrust, race-based medical practices, and immigration health. An anonymous feedback survey was conducted at the end of the year to assess curriculum reception and ease of clinical application. This study looks at 17 participant responses from that survey. The survey prompts assessed curriculum content, participant reception (feeling included and respected), and applicability of the curriculum to clinical practice. RESULTS: The primary outcome was to assess if this curriculum would influence clinical practice positively. The secondary outcomes assessed if the sessions met the objectives, if the content was clear and organized, and if facilitators were knowledgeable about their topics. 100% of survey participants felt that these sessions would help improve their clinical practice. 100% strongly agreed that the facilitators were knowledgeable. 94% agreed that the content was clear and 82% agreed it met the objectives

CONCLUSIONS: Based on the survey responses, we found that this curriculum was well received and would positively impact clinical practice and patient care. The supporting data was also reflective of widespread acceptance. Moving forward, we hope to expand this lecture series into a three-year rolling curriculum that other residency programs could easily adopt at their institutions.

TRAINING PRIMARY CARE RESIDENTS TO PROVIDE MEDICATIONS FOR STIGMATIZED CONDITIONS: A NATIONAL SURVEY OF PROGRAM LEADERS

Nina Tan¹; ADELAIDE H. MCCLINTOCK¹; Helene Starks³; E. B. Schwarz²

¹Internal Medicine, University of Washington Department of Medicine, Seattle, WA; ²Internal Medicine, University of California San Francisco, San Francisco, CA; ³University of Washington School of Medicine, Seattle, WA. (Control ID #3864610)

BACKGROUND: Stigma in health care impairs diagnosis and treatment and adversely affects health outcomes. As quality healthcare requires thoughtful training to minimize stigma, we surveyed residency program directors to compare approaches to training primary care residents in the use of medications to treat four stigmatized conditions that impact large numbers of US patients: opioid use disorder (OUD), high risk sexual activity warranting pre-exposure prophylaxis for HIV, gender affirming care, and undesired pregnancy. METHODS: We identified 244 program leaders of internal medicine (IM) primary care residency programs using the American Medical Association’s FREIDA database. We distributed anonymous unique links to a 15-question, institutional review board approved REDCap survey to program leaders, offering a $35 gift card for survey completion. We sent five reminders to encourage survey completion.

RESULTS: Surveys were completed by 68 program leaders (28% response rate). The majority of respondents felt that medications to manage these conditions fall within the scope of primary care (96% OUD, 100% HIV prevention, 72% gender affirming care, 63% undesired pregnancy) and recognized that patients currently have difficulty accessing these medications (88% OUD, 72% HIV prevention, 87% gender affirming care, 87% undesired pregnancy). While most programs reported training residents to prescribe medications for OUD (72%) and HIV prevention (84%), and 47% provide training in gender-affirming care, only 19% provide training in managing undesired pregnancy with over half of programs reporting no plans to offer this training (51%). The most common barrier to providing training was “lack of trained faculty preceptors” with over 50% of programs reporting that less than 10% of faculty were trained to manage gender affirming care and undesired pregnancy.

CONCLUSIONS: In this national survey of IM primary care residency program leaders, we found that most believe that medications for common stigmatized conditions are within the scope of primary care and that access to medications for these conditions remains limited. However, while most programs train residents to prescribe medications for OUD and HIV prevention, few offer training on medication abortion for undesired pregnancy with a commonly cited barrier being lack of trained faculty preceptors. This suggests that limited access to primary care for stigmatized conditions may be perpetuated and amplified without adequately trained physicians in the rising workforce. It also suggests that faculty development to learn and teach new clinical skills, such as medication management of gender affirming care and undesired pregnancy, may help improve residency training and primary care of these common stigmatized conditions.

TRANSITIONING HOME: EXPLORING PATIENT SOCIAL RESOURCE ENGAGEMENT AT DISCHARGE

Sonja Raaum¹; Alycia A. Bristol²; Kirsten Schmutz²; Katie Feldner²; Erin P. Johnson²; Andrea Wallace²

¹Internal Medicine, University of Utah Health, Salt Lake City, UT; ²College of Nursing, University of Utah Health, Salt Lake City, UT. (Control ID #3876043)

BACKGROUND: Hospital discharge is a vulnerable time for patients that requires engaging social resources to effectively self-manage medical conditions. COVID-19 disrupted patients’ social networks, but factors impacting patients’ preparation and navigation of the discharge process during COVID-19 have not been explored. The purpose of this study was to explore patient experiences with hospital discharge, the transition to home, and how COVID-19 impacted post-discharge processes.

METHODS: As part of a parent study (n=3815) examining the effect of implementing a systematic assessment of social needs and support in acute medical-surgical settings, patients were asked for their willingness to be interviewed 1-2 months after discharge. 38 interviews were conducted in April-May 2020 that concerned discharges from February-March 2020. Interviews were recorded and transcribed. Qualitative content analysis was performed using individual and family self-management theory (IFSMT) to code context and process factors that patients identified. Each transcript had a primary and secondary coder to reach a consensus, and subsequent coder meetings identified themes.

RESULTS: Several themes were identified according to the IFSMT framework. Themes that impacted patient self-management in the context of discharge planning included: timing discharge education to allow presence of support system, clearly communicating pain management expectations, and considering discharge location distance from the hospital. Those who expressed needs related to social resources were more likely to describe challenges associated with the discharge experience. Once home, COVID impacted the process of self-management by way of paid and unpaid support. For paid support, most valued the ability to meet via telehealth and to stay at home, but some reported difficulties accessing adequate support services (e.g., physical therapy). For personal support, several noted that an unexpected loss of support from family or friends would hinder self-management and ability to cope. Throughout, patients provided examples of unexpected support and concrete examples of how personal networks aid patients in anticipating needs in the context of transitions from hospital to home.

CONCLUSIONS: COVID-19 provided an opportunity to explore the role of social resources in successful patient transitions home. Our findings identified specific areas where the healthcare team can improve patients’ self-management – prioritize having support present during discharge, robust anticipatory planning around pain management and logistics, and explore reliance on both paid and unpaid support at home. Anticipating medical needs remains elusive for many patients but is key to a successful transition home from the hospital.

TRENDS IN FOOD INSECURITY AMONG ADULT PATIENTS BEFORE, DURING, AND AFTER THE NEW YORK CITY COVID-19 SHUTDOWN

Ines M. Robles Aponte^1,2; Ngoc Duong^1,2; Dodi Meyer^1,2; Nathaniel Kratz^1,2; Sandra S. Albrecht³; Maria Baquero⁴; Aldo Crossa⁴; Francesca Gany⁵; Elizabeth Kelman⁴; Morgan A. Finkel^1,2; Evianna Cruz Herrera^1,2; Amelia Shapiro²; Patricia Peretz²; Ying Kuen K. Cheung^1,2; Jennifer A. Woo Baidal^1,2

¹Columbia University, New York, NY; ²NewYork-Presbyterian, New York, NY; ³Mailman School of Public Health at Columbia University, New York, NY; ⁴NYC Department of Health and Mental Hygiene, New York, NY; ⁵Memorial Sloan Kettering Cancer Center, New York, NY. (Control ID #3855372)

BACKGROUND: Food insecurity has consistently been associated with adverse health outcomes, but sparse information exists on food insecurity among broad adult patient populations. We tested the hypotheses that the proportion of adult patients with food insecurity would increase during the COVID-19 shutdown anddecrease after the shutdown. We also examined differences in food insecurity according to race/ethnicity.

METHODS: Using an interrupted time series design, we examined electronic health records in an urban academic multi-site healthcare system in New York City (NYC) predominantly serving Hispanic/Latino neighborhoods with high food insecurity prevalence. We included a sequential sample of patients age >18 years with a Hunger Vital Sign™ food insecurity screening questionnaire completed as part of routine clinical care between March 2019 and June 2022. We defined food insecurity as a response of “always” or “sometimes” to either of 2 questions. We estimated monthly food insecurity as the proportion of screened patients with food insecurity. We defined three time points: before (Mar 2019-Feb 2020), during (Mar 2020- May 2020), and after (June 2020-June 2022) the COVID-19 NYC shutdown. We fit segmented regression models to estimate monthly food insecurity and changes over time with clustered standard errors, adjusting for patient demographics (age, sex, and race/ethnicity). We tested for effect modification by race/ethnicity using a 2-way interaction term between race/ethnicity and change over time (continuous months) in adjusted models.

RESULTS: Among 47,566 eligible patients, mean age was 43.6 ±18.4 years; 78.1% were female. 64.2% reported Hispanic/Latino, 10.6% non-Hispanic Black, 7.4% non-Hispanic white, and 5.6% other race/ethnicity. Patients completed mean 2.1 ± 1.7 food insecurity screens over 3 years. Average monthly food insecurity was 27.6% overall (32.1% before, 32.8% during, and 24.8% after the shutdown). In adjusted models, food insecurity increased 2.4% per month on average (95% CI: 0.3%-4.4%) during the shutdown compared to before. After the shutdown, food insecurity decreased by 0.1% per month (95% CI: -0.3% to 0.2%) compared to before the shutdown. After the shutdown, food insecurity was higher among Hispanic/Latino and non-Hispanic Black patients than non-Hispanic white patients (26.6%, 24.5%, and 15.0%, respectively; p-value < 0.001). Food insecurity among Hispanic/Latino and non-Hispanic Black patients decreased less than non-Hispanic white patients after the shutdown (p-interaction 0.003 and 0.03, respectively).

CONCLUSIONS: Despite overall improvements in food insecurity, racial/ethnic disparities in food insecurity widened after the COVID-19 shutdown. Clinical screening can be used to identify patient populations disproportionately burdened by food insecurity for future interventions.

UNCOVERING AND MITIGATING RACIAL BIAS IN MORTALITY PREDICTION MODELS USING HEALTH CARE DATA

Amol S. Navathe^1,2; Kristin A. Linn^3,2; Sae-Hwan Park¹; Caleb M. Hearn¹; Matthew Maciejewski⁴; Craig J. Kreisler⁵; Christopher B. Roberts⁶; Sumedha Chhatre⁷; Mateo U. Rosania²; Ravi B. Parikh¹

¹Medical Ethics and Health Policy, University of Pennsylvania, Philadelphia, PA; ²Leonard Davis Institute of Health Economics, Philadelphia, PA; ³Biostatistics, Epidemiology, and Informatics, University of Pennsylvania, Philadelphia, PA; ⁴HSR&D, Durham VA Medical Center, Durham, NC; ⁵VA Medical Center Corporal Michael J Crescenz, Philadelphia, PA; ⁶VA Center for Health Equity Research and Promotion, Pittsburgh, PA; ⁷Psychiatry, University of Pennsylvania, Philadelphia, PA. (Control ID #3875913)

BACKGROUND: Mortality risk prediction models are used widely in health care for risk-stratification and risk-adjustment. Such models may be vulnerable to bias against minority groups. We investigated racial bias in population-level mortality prediction models used by the Veterans Administration (VA), Medicare, and other policymakers.

METHODS: Our initial focus was the VA Care Assessment Needs (CAN) score, which primary care teams use in patient management and resource allocation. The mortality CAN score predicts 1-year all-cause mortality for Veterans using more than 100 clinical and administrative variables. First, we investigated the CAN for 2 common bias metrics: (1) equality of opportunity, which requires equivalent false negative rates (FNRs); and (2) demographic parity, which requires equivalent proportions of Black and White patients to be classified as high-risk. Second, we systematically applied several literature-based methods to mitigate observed bias toward Black patients. Third, we simulated the impact of an “Equitable CAN” score on the racial distribution of palliative care referrals. Fourth, we examined other mortality risk scores (Medicare Hierarchical Condition Categories (HCC), Elixhauser comorbidities) in VA and Medicare data for similar racial bias.

RESULTS: We found consistent evidence of inequity in predictions against Black patients across both bias metrics. Among 989,977 Black and 3,945,605 White Veterans, mean CAN score was lower for Blacks (41.6 vs. 52.1; P <0.001). Using a CAN threshold of 75 to indicate high risk, the FNR was 53% higher for Black vs White Veterans (29.6% vs. 19.4%; P<0.001). Age was a leading mechanism of bias: Black Veterans were younger than Whites (mean age 58.2 vs. 64.5 years, p<0.001), despite similar comorbidity (mean comorbidity number 2.5 vs 2.3). Black Veterans died at younger ages than Whites (72.0 vs. 78.6 years, P<0.001). An "Equitable” CAN that normalized age distributions to a common mean resulted in similar CAN scores (mean score difference of original vs. Equitable CAN: 10.3 vs. 3.0) and mitigated inequity (FNR difference: 10.3 vs. 1.3 percentage-points [pp]), while other statistical approaches did not. In simulations, an Equitable CAN increased the percentage of Black Veteran decedents who would receive palliative care by 8.1pp from a baseline of 15.5% (a 52% increase). Our findings replicated across other risk-scores and populations (e.g., Original vs. Equitable HCC FNR gap 4.1 vs. 0.1pp).

CONCLUSIONS: We uncovered a generalizable mechanism of inequity in mortality prediction related to faster aging of Black individuals – a “weathering effect” likely related to adverse social drivers of health. Age normalization markedly reduced bias and is currently being tested in the VA to ameliorate disparities in areas like palliative care and hospice referral. Our findings highlight pervasive unintended racial bias against.

Blacks and inform mitigation approaches in other mortality prediction models.

UNEVEN VA MENTAL HEALTHCARE EXPERIENCES AMONG TRANSGENDER VETERANS WITH DEPRESSION

Joy L. Lee¹; Claire E. Williams³; Masheka Fuqua⁴; Diana Natividad⁴; Michael Weiner²

¹General Internal Medicine, Indiana University School of Medicine, Indianapolis, IN; ²Department of Medicine, Indiana University, Indianapolis, IN; ³Health Services Research, Regenstrief Institute, Indianapolis, IN; ⁴Richard L Roudebush VA Medical Center, Indianapolis, IN. (Control ID #3874746)

BACKGROUND: Transgender Veterans experience higher rates of depression than cisgender Veterans. They also often report delaying seeking mental healthcare, citing fear of discrimination. The VA, the largest integrated healthcare system in the US, has made a stated commitment toward reducing health disparities for transgender Veterans and delivering gender-affirming care. In this study, we describe the mental healthcare experiences of transgender Veterans with depression, including identifying facilitators and barriers toward the delivery of gender-affirming care.

METHODS: We conducted in-depth, semi-structured interviews with transgender Veterans with depression to understand their mental healthcare experiences within and outside of the VA. Data collection took place in 2021 and included Veterans from different US regions. Each interview was audio-recorded, transcribed, and coded for themes using an immersion/crystallization approach.

RESULTS: Twenty-six Veterans with depression participated in the study. The participants appeared very knowledgeable about the parameters of VA care for transgender Veterans, often quoting exact VA directives on available services. Health system facilitators cited included the LGBTQ+ program, mental health provider availability (relative to non-VA providers), and ease of connecting with services. Barriers included perceived lack of provider knowledge about transgender experiences, and lack of available providers. Interpersonal facilitators included participant experiences with a wide range of involved specialties (e.g. speech pathologists, physical therapists, mental health counselors) with providers who went “above and beyond” to deliver care, while barriers included Veteran experiences with mis-gendering and misnaming by VA staff. Given that study enrollment took place during the COVID-19 pandemic, many participants had experience with receiving care through video visits. Video visits were valued not only for their convenience, but the “safe space” that they created. One patient said, “Being able to do [video visits] in the space where I’m most comfortable, I’ve been able to express myself a lot easier. It makes for a more productive session. When you go somewhere that’s not your personal space, I never felt comfortable just sharing things.”

CONCLUSIONS: While the VA has made a commitment toward improving care experiences for transgender Veterans, mis-gendering by VA staff remains common. Notably, clinician availability and knowledge were cited by some participants as facilitators and some as barriers to care, suggesting system (clinician availability) and interpersonal (clinician knowledge) variations across VA sites. Greater availability of mental healthcare providers and education on gender affirming mental healthcare may be needed to improve care experiences.

USE OF COMPANIONS WHO SUPPORT MEDICAL COMMUNICATION AMONG OLDER ADULTS WITH COMMUNICATION DISABILITIES

Jennifer Y. Osh*ta

Medicine, Clinical and Translational Sciences, University of Vermont Larner College of Medicine, Burlington, VT. (Control ID #3876209)

BACKGROUND: Persons with communication disabilities (CDs) experience multiple healthcare disparities including increased likelihood of emergency department visits, increased risk of an adverse medical event, and decreased satisfaction with care. Prior literature indicates that companionship during medical encounters increases patient satisfaction, autonomy related behaviors and patient safety. Little is known about utilization of companions who aide medical communication among older adults with CDs. We aimed to: 1) Describe the prevalence of using a companion who aids communication at doctors’ visits among older adults, by no, single, and multiple CDs; 2) Compare the adjusted likelihood of bringing a companion by presence of CDs; 3) Identify social factors impacting use.

METHODS: Cross-sectional analysis of the 2015 cohort of the National Aging Health and Trends Survey, a longitudinal, nationally representative survey of Medicare beneficiaries. The predictor of interest was the presence of no, single, or multiple CDs (hearing, expressive and/or cognitive). The outcome was bringing someone to the doctor who helped with communication. NHATS analytic weights were utilized. Descriptive statistics were used in the unadjusted analysis. We used a generalized linear model with a Poisson family distribution to obtain an adjusted prevalence ratio (APR), while controlling for sociodemographics, health, and other disability factors. As a secondary analysis, we stratified the population by race, gender, and income, to assess if they moderate companion use. Finally, we calculated adjusted predictive probabilities using the Stata margins command.

RESULTS: An estimated 60% (CI: 56.9-62.9) of older adults with CDs do not bring someone with them who aides communication to doctors’ visits. After adjustment, having CDs increases the likelihood of companionship when compared to no CD (Single CD APR: 1.40 [CI 1.25-1.57]; Multiple CDs APR: 1.47 [CI 1.26-1.72]). However, when stratified by social groups the predicted probabilities of using a companion differed from the entire population. Older black adults were less likely than white adults to use a companion

(No CD: Black 19.2%, White 21.9; Single CD: Black 24.8%, White 30.7; Multiple CDs: Black 24.3%, White 32.6%). While having more than one CD increased the predicted probability of using a companion among all groups, among older black adults, those with multiple CDs were no more likely to use a companion when compared to those with single CDs.

CONCLUSIONS: Slightly over half of older adults with CDs do not bring a communication companion to medical visits. Beyond presence of CDs, social factors also impact use of a companion.

UTILIZING A REMOTE POVERTY SIMULATION IN MEDIAL EDUCATION

Bethany M. Langner

Shilpa Register, Dawn Taylor Peterson. The University of Alabama at Birmingham Heersink School of Medicine, Birmingham, AL. (Control ID #3860021)

BACKGROUND: Social determinants of health have a complex impact on patient care and outcomes.

Therefore, an understanding of poverty and how it relates to patient care is critical to medical student training. With the transition to online platforms in education, a remote version of the poverty simulation allows for a more cost effective and time efficient mode of delivery than an in-person poverty simulation

METHODS: A remote version of the poverty simulation was implemented for first year medical students at the beginning of their training. The learning experience consisted of content review and a simulation prebriefing lasting 30 minutes, the virtual simulation which lasted 30 minutes, and a structured debriefing led by trained faculty members lasting 45 minutes. All of these activities were online. In the simulation experience, students were provided $1000 to get through the month, which forced them to make financial compromises. Students were allowed to redo the simulation and make different choices to see how this altered the outcome of their living situation. After the virtual simulation, students engaged in a live, structured debriefing session over a web conferencing platform with trained faculty members. A qualitative survey analysis was used to examine anonymous survey data from medical students that participated in the poverty simulation.

RESULTS: A total 337 students from two different cohorts completed the poverty simulation and post-test surveys, which consisted of nine Likert scale questions and two open-ended questions. Students overwhelmingly agreed or strongly agreed that objectives for the event were met (97%), the learning experience was valuable (96%), the feedback session was valuable (94%), the experience will improve their performance in a clinical setting (92%), and they would recommend this event to others (94%). Statements such as, “I learned about the difficulties of being moral and ethical when survival is difficult to achieve” and “the simulation was more engaging than just reading studies” were common in the post-test surveys. A theme for improvement in the simulation was to “include mental health and/or addiction,” while another student said, “I wish the simulation could have been lengthened to see longer-term outcomes.”

CONCLUSIONS: Implementing a poverty simulation during the first year of medical school ensures that medical students receive this knowledge early in their training even before they see their first patient. A barrier to offering these critical simulations in-person is coordinating time with preceptors and participants to participate in the simulation. We found that a remote simulation can overcome these barriers and still provide appropriate training to students.

VA FACILITY-LEVEL VARIATION IN RACIAL DISPARITIES IN ANTICOAGULATION FOR ATRIAL FIBRILLATION

Utibe R. Essien, Nadejda Kim, Leslie R. Hausmann, Maria K. Mor, Chester ". Good, Terrence Michael A. Litam, Walid F. Gellad, Michael J. Fine

Center for Health Equity Research and Promotion, VA Pittsburgh Healthcare System University Drive Division, Pittsburgh, PA. (Control ID #3874307)

BACKGROUND: Although anticoagulation reduces stroke risk in atrial fibrillation (AF), prior nationwide VA studies conducted at the patient-level show significantly lower prescribing rates for any anticoagulant and direct oral anticoagulant (DOAC) therapy in Black than White individuals. Because little is known about whether disparities in anticoagulant prescribing for AF exist at the facility level, we compared anticoagulant initiation for White and Black patients with AF across VA facilities.

METHODS: We identified a national cohort of patients enrolled in VA with a new outpatient diagnosis of non-valvular AF from 1/1/2020 to 12/31/2021. We excluded patients with an AF diagnosis or anticoagulant therapy in the 2 years prior to their index AF diagnosis, pre-existing valvular heart disease, and those receiving hospice care or who died within 180 days of an index AF diagnosis. For all VA facilities with at least 10 Black patients with AF, we calculated facility-level rates of anticoagulant initiation (any or DOAC) within 180 days of an index diagnosis and assessed the difference in anticoagulant initiation between White and Black patients. We characterized variation across VA facilities using risk differences by percentile points, adjusted for year of AF diagnosis, area deprivation index, VA enrollment priority group, CHADS₂VA₂Sc stroke risk score, bleeding risk, and comorbid renal and liver disease.

RESULTS: We identified 28,434 patients with AF at 82 VA facilities with at least 10 Black patients with AF.

The facility-level ranges of the number of Black and White patients were 2.7% to 56.5%, and 37.0% to

96.1%, respectively. The facility-level percentages of patients initiating any anticoagulant therapy for Black and White patients were, 40.4% to 93.3%, and 51.0% to 88.0%, respectively; for DOAC therapy, these ranges were 40.0% to 90.9% for Black patients and 49.0% to 86.8% for White patients. The adjusted White-Black facility-level differences ranged from -29.3% to 13.8% for any anticoagulant therapy and -27.4% to 14.2% for DOAC therapy (Figure). Only 4/82 facilities had a statistically significant risk difference for either any anticoagulant or DOAC therapy, with 3 of 4 of these facilities showing higher prescribing frequencies in Black than White patients.

CONCLUSIONS: Despite wide variation in VA facility-level adjusted racial differences in any anticoagulant or DOAC therapy for patients with AF, statistically significant racial differences were uncommon and more often favored Black versus White patients at the facility-level.

VALIDITY EVIDENCE FOR THE PHYSICIAN ORAL LANGUAGE OBSERVATION MATRIX AS A MEASURE OF MEDICAL SPANISH PROFICIENCY

Pilar Ortega²; Steven Gregorich³; Leah S. Karliner⁴; Javier González⁵; Cristina Pérez-Cordón⁶; Reniell Iñiguez²; José Alberto Figueroa⁷; Karen Izquierdo¹; Lisa C. Diamond¹

¹Immigrant Health and Cancer Disparities Service, Memorial Sloan-Kettering Cancer Center, New York, NY; ²University of Illinois Chicago, Chicago, IL; ³University of California San Francisco, San Francisco, CA; ⁴Medicine, UCSF, San Francisco, CA; ⁵Immigrant Health and Cancer Disparities, Memorial Sloan Kettering Cancer Center, New York, NY; ⁶United Nations, New York, NY; ⁷Northwestern University, Evanston, IL. (Control ID #3875132)

BACKGROUND: The lack of a standardized language assessment process for medical students and physicians communicating in Spanish threatens patient safety and quality of care. Many medical students and physicians take medical Spanish courses and/or report having and using some Spanish skills. However, most never have their skills formally assessed and rely on self-reported proficiency, which can be inaccurate. The purpose of this study is to evaluate the validity of the Physician Oral Language Observation Matrix (POLOM) in assessing medical student communication with Spanish-speaking patients.

METHODS: Forty-one medical students completed standardized patient (SP) encounters in Spanish. The students’ performance on the videorecorded encounters was scored by trained, reliable raters. We examined the POLOM’s convergent validity via correlations with other medical Spanish proficiency measures: (1) self- rating on the Interagency Language Roundtable scale for healthcare (ILR-H) and (2) score on the Clinician Cultural and Linguistic Assessment (CCLA). We measured Criterion validity via the POLOM’s correlations with clinical outcomes of the students’ SP encounters: (1) an SP scale about the student’s communication/interpersonal skills, (2) an SP checklist about the completeness of the interview, and (3) a faculty rating of the clinical note the student wrote after the encounter. We report Pearson correlations (r), and—for criterion validity—canonical correlations (r) representing the association of each language proficiency measure with all three clinical outcomes, simultaneously.

RESULTS: Students in the sample were mostly female (56%), 48% white, 27% Hispanic/Latinx, 17% Asian, and 7% Black. The POLOM was strongly correlated with student ILR-H self-rating (r=.72) and CCLA Objective score (r=.59), providing evidence of convergent validity. In addition, the POLOM was substantially correlated with the SP scale (r=.35), the SP checklist (r=.32), and the faculty scoring of the student’s clinical note (r=.25), providing criterion validity evidence of POLOM scores. Of the three language proficiency measures, the POLOM had the stronger multivariate relationship with the three clinical outcomes, r=.46 versus r=.44 (ILR-H) and r=.41 (CCLA).

CONCLUSIONS: The POLOM is the first observational tool that can be used to assess oral Spanish proficiency during an SP encounter. The POLOM has demonstrated evidence of convergent and criterion validity as a measure of medical students’ Spanish proficiency during SP encounters. Compared to other available healthcare language examinations or tools, POLOM ratings were more strongly correlated with clinical objectives of medical Spanish curricula. Future work will expand the scalability of the POLOM to make it available as a medical Spanish proficiency assessment. Additional research is needed to evaluate how the POLOM can be implemented with resident and practicing physicians, applied to other health professions, and adapted to other languages.

VARIATION AND CHALLENGES IN PRE-HEALTH ADVISING RESOURCES IN CALIFORNIA’S INSTITUTIONS

Francine Rios-Fetchko¹; Mariam Carson¹; Manuel Tapia²; Janet Coffman³; Alicia Fernandez¹

¹Department of General Internal Medicine, ZSFG, University of California San Francisco, San Francisco, CA; ²Department of Family and Community Medicine, University of California San Francisco, San Francisco, CA; ³Institute for Health Policy Studies, University of California San Francisco, San Francisco, CA. (Control ID #3876717)

BACKGROUND: California’s health professionals do not reflect the state’s population growth and increased diversity. Pre-health advising presents a modifiable factor that can help increase the diversity of the health workforce. Colleges are known to vary in resources, structure, and services invested in pre-health advising, but data are lacking and there is no estimate of any pre-health advising resource gap. We surveyed all CSU (California State Universities), University of California (UC) institutions and select private schools to quantify any gap in resources and gather perceived challenges in pre-health advising.

METHODS: We conducted structured 60-minute interviews with closed and open-ended questions from June 2022 to October 2022 via videoconference with pre-health advisors from all CSUs and UC, and selected 6 private colleges with varying student selectivity to match UCs and CSUs. Two investigators independently analyzed interviews using a Grounded Theory Approach and the full team reviewed transcripts and themes.

RESULTS: 18/23 CSU, 9/9 UC and 6/6 private colleges participated. Of 4 non-participating CSUs, one is a dedicated non-pre-health campus. Pre health advisor FTE varied from mean CSU 1: 24,620 graduates (range: 1: 1,059 – 1: 150,520) to UC 1: 4,526 graduates (range: 1: 1,912 – 1: 10,920) and privates 1:1,794 graduates (range: 1: 722 – 1: 5,300). On average, UC schools had 5.4 times more FTE pre-health advising per graduate than CSU schools. Challenges across all institutions included: advising capacity, competing priorities, difficulties in accessing clinical opportunities for students, challenges in course work navigation specific to transfer students from community college, and challenges in identifying pre-health students at opportune times for advising input. These challenges were reported as particularly common and severe by CSU participants. Both CSU and UC participants noted that advising challenges were particularly salient for students from underrepresented racial/ethnic groups and first-generation college students where financial pressures, lack of access to academic opportunities earlier in their educational careers, and familial responsibilities complicate college navigation. All participants noted local innovations to better serve their students, including leveraging peer advisors, offering health professions career exploration courses, establishing partnerships with local physicians and organizations, and developing advisor networks

CONCLUSIONS: UC schools have over five times more FTE pre-health advising per graduate than CSU schools, and both California’s public systems have far fewer advisors per student than private colleges. While pre-health advising presents common challenges on all campuses, much greater investment is needed in California’s public institutions, particularly California State Universities, to increase equity in access to advising for pre-health professional students.

Scientific Abstract - Geriatrics and Palliative Care

ADVERSE CHILDHOOD EXPERIENCES (ACES): IMPLICATIONS FOR PHYSICAL, COGNITIVE, AND FUNCTIONAL IMPAIRMENT IN A NATIONAL SAMPLE OF OLDER COMMUNITY-DWELLING ADULTS

Victoria M. Lee¹; Anita S. Hargrave²; Nadra E. Lisha²; Alison Huang²

¹School of Medicine, University of California San Francisco, San Francisco, CA; ²Medicine, University of California San Francisco, San Francisco, CA. (Control ID #3854748)

BACKGROUND: Prior research on the health implications of ACEs has focused on early or midlife adults, not older adults who bear the greatest burden of health-related functional impairment. We examined associations between ACEs and mobility impairment, cognitive impairment, and functional disability in older community-dwelling adults.

METHODS: Cross-sectional analyses of the National Social Life, Health, and Aging Project (2015-2016), a national cohort of community-dwelling older U.S. adults. Participants were interviewed about their history of ACEs (childhood experience of violence/abuse, witnessing of violence, financial insecurity, parental separation, or serious illness), underwent standardized physical performance testing (tandem balance, 3-meter walk, chair stand test) and cognitive testing (survey adaptation of the Montreal Cognitive Assessment), and reported functional disability (difficulty with activities of daily living). Multivariable logistic regression models examined associations between ACE exposures and each functional impairment and functional disability outcome, adjusting for potential demographic confounders.

RESULTS: Among the 3387 participants (ages 50-97; 54% female; 26% ethnic minority), 44% reported a history of one or more types of ACEs, including 14% reporting childhood experience of violence and 16% childhood witnessing of violence. Thirty-five percent met criteria for mobility impairment and 24% cognitive impairment, and 24% reported functional disability. After adjusting for age, gender, race, and ethnicity, participants reporting exposure to one or more ACEs were more likely to demonstrate mobility impairment (OR 1.30, 95% CI 1.11–1.52) and cognitive impairment (OR 1.26, 95% CI 1.03–1.54), as well as report functional disability (OR 1.69, 95% CI 1.38–2.07), compared to those with no history of ACEs. Participants with a history of multiple simultaneous ACEs were even more likely to have mobility impairment (OR 1.37, 95% CI 1.10–1.72) or cognitive impairment (OR 1.38, 95% CI 1.07–1.79) and report functional disability (OR 1.97, 95% CI 1.59–2.43). In analyses distinguishing between specific types of ACEs, poor childhood health was associated with a 65% to 74% increased odds of almost all measured outcomes, and childhood experience of violence was associated with a 38% greater odds of physical mobility impairment (95% CI 1.11–1.71) and 86% increased odds of functional disability (95% CI 1.49–2.33).

CONCLUSIONS: Adverse childhood experiences may be important markers of future physical and cognitive functional impairment late in life, suggesting that efforts to screen for or mitigate ACEs may have implications for aging-associated functional decline. Clinicians may be able to provide more thoughtful and compassionate care to older adults by exploring the potential role of early life traumatic experiences in shaping or further complicating their functional challenges.

AN AUDIOLOGY AND GERIATRICS COLLABORATION TO IMPROVE HEARING CARE SERVICE UTILIZATION AMONG OLDER VETERANS

Lindsey Ulin¹; Hema Pingali¹; Andrea W. Schwartz^2,3; Michelle Martinchek²; Danielle L. Stiles²

¹Internal Medicine, Brigham and Women's Hospital, Boston, MA; ²VA Boston Healthcare System, Boston, MA; ³New England Geriatrics Research Education and Clinical Center, Boston, MA. (Control ID #3869246)

BACKGROUND: Hearing loss is associated with increased mortality, a higher risk of dementia and depression, and worse quality of life in older adults. Hearing loss is prevalent among older veterans but is undertreated. The Veterans Affairs (VA) health care system provides universal access to audiology care but many veterans may not be accessing benefits. We report lessons learned from a quality improvement initiative to increase the use of audiology services by veterans in a Geriatrics clinic.

METHODS: Geriatrics schedules were reviewed weekly by an audiology team to identify and electronically communicate to Geriatrics providers the audiology needs of veterans with upcoming appointments. Geriatrics providers could then discuss hearing care needs during the visit and place appropriate referrals including audiology and/or ear cleaning, if indicated. We reviewed 50 charts randomly selected via convenience samples from 2017-2018 to determine the impact of the audiology and geriatrics partnership on veterans' access to appropriate audiology care.

RESULTS: The audiology review of the 50 charts sampled revealed that 18 patients never had audiology testing (36%), and 14 patients had hearing aids older than 2 years and were due for updates (28%). Of the 50 patients reviewed, 10 (20%) no showed their Geriatrics appointment. Of the 40 who presented to Geriatrics, 8 received referrals to audiology (20%) and 3 received referrals for ear cleaning (8%). Of the 8 veterans referred for audiology, 2 no showed (25%), 1 did not qualify for hearing aids (13%), 1 declined hearing aids (13%), and 4 received new hearing aid prescriptions (50%). Of the veterans referred for ear cleaning, 1 no showed (33%). Of the 50 veterans, 39 had a diagnosis of dementia on their problem list (78%), 25 had a diagnosis of hearing impairment (50%), and 30 died during the follow-up period to 2022 (60%).

CONCLUSIONS: Hearing loss was highly prevalent in a sample of veterans in a Geriatrics clinic, most of whom had dementia. An Audiology-Geriatrics partnership led to provision of new hearing aids or ear cleaning in 5 veterans of the 50 charts reviewed (10%). However, audiology referrals were not consistently placed when indicated by audiology review, in some cases because of patients declining. There was also a 20% or greater no show rate to geriatrics and audiology appointments. This review highlights possible barriers to receiving audiology care among older veterans and suggests that even if care is free and accessible, a dedicated review to identify high-risk patients may be insufficient to ensure care is received.

High rates of dementia among the clinic population may contribute to high no show rates, as well as the need to address multimorbidity taking precedence over hearing impairment. Given hearing impairment’s impact on memory, mood, mobility, and more, and as hearing aids have recently been approved to be available over the counter, the need to characterize and address the barriers to hearing care is all the more urgent.

AN RCT OF GAMIFICATION WITH SOCIAL SUPPORT TO INCREASE PHYSICAL ACTIVITY IN OLDER ADULTS AT RISK FOR ALZHEIMER'S: THE STEP 4 LIFE TRIAL

Scott R. Greysen¹; Gavin Turner²; Ai Leen Oon¹; Kristin Harkins³; Jason H. Karlawish³

¹DGIM, University of Pennsylvania, Philadelphia, PA; ²University of Pennsylvania Perelman School of Medicine, Philadelphia, PA; ³Division of Geriatrics, University of Pennsylvania, Philadelphia, PA. (Control ID #3872361)

BACKGROUND: Higher levels of physical activity are associated with reduced risk of cognitive and functional decline in Alzheimer’s Dementia and Related Dementias (ADRD). Remote gamification interventions informed by behavioral science using participants’ social networks are effective in populations at risk for cardiovascular disease but have not been studied in populations at risk for ADRD. Innovative and scalable approaches that achieve sustained increases in physical activity have great potential to impact the risk of developing ADRD.

METHODS: 18-week RCT with 12-week intervention and 6-week follow-up using gamification with social incentives among participants already enrolled in a national cohort of older adults (age 55-75) at risk of developing ADRD. All participants (n=90) were supplied with a Fitbit device to track daily steps. The control group (n=49) set daily step goals and received feedback from the device but no other interventions. The intervention group (n=41) entered a 12-week game designed using insights from behavioral economics that assigned points and levels for achieving daily step goals and was reinforced by a support partner (family/peer) who received information on the participant’s progress and agreed to provide support and encouragement. During the follow-up period, daily steps were monitored for both groups. The primary outcome was mean change in daily step count from baseline through the 12-week intervention, and the secondary outcome was sustained change in step counts over the follow-up period. Exploratory outcomes included a-priori subgroup analyses of participants with higher social support and feasibility of collecting remote functional and cognitive performance measured by Timed Up-and-Go (TUG) and MOCA exams

RESULTS: We recruited 90 participants: mean age 70.4 (SD 2.9), 78% female, 71% married, 81% retired, 29% income >100k, 69% prior wearable use. Participants in the intervention arm had significantly greater change in mean daily step count (adjusted difference, 1792 (95% CI: 1238, 2346; p<.0001), which was sustained in the follow-up period (adjusted difference, 1216 (95% CI: 440, 1993); p=0.002). Mean daily step counts for Participants with higher social support were not significantly different than other participants. Follow-up TUG and MOCA scores were not significantly different than baseline scores for either arm.

CONCLUSIONS: A 12-week remote intervention using gamification and social incentives in a cohort of older adults at risk for ADRD was effective in promoting higher physical activity that was sustained during follow-up. Remote, repeated assessments of functional and cognitive measures were feasible but did not change significantly for control or intervention participants over 18 weeks. Future studies with longer intervention and follow-up should explore impact of sustained physical activity on prevention and progression of functional and cognitive changes associated with ADRD.

ASSOCIATION BETWEEN PHYSICIAN AGE AND PATTERNS OF END-OF-LIFE CARE DELIVERED TO PERSONS WITH DEMENTIA

Jessica J. Zhang¹; Hiroshi Gotanda²; David B. Reuben³; Anne Walling^4,5; Haiyong Xu⁴; Yusuke Tsugawa^4,6

¹Department of Medicine, University of California Los Angeles David Geffen School of Medicine, Los

Angeles, CA; ²Division of General Internal Medicine, Cedars-Sinai Medical Center, Los Angeles, CA; ³Multicampus Program in Geriatric Medicine and Gerontology, University of California Los Angeles David Geffen School of Medicine, Los Angeles, CA; ⁴Division of General Internal Medicine and Health Services Research, University of California Los Angeles David Geffen School of Medicine, Los Angeles, CA; ⁵Greater Los Angeles Veterans Affairs Healthcare System, University of California Los Angeles David Geffen School of Medicine, Los Angeles, CA; ⁶Department of Health Policy and Management, UCLA Fielding School of Public Health, Los Angeles, CA. (Control ID #3874674)

BACKGROUND: Physician age may be associated with patterns of end-of-life (EOL) care delivered to persons with dementia. Given the projected increase in persons with dementia and increasing concerns about the quality of EOL care received by persons with dementia, we determined whether patterns of EOL care differ by physician age.

METHODS: Using a nationally representative sample of Medicare fee-for-service beneficiaries aged 66 years and older with dementia who died in 2016-2019, we conducted a cross-sectional study examining the association between physician age and patterns of EOL care. We attributed beneficiaries to a physician who had the largest number of primary care visits during the last 6 months of life. Physician age was categorized as <40, 40-49, 50-59, or ≥60 years old. Our outcome measures included: (i) advance care planning (ACP) and palliative care (ACP, palliative care counseling, and hospice enrollment in the last 90 days of life), and (ii) high-intensity EOL care (emergency department visits, hospital admissions, intensive care unit [ICU] admissions, mechanical ventilation, cardiopulmonary resuscitation, or feeding tube placement in the last 30 days of life).

RESULTS: We included 287,514 beneficiaries with dementia who died in 2016-2019. Beneficiaries with dementia under the care of younger physicians had a higher proportion of ACP (P-for-trend <0.001 after accounting for multiple comparisons), palliative care counseling (P-for-trend <0.001), and hospice enrollment (P-for-trend<0.001) compared to those under the care of older physicians. Physician age was also associated with a lower proportion of emergency department visits (P-for-trend=0.005), hospital admissions (P-for- trend=0.002), mechanical ventilation (P-for-trend <0.001), or cardiopulmonary resuscitation (P-for- trend<0.001) in the last 30 days of life, although there was no evidence that ICU admissions or placement of feeding tubes in the last 30 days of life was associated with physician age.

CONCLUSIONS: We found that younger physician age was associated with more ACP and palliative care and less intensive EOL care for persons with dementia compared to older physician age.

ASSOCIATION BETWEEN PHYSICIANS’ GERIATRIC TRAINING AND PATTERNS OF END- OF-LIFE CARE DELIVERED TO PERSONS WITH DEMENTIA: A CROSS-SECTIONAL STUDY Jessica J. Zhang¹; Hiroshi Gotanda²; David B. Reuben³; Anne Walling^4,5; Haiyong Xu⁴; Yusuke Tsugawa^4,6

¹Department of Medicine, University of California Los Angeles David Geffen School of Medicine, Los Angeles, CA; ²Division of General Internal Medicine, Cedars-Sinai Medical Center, Los Angeles, CA; ³Multicampus Program in Geriatric Medicine and Gerontology, University of California Los Angeles David Geffen School of Medicine, Los Angeles, CA; ⁴Division of General Internal Medicine and Health Services Research, University of California Los Angeles David Geffen School of Medicine, Los Angeles, CA; ⁵Greater Los Angeles Veterans Affairs Healthcare System, University of California Los Angeles David Geffen School of Medicine, Los Angeles, CA; ⁶Department of Health Policy and Management, UCLA Fielding School of Public Health, Los Angeles, CA. (Control ID #3874672)

BACKGROUND: Geriatric training is designed to prepare physicians to meet the complex needs of older adults, including persons with dementia at the last stages of their lives. Given the projected increase in persons with dementia and increasing concerns about the quality of end-of-life (EOL) care received by persons with dementia, we determined whether patterns of EOL care differ between physicians with versus without formal geriatric training.

METHODS: Using nationally representative data of Medicare fee-for-service beneficiaries with dementia who died in 2016-2018 (n=100,778), we conducted a cross-sectional study examining the association between geriatric training and patterns of EOL care among persons with dementia using a broad range of measures. We attributed beneficiaries to a physician who had the largest number of primary care visits during the last 6 months of life and determined whether the physician was trained in geriatrics. Our outcome measures included: (i) advance care planning (ACP) and palliative care (e.g., ACP, palliative care counseling, and hospice enrollment in the last 90 days of life), and (ii) high-intensity EOL care (e.g., emergency department visits, hospital admissions, intensive care unit [ICU] admissions, mechanical ventilation, cardiopulmonary resuscitation, or feeding tube placement in the last 30 days of life).

RESULTS: Beneficiaries with dementia under the care of physicians with geriatric training had a higher proportion of ACP (adjusted proportion, 15.5% vs. 13.0%; P=0.003 after accounting for multiple comparisons), palliative care counseling (22.7% vs. 20.7%; P=0.02), and hospice enrollment (64.5% vs. 59.9%; P<0.001). Geriatric training was also associated with a lower proportion of emergency department visits (54.2% vs. 58.1%; P<0.001), hospital admissions (49.1% vs. 52.3%; P<0.001), ICU admissions (24.8% vs. 26.8%; P=0.003), and use of mechanical ventilation (10.9% vs. 12.9%; P<0.001) in the last 30 days of life. There was no evidence that use of cardiopulmonary resuscitation or placement of feeding tubes differed between the two groups.

CONCLUSIONS: We found that persons with dementia cared for by physicians with geriatric training received more ACP and palliative care and less intensive EOL care. Our findings suggest that interventions for physicians without formal geriatric training who care for persons with dementia may be considered, if evidence emerges that short-duration training is effective in improving the quality of EOL care.

A TAILORED OUTREACH PROGRAM TO ENGAGE PATIENTS AND PHYSICIANS IN DEPRESCRIBING CHRONIC BENZODIAZEPINES IN PRIMARY CARE

Sul Gi Chae^1,3; Julia Lindenberg¹; Emma Lee¹; Kaden Shen²; Timothy Anderson¹

¹General Medicine and Primary Care, Beth Israel Deaconess Medical Center, Boston, MA; ²Northeastern University Bouve College of Health Sciences, Boston, MA; ³Pharmacy, Beth Israel Deaconess Medical Center, Boston, MA. (Control ID #3873927)

BACKGROUND: Benzodiazepines are commonly prescribed as chronic treatments for anxiety and insomnia despite little evidence of clinical benefit and strong evidence of increased risks for adverse drug effects. Using a population health conceptual framework, we sought to develop and evaluate a benzodiazepine deprescribing quality improvement program in a large US academic primary care clinic.

METHODS: Pharmacy data was used to develop a clinic registry of older adults prescribed chronic benzodiazepines and identify concurrent sedating medications. Outreach processes were designed to match other population health practices in the clinic. Primary care physicians (PCPs) received one email providing benzodiazepine tapering resources, offering a voluntary deprescribing training, and asking them to opt out patients not appropriate for outreach. Patients were mailed one letter including discussing benzodiazepine risks, patient-specific information regarding heightened risk, and advising them to discuss deprescribing with the clinic’s Safe Prescribing Team or their PCP. A pharmacist was available to coordinate tapering for all patients. The primary outcomes were the number of patients who discussed deprescribing and number with a benzodiazepine dose reduction at 90 days.

RESULTS: A total of 559 patients and 50 PCPs were included in the study. This registry consists of 328 female (58%), 462 (82%) white, 388 patients who are between 65-74 years old (69%) and older (31%). One PCP requested a deprescribing training, 4 PCPs (8%) opted all patients out of the program, and 21 PCP (42%) opted out some patients. In total, 133 patients (24%) were opted out. Of the remaining 426 patients, 38 (9%) patients contacted the team and received pharmacist-led education. Subsequently 15 (40%) of patients started a taper. At 90 days, 4 patients discontinued benzodiazepines, 2 patients have had 50% dose reduction, 2 patients have had a 25% dose reduction, 5 patients have had less than 25% dose reduction, and 2 patients returned to prior dose after trying taper. Retrospective chart review found that at the end of 90 days after the outreach, 110 patients (27%) had conversations about sleep, anxiety or benzodiazepine use with a clinician.

80 of those patients (73%) had the conversation with their PCP.

CONCLUSIONS: A one-time low-cost clinic-based education mailing on benzodiazepine risks led to deprescribing conversations for a minority of patients but when deprescribing conversations occurred, nearly half of patients started a benzodiazepine taper. Direct patient outreach resulted in greater engagement than PCP outreach. To make deprescribing a part of routine, primary care practice will require innovative population-health efforts to strengthen patient-engagement and team-based care.

BARRIERS ORGANIZATIONAL LEADERS FACE PREPAIRING TO IMPLEMENT GOALS OF CARE INITIATIVES IN PRIMARY CARE

Alexis Hunyh³; Matthew McCaa¹; Raziel Gamboa¹; Kieran Holzhaur⁴; Arpan Patel^4,7; Karl A. Lorenz^1,2; David Bekelman^5,6; Anne Walling^4,7; Karleen Giannitrapani^1,2

¹Center For Innovation to Implementation, VA Palo Alto Health Care System, Palo Alto, CA; ²Division of Primary Care and Population Health, Stanford University School of Medicine, Stanford, CA; ³Center for Healthcare Innovation, Implementation, and Policy, VA Greater Los Angeles Healthcare System, Los Angeles, CA; ⁴VA Greater Los Angeles Healthcare System, Los Angeles, CA; ⁵VA Eastern Colorado Health Care System, Aurora, CO; ⁶University of Colorado Anschutz Medical Campus School of Medicine, Aurora, CO; ⁷University of California Los Angeles, Los Angeles, CA. (Control ID #3875900)

BACKGROUND: Across VA nationally, most goals of care conversations (GoCCs) occur in the inpatient setting when patients are close to death; only 39% occur in the primary care. In advance of proactively facilitating GoCC uptake in a pragmatic trial to enhance implementation of the VA Life Sustaining treatment

(LST) decisions initiative, we aimed to identify current barriers primary care practices face implementing

GoCC for seriously ill patients.

METHODS: We conducted semi-structured interviews with 20 interdisciplinary organizational leaders (physicians, nurses, and social work) from three geographically disparate healthcare systems. We used the Consolidated Framework for Implementation Research (CFIR) to inform interview guide development. Our analytic approach relied on thematic analysis, leveraging CFIR in conjunction with team and role theories to inform our code schema.

RESULTS: We present 4 themes representing organizational leaders’ perspectives of barriers.

1) Providers say “why bother” to having GoCC when there are high competing demands in a PC visit with too little time to prioritize anything but immediate concerns and “people change their minds.”

2) The information documented in a LSTs note is not used by other providers to inform patient care choices: “they [Other provider] don't have time to read through all the notes, and if they're stable [when] admitted inpatient [we] will re ask those questions.”

3) Providers lack expertise in serious illness communication and lack clarity on how to coordinate sharing the tasks of completing GoCC, invoking the idea that various pieces of GOCC are better fulfilled by someone

else with either more expertise or more time.

4) Providers lack consensus on what a good quality GOCCs looks like, and challenge the notion that documentation in the LST note alone is enough to know that a quality conversation occurred. “We know when I've a good [life sustaining treatment] note…But I don't know if that is sufficient and if there's some other way to measure success….Something that's actionable that the team can go off.. You know what is a good [or] sufficient LST discussion look like?”

CONCLUSIONS: We identified key barriers for implementing GOCCs. We recently implemented a randomized controlled trial to test whether provider and patient facing strategies improve the frequency of documented GOCCs. Future work should also assess the quality of conversations documented.

CARE CASCADES FOLLOWING LOW-VALUE CERVICAL CANCER SCREENING IN VETERANS DUALLY ENROLLED IN THE VETERANS HEALTH ADMINISTRATION AND MEDICARE

Aimee N. Pickering^3,1; Xinhua Zhao¹; Florentina E. Sileanu¹; Elijah Lovelace¹; Liam Rose¹; Aaron Schwartz¹; Jennifer A. Hale¹; Loren J. Schleiden¹; Walid F. Gellad^1,3; Michael J. Fine^1,3; Carolyn T. Thorpe²; Thomas R. Radomski^3,1

¹Center for Health Equity Research and Promotion, VA Pittsburgh Healthcare System, Pittsburgh, PA; ²School of Pharmacy, University of North Carolina System, Chapel Hill, NC; ³Division of General Internal Medicine, University of Pittsburgh School of Medicine, Pittsburgh, PA. (Control ID #3875823)

BACKGROUND: Although low-value cervical cancer screening occurs infrequently within VA, it may lead to unnecessary services, defined as a care cascade. Our objective was to compare the rates and costs of care cascade services in Veterans dually enrolled in VA and Medicare who received low-value cervical cancer screening with those who did not.

METHODS: We constructed a cohort of Veterans dually enrolled in VA and Medicare for whom cervical cancer screening would be considered low-value: females aged ≥65 without risk factors for cervical cancer (e.g., abnormal Papanicolaou (pap) smear). Our exposure group was Veterans who underwent cervical cancer screening in VA or Medicare during the first half of FY18; our control group was Veterans who had a primary care or gynecology visit but no cervical cancer screening during this time. Six months following the initial test or outpatient visit, we identified the following cascade services within VA and Medicare: 1) subsequent cervical cancer screening; 2) outpatient gynecology visit or non-gynecology visit for an abnormal pap smear; 3) cervical procedures (e.g., colposcopy; diagnostic excisional and ablative procedures); and 4) hysterectomy. We compared the rates and costs of cascade services in the exposure and control groups, adjusting for patient and facility-level covariates using negative binomial or linear regression models with inverse probability of treatment weighting. We also determined the location of downstream care.

RESULTS: Among 19,815 Veterans in the cohort, the mean age was 74.7 (8.4); 80.1% were non-Hispanic white. Overall, 468 (2.4%) underwent low-value cervical cancer screening (260 in VA; 208 in Medicare). The exposure group received 44.6 (95% CI 26.1-63.2) additional cascade services per 100 Veterans compared to controls. The difference in rates of cascade services was highest for related outpatient visits (40.1 per 100 Veterans) and repeat screening (2.8 per 100 Veterans) while other services occurred at low rates. The exposure group incurred an additional $34.2 (95% CI 14.4-54.2) per Veteran due to cascade services. Among those who had cervical cancer screening within VA, the rate of cascade services was 35.8 per 100 Veterans in VA and 2.7 per 100 Veterans in Medicare. For those who had screening within Medicare, the rate of cascade services was 5.8 per 100 Veterans in VA and 38.0 per 100 Veterans in Medicare.

CONCLUSIONS: Care cascade services were common after low-value cervical cancer screening in Veterans, and both the initial screening and subsequent cascade services occurred within and outside of VA. These findings demonstrate the importance of identifying low-value cascade services and measuring health service utilization occurring in healthcare systems outside of VA to fully capture the burden of low-value cervical cancer screening on Veterans.

CHALLENGES EXPERIENCED BY FAMILY MEMBERS OF HOSPITALIZED OLDER ADULTS WITH DEMENTIA WHEN MAKING “IN-THE-MOMENT” DECISIONS REGARDING INTENSITY OF CARE

Rashmi K. Sharma¹; Janaki M. Torrence¹; Elizabeth W. Dzeng²; J R. Curtis³; Ruth A. Engelberg⁴

¹Medicine, University of Washington, Seattle, WA; ²Medicine, University of California San Francisco, San Francisco, CA; ³Medicine, University of Washington, Seattle, WA; ⁴Medicine, University of Washington, Seattle, WA. (Control ID #3872293)

BACKGROUND: Family members of hospitalized older adults with dementia often face complex, high- stakes, “in-the-moment” decisions about ICU care or other high intensity interventions, and experience significant psychological distress as a result. Prior studies have focused on the ethical, intrapersonal, and communication-related challenges specific to surrogate decision making, but few have explored challenges that may be specific to families of patients with dementia. We sought to understand specific challenges faced by family members of hospitalized older adults with dementia as they navigated decisions about intensity of care.

METHODS: This was a qualitative study of family members of hospitalized older adults with dementia between February 2020 and December 2020. Hospitalized patients > 50 years old with an ICD diagnosis of dementia were identified from the EHR along with their family member. Family members were then approached for a semi-structured phone interview about their experiences with decision making regarding intensity of care. Interviews were audiotaped, transcribed, and coded by seven investigators using an inductive and constant comparative approach.

RESULTS: Thirty-one interviews were conducted of which 30 comprised the analytic sample. Mean participant age was 63 years and 66.7% were female. Seventy-three percent of participants identified as white, 17% as Black, 7% as Asian, and 3% as other race. The majority of participants were either an adult child (53.3%) or spouse (26.7%). Three key themes emerged regarding specific decision-making challenges

in the presence of dementia: 1) discordance between family and clinician perceptions of the patient’s baseline function and quality of life which were central factors in the way both parties constructed decision preferences (family) and recommendations (clinicians); 2) family perceptions that clinicians attributed acute changes to dementia and were reluctant to present certain important decision options (e.g., pursuing diagnostic interventions); and 3) family’s intrapersonal tensions between wanting more comfort-focused care within the context of dementia progression and maintaining hope that their loved one may still experience an acceptable quality of life with additional interventions.

CONCLUSIONS: Family members of hospitalized patients with dementia faced several dementia-specific challenges when making in-the-moment decisions about intensity of care. Interventions that account for these dementia-specific decision-making challenges are needed to better support family members of older adults with dementia as they navigate high-stakes, “in-the-moment” decisions.

COMMUNICATION PARTICIPATION IN PRIMARY CARE VISITS: THE PHYSICIAN, PATIENT, CAREGIVER TRIAD

Leah S. Karliner¹; Jennifer Livaudais-Toman²; Celia P. Kaplan³; Lisa C. Diamond⁴; Debra L. Roter⁵

¹Medicine, UCSF, San Francisco, CA; ²Division of General Internal Medicine, University of California, San Francisco, Greenwich, CT; ³Medicine, University of California San Francisco, San Francisco, CA; ⁴Immigrant Health and Cancer Disparities Service, Memorial Sloan-Kettering Cancer Center, New York, NY; ⁵Health, Behavior and Society, Johns Hopkins Bloomberg School of Public Health, Baltimore, MD. (Control ID #3873987)

BACKGROUND: Family caregivers often accompany older patients to medical visits. To optimize care, it is important to understand communication during these triadic visits. In this study we examine the presence of a caregiver on the amount of biomedical and psychosocial-emotional talk during primary care visits.

METHODS: We leveraged a dataset of audiotaped primary care visits with 189 older ethnically Chinese and Latinx patients. We classified each visit according to a family caregiver’s presence (‘accompanied’) or absence (‘unaccompanied’). We used the Roter Interaction Analysis System to analyze the audiotapes, identifying and comparing frequency of biomedical and psychosocial-emotional communication elements between accompanied and unaccompanied visits. For accompanied visits, we first excluded and then included caregiver talk.

RESULTS: Of the 189 visits, 60 (32%) were accompanied. Patients who were accompanied were on average older (77.5 vs. 67.2; p<.001), and more often had limited English proficiency (87% vs 62%; p<.001). Half (n=96) of all visits were fully, and 15 partially, language concordant in Chinese, English or Spanish; 66 visits had a professional interpreter present. For 11 (18%) accompanied visits, caregivers acted as interpreters. Doctors did not vary the frequency of biomedical or psychosocial-emotional talk between accompanied and unaccompanied visits. Patients in accompanied visits had less of every talk element compared with patients in unaccompanied visits, including medical information, medical questions, psychosocial and lifestyle information, psychosocial questions, and emotional talk (all p <.02). When adding caregiver talk, there was no longer a difference between accompanied and unaccompanied visits for any element except medical questions which was higher for accompanied visits (p=.02).

CONCLUSIONS: Having a caregiver present did not change measured physician communication behavior.

Caregivers were active participants for both biomedical and psychosocial-emotional communication. It remains unclear the best ways to elicit patients’ communication preferences and to balance communication with patients and their caregivers during visits.

DETECTING BEHAVIORAL CHANGE FACTORS AMONG HEALTHCARE PROVIDERS REGARDING GOALS OF CARE COMMUNICATION

Arpan Patel^1,2; Sylvia H. Paz³; Marian L. Katz⁴; Alicia Bergman⁵; Vanessa Ryan¹; Karen L. Spritzer³; David Ganz²; Neil Wenger⁶; Anne Walling^7,2; Ron D. Hays³

¹Digestive Diseases, University of California Los Angeles, Los Angeles, CA; ²VA Greater Los Angeles Healthcare System, Los Angeles, CA; ³General Internal Medicine and Health Services Research, University of California Los Angeles, Los Angeles, CA; ⁴Research, Veterans Health Administration, North Hills, CA; ⁵CSHIIP, VA Greater Los Angeles Healthcare System, Los Angeles, CA; ⁶Medicine, University of California, Los Angeles, Los Angeles, CA; ⁷Medicine, University of California Los Angeles, Los Angeles, CA. (Control ID #3875856)

BACKGROUND: Discussions between patients with serious illness, their families, and healthcare providers regarding goals of care (GOC) may lead to improved quality of care and outcomes, but such discussions are rarely performed. Efforts to improve GOC communication have used formal communication skills training, with the goal of increasing the comfort of healthcare providers to perform such conversations. However, training may have other effects. To help guide improvement efforts, we aimed to identify other correlates of behavior change.

METHODS: We developed an instrument to detect correlates of behavior change for GOC communication (BCGC) by adapting the Determinants of Implementation Behavior Questionnaire. We performed cognitive interviews with 23 healthcare providers across 2 institutions (West Los Angeles VA Medical Center, UCLA Medical Center) representing different disciplines (74% physicians, 17% advance practice practitioners, and 9% social workers) and specialties (52% medical and surgical specialties [non-palliative care], 39% internal medicine, 9% palliative care). Rapid qualitative analysis methods were used to iteratively adjust phrasing and content of items. We administered the modified instrument cross-sectionally (N=100) to providers across both institutions and then in a pre-post format (N=25) to providers participating in formal GOC communication skills training programs. We estimated item-scale correlations, confirmatory factor analysis, responsiveness to change, and short-form analysis.

RESULTS: The final BCGC instrument contained 34 questions and represented 9 behavior change domains (Knowledge, Skills, Beliefs about Capabilities; Organizational Context; Provider’s Expectations; Role and Identity; Beliefs about Consequences; Attention; Intention; Priority; and Perceptions about Patient Engagement). Item-scale correlations (α>0.46) and confirmatory factor analyses (Comparative Fit Index=0.94, Root Mean Square Error of Approximation=0.078) were acceptable. Five domains (Knowledge, Skills, Beliefs about Capabilities; Organizational Context; Provider’s Expectations; Priority; and Perceptions about Patient Engagement) changed significantly after a single GOC communication skills training (Effect Sizes (Cohen’s d) ranging from 0.500-1.292, p<0.05). Short-form versions of 2 domains (Knowledge, Skills, Beliefs about Capabilities [11 items to 3 items]; Organizational Context [8 items to 3 items]) also changed significantly after training (Effect Sizes 1.220 and 0.597, respectively, p<0.05).

CONCLUSIONS: The BCGC survey instrument had favorable psychometric properties but should be further evaluated in a larger sample of healthcare providers. We demonstrated that some behavior change domains (but not all) changed after a single formal GOC communication skills training. Using these domains as intermediate measures can be helpful in guiding future improvement efforts aimed at addressing provider and organizational barriers to GOC communication.

DEVELOPMENT AND PILOT TESTING OF AN ARTIFICIAL INTELLIGENCE-BASED FAMILY CAREGIVER NEGOTIATION TRAINING PROGRAM: NEGOTIAGE

Alaine Murawski¹; Vanessa Ramirez-Zohfeld¹; Allison Schierer¹; Johnathan Mell²; Jeanne Brett³; Lee Lindquist¹

¹Medicine - Geriatrics, Northwestern University Feinberg School of Medicine, Chicago, IL; ²University of Central Florida, Orlando, FL; ³Northwestern University, Evanston, IL. (Control ID #3874370)

BACKGROUND: Family caregivers of older people with Alzheimer’s disease (PWD) often experience conflicts as they navigate the healthcare system for their loved ones (e.g., determining treatment necessity, billing errors, home health extensions). With these conflicts, family caregivers experience unnecessary frustration, anxiety, and stress. Many family caregivers lack the negotiation training and dispute resolution skills to resolve these conflicts. We sought to develop and pilot test an artificial-intelligence negotiation training program, NegotiAge, for family caregivers.

METHODS: With NIH funding [R01AG068421], we convened a group of national negotiation experts, geriatrician, social worker, and community-based family caregivers (N=9; Illinois, Florida, New York, California). Content matter experts created short informational videos/didactics to teach negotiation skills. Family caregivers generated dialogue surrounding caregiver conflicts. Computer science experts used the dialogue with the Interactive Arbitration Guide Online (IAGO) platform to develop avatar-based agents (e.g. sibling, older adult, physician) for caregivers to practice negotiating with on their own time. We then pilot tested the NegotiAge training program with family caregivers to assess feasibility and usability.

RESULTS: NegotiAge.com was developed to facilitate negotiation skills training for family caregivers of PWD. Family caregivers progress through didactic material, then receive scenarios to negotiate (e.g. physician recommends gastric tube in PWD, sibling disagrees with home support, PWD refusing support). Caregivers negotiate in real-time with avatars who are designed to act like humans, including emotional tactics and irrational behaviors. Caregivers send/receive offers, using tactics until either mutual agreement or time expires. Immediate feedback is generated from the response chosen/tactics used to assist with the negotiation skills training. Pilot testing was conducted with family caregivers of PWD (n=15). and showed that the negotiation and conflict resolution training program was feasible and usable for family caregivers. Subjects found the program to be highly satisfying with real-world applicability.

CONCLUSIONS: NegotiAge is an Artificial Intelligence-based online Caregiver Negotiation Program, that is usable and feasible for family caregivers to become familiar with navigating conflicts commonly seen in caring for older adults with dementia. The overarching goal is to empower family caregivers to better communicate when advocating for their loved ones and make conflict resolution easier. Further Multiphase Optimization Strategy (MOST) randomized testing will be conducted next to optimize this program. Ultimately, we hope to improve the quality of care of older adults while decreasing the stress, burden, and frustration experienced by their family caregivers.

DEVELOPMENT OF A GERIATRIC MODULE FOR A SOCIAL DETERMINANTS OF HEALTH SCREENING TOOL

Naila Edwards^1,6; Meaghan Kennedy^2,3; Alicia Cohen^4,5; Andrea W. Schwartz^7,8; MICHELLE MARTINCHEK¹

¹VA New England Geriatric Research Education and Clinical Center, Boston, MA; ²VA Bedford Healthcare System, Bedford, MA; ³Department of Family Medicine, Boston University School of Medicine, Boston, MA; ⁴VA Providence Healthcare System Center of Innovation in Long Term Services and Supports, Providence, RI; ⁵Department of Family Medicine, Brown University Warren Alpert Medical School, Providence, RI; ⁶Department of Geriatrics, Harvard Medical School, Boston, MA; ⁷Geriatrics, VA New England Geriatric Research Education and Clinical Center, Boston, MA; ⁸Medicine, Harvard Medical School, Boston, MA. (Control ID #3876415)

BACKGROUND: Social determinants of health are increasingly being addressed in healthcare settings through routine screening for health-related social needs (e.g., food insecurity). Older adults face a unique set of challenges that may not be included on existing screening tools designed for general patient populations. The Assessing Circ*mstances and Offering Resources for Needs (ACORN) tool is a Veterans Affairs (VA) social needs screening tool that has been successfully implemented in VA clinical settings. We sought to develop a geriatric module to be used in conjunction with the core ACORN screener to assess for unmet social needs in older veterans in a VA outpatient geriatrics clinic.

METHODS: We conducted a literature review of social needs domains affecting health outcomes in older adults. From this literature review, we identified 17 domains that were applicable to this population. We then compared domains to the current geriatric clinic template and identified 14 domains that were prominent in the literature, but not being addressed. Then we compared these 14 domains to the original ACORN screener and found 3 relevant domains that were not being screened for by either ACORN or geriatric clinic template. We interviewed key informants and subject matter experts including clinicians, researchers and teams experienced in implementing social needs screening in geriatrics settings. Key informants also reviewed the current ACORN screener along with the current geriatric clinic template. Once a consensus was reached regarding the domains to be used in the ACORN-Geri screener, established screening instruments for each domain were reviewed.

RESULTS: Feedback from key informants was discussed amongst an interprofessional team. Domains selected for inclusion in the ACORN-Geri module were: 1) consistently identified as important, and 2) not routinely assessed in the geriatric clinic. The domains chosen were elder abuse/mistreatment, medication affordability, and caregiver burden. The module was then created by adapting established screening tools and developing new questions in these domains. We incorporated screening processes used in other VA clinic settings during workflow development to optimize implementation and leverage prior lessons learned.

CONCLUSIONS: Standard social determinants of health screening tools may not assess certain domains that are unique to older adults. There is an opportunity to identify vulnerable older adults, in particular older veterans, and link them with resources available within or outside the VA. Through an iterative, interprofessional process, we identified 3 domains for inclusion in a geriatric add-on module for an existing VA social needs screener. The new module will be piloted as part of a quality improvement project in a geriatric consultative clinic. The screener is intended for use in VA community-based settings with substantial older adult populations.

DOES A COLLABORATIVE PALLIATIVE CARE INTERVENTION INFLUENCE ADVANCED CARE PLANNING IN PATIENTS WITH HEART AND LUNG DISEASES?: SECONDARY OUTCOMES FROM A RANDOMIZED CLINICAL TRIAL

Allison V. Lange¹; William Feser³; Anna E. Baron³; David Bekelman^2,4

¹Pulmonary Sciences and Critical Care Medicine, University of Colorado Anschutz Medical Campus School of Medicine, Aurora, CO; ²Medicine, University of Colorado, Denver, CO; ³Biostatistics and Informatics, Colorado School of Public Health, Aurora, CO; ⁴Rocky Mountain Regional Veterans Affairs Medical Center, Denver-Seattle Center of Innovation, Aurora, CO. (Control ID #3874456)

BACKGROUND: Chronic heart failure (CHF), chronic obstructive pulmonary disease (COPD), and interstitial lung disease (ILD) are leading causes of hospitalization and mortality. However, few patients with these illnesses engage in goals of care conversations (GoCC). We determined the effect of a collaborative palliative care intervention for patients with CHF, COPD, and ILD on readiness to engage in advance care planning (ACP), advanced directive (AD) documentation and GoCC.

METHODS: The Advancing Symptom Alleviation with Palliative Treatment (ADAPT) study was a randomized clinical trial of a collaborative palliative care intervention. The study enrolled patients with COPD, ILD, or CHF from 2 VA health systems who were at high risk for hospitalization and death and who reported poor quality of life. As part of the intervention, a nurse and social worker had a GoCC visit with each patient. GoCC were documented in the electronic medical record. Patients were encouraged to complete an AD. At baseline and 6 months, GoCC, AD, and readiness to engage in advance care planning (ACP Engagement Survey-4, example question, “How ready are you to decide who you want your medical decision maker to be?” 1=I have never thought about it, 2=I have thought about it, but I am not ready to do it; 3=I am thinking about doing it in the next 6 months; 4=I am definitely planning to do it in the next 30 days; 5=I have already done it) were measured.

RESULTS: Of 306 randomized patients (154 intervention, 152 usual care), 57.8% had COPD, 21.9% HF, 16% both COPD/HF, 4.2% ILD. Mean age was 68.9 years, and patients were primarily white (80.1%) and male (90.2%). At baseline, patients in the intervention and control groups had similar rates of readiness to engage in ACP, AD completion, and GoCC. At 6 months, mean readiness to engage in ACP improved by 0.52 in the intervention group and 0.03 in the control group (difference, 0.49; effect size, 0.43; p=0.0005). Documented GoCC increased in the intervention group (intervention at baseline: 56/154 (36%), 6 months: 131/145 (90%), control baseline: 50/152 (39%), 6 months: 59/145 (34%), effect size 1.40; p<0.0001), while percent of patients who had a documented AD did not change (intervention at baseline: 56/154 (36%), 6 months: 58/145 (40%), control at baseline: 61/152 (40%), at 6 months: 62/145 (43%), p=0.64)

CONCLUSIONS: Nurses and social workers conducting GoCC as a part of a palliative care intervention improved readiness to engage in ACP and documented goals of care conversations. Future work should investigate GoCC quality and how GoCC contribute to person-centered care.

EDUCATING AGE-FRIENDLY RESEARCHERS

Bryanna De Lima¹; Allison Lindauer²; Elizabeth Eckstrom^1,3

¹Medicine, Oregon Health & Science University School of Medicine, Portland, OR; ²Layton Aging and Alzheimer's Disease Center, Oregon Health & Science University, Portland, OR; ³Oregon Health & Science University Oregon Clinical & Translational Research Institute, Portland, OR. (Control ID #3868155)

BACKGROUND: Older adults have a high disease burden but are often underrepresented in research studies due to recruitment and retention obstacles. Researchers with geriatric expertise have identified potential solutions to these barriers but these practices have yet to be disseminated to the general research community. Our team created an interactive webinar series for researchers to share practical tips and tools to increase inclusion of older adults in research.

METHODS: Forty researchers (e.g., research assistants, data managers, project coordinators, co- investigators) participated in a 6-session webinar series from October to November 2022. Sessions comprised of 20-30 minute didactics and 30-40 minute group discussions. Participants completed pre- and post-program surveys. Responses were analyzed to determine changes in confidence levels and to understand how participants rated the webinar series.

RESULTS: Thirty-nine participants completed both pre- and post-program surveys. Self-reported confidence levels improved across all categories (see Table). On a scale of 1-10, the average rating was 8.2 for the helpfulness of the series, 8.6 for the likeliness of utilizing what they learned, and 8.8 for the likeliness to recommend the webinar to others. Participants’ comments included: “The main takeaway I’ve gained from this series is how I can better communicate and facilitate older adults in research”; “small actions on the part of the research team can have a big impact on older adults’ comfort and willingness to participate in research”; “the real-world examples of what has occurred for others and also what they learned from the experience, and even the real-time ideas other teams had and shared were really helpful and inspiring.” The major critique received was to include more information on industry-sponsored clinical trials.

CONCLUSIONS: Participants positively responded to our Age-Friendly Research webinar series with reported improvements in confidence and knowledge. Adaptations to the series should incorporate suggestions for industry-sponsored trials.

ENCOURAGING SERIOUS ILLNESS CONVERSATIONS IN THE GENERAL MEDICINE INPATIENT SETTING

Myrna K. Serna¹; Cathy Yoon²; Julie Fiskio²; Joshua Lakin³; Anuj Dalal²; Jeffrey L. Schnipper¹

¹Internal Medicine, The University of Texas Medical Branch at Galveston Department of Internal Medicine, Galveston, TX; ²Hospital Medicine Unit, Brigham and Women's Hospital, Boston, MA; ³Psyschosocial Oncology and Palliative Care, Dana-Farber Cancer Institute, Boston, MA. (Control ID #3872443)

BACKGROUND: Despite benefits of early Serious Illness Conversations (SICs), including increased satisfaction and earlier hospice referral, rates of SICs remain low. The primary aim of this study is to assess if standardized documentation of SICs increase following implementation of interventions encouraging providers to have more of these conversations in the inpatient setting.

METHODS: In this interrupted time series analysis, encounters ≥18 years old with an Epic (Verona, WI) Readmission Risk Score (RRS) of >28% admitted to a general medicine service at an academic medical center from January 2019 to December 2020 (pre-implementation), January 2021 (wash-in) or February 2021 to October 2021 (post-implementation) were included. The Expert Recommendations for Implementing Change (ERIC) framework guided development of interventions to encourage SICs and standardized documentation within a structured electronic health record module, which complemented pre-existing faculty training in SICs. A multivariable segmented logistic regression model was used to evaluate the immediate effect of the interventions (y-intercept change, “change in step”) and the difference in temporal trends (change in slope) of the odds of standardized documentation of an SIC per month. The model was adjusted for patient hospitalizations in the year prior to index admission, Elixhauser/Van Walraven comorbidity score, and COVID-19 inpatient census on day of patient admission.

RESULTS: Major interventions included stakeholder engagement to identify barriers and facilitators of implementation (using the Consolidated Framework for Implementation Research (CFIR)), weekly emails sent to providers with admitted patients who may benefit from an SIC, encouraging use of a Quality and Safety Dashboard to facilitate identification of patients with elevated RRS and SIC needs, identification of clinical champions, and educational sessions of inpatient teams, among others. An interrupted time series model did not show a change in y-intercept (odds ratio (OR) 0.58, 95% confidence interval (CI) 0.18-1.89) but did show a statistically significant increase in the temporal trend (OR 1.20, 95% CI 1.00-1.43).

CONCLUSIONS: Implementation of context-specific interventions guided by the ERIC framework resulted in increased improvement over time in standardized documentation of SICs for encounters with elevated comorbidity burden. Continued assessment of barriers and facilitators to conducting these conversations and refinement or use of additional interventions is needed to support clinicians in conducting these conversations and providing patient-centered advance care planning.

ENHANCING COMMUNITY-DWELLING OLDER ADULTS’ SELF-CONFIDENCE TO MANAGE THEIR HEALTH AND WELL-BEING UTILIZING THE 4MS MODEL

Atreya Mishra¹; Claire T. Larkin¹; Vallari Shah²; Valerie Gruss³; Michael Koronkowski²; Memoona Hasnain¹

¹Medicine, University of Illinois Chicago College of Medicine, Chicago, IL; ²Pharmacy, University of Illinois Chicago, Chicago, IL; ³Nursing, University of Illinois Chicago, Chicago, IL. (Control ID #3875197)

BACKGROUND: The older adult population faces significant health challenges in managing chronic diseases, behavioral health, mobility, medication usage, and health literacy. To address this problem, we designed an educational workshop using the Institue of Health Improvements’s Age-Friendly 4Ms framework for high-quality care in a community-based participatory research (CBPR) partnership with the Pat Crowley House in Chicago, IL. Our objective was to explore whether education on the Age-Friendly 4Ms improved older adults’ self-confidence in managing their health and well-being.

METHODS:

A one day educational session was conducted educating the residents on each of the 4Ms: “What Matters”, “Medication”, “Mentation”, and “Mobility”. A total of 6 residents participated in the session. The effectiveness of the session was assessed using an identical pre-and-post survey identifying participants’ confidence in managing various aspects of their health related to the 4Ms. Ordinal data was collected from the surveys.

RESULTS: Participants were engaged throughout the workshop and had 100% participation in the pre- and post-surveys. Pre-post test analysis showed a positive trend towards increased self-confidence in talking with healthcare providers, managing medications, staying active, and understanding age-related cognitive changes.

CONCLUSIONS: Overall, the 4Ms model provided a valuable tool for addressing the needs of the residents at the Pat Crowley House. This community-based research project illustrates how the 4Ms model can be adapted into an interactive educational workshop for older adults. Education on 4Ms allows for them to have better conversations with healthcare providers and to increase their knowledge regarding challenges they commonly face. Our research was limited by a small sample size. There is a need for future research to study the impact of this intervention on a larger sample size and assess its long-term health outcomes. We believe that such an approach can be invaluable in improving the confidence of this vulnerable population in managing their healthcare.

EQUIPPING RESIDENTS FOR END-OF-LIFE CONVERSATIONS IN PRIMARY CARE

Brooke Wagen, Simisola Kuye, Sarah S. Mills, Rajvi Patel, Jaya Vasudevan, Heather L. Herrington, Asma Nuri, Michael Pignone, Robin Reister

Internal Medicine, The University of Texas at Austin Dell Medical School, Austin, TX. (Control ID #3873177)

BACKGROUND: Medical education programs largely lack formal teaching around end-of-life (EOL) care, despite the acknowledged need for physicians to be trained with these skills. Although there is general agreement about the importance of EOL care discussions, residency leaders highlight time constraints for teaching as a major limiting factor. We sought to develop and test an efficient outpatient educational intervention for residents in EOL care planning as part of a larger QI program.

METHODS: Our intervention took place within the Central Texas Veterans Affairs internal medicine resident clinic during the 2021-22 academic year. Curriculum was designed by resident team leaders in collaboration with palliative care faculty. Teaching was delivered over two one-hour lunch sessions – one didactic and one practice-based. Educational aims for the first session were the importance of EOL

planning in primary care (instead of in emergency settings), criteria to select appropriate patients for EOL planning visits, and differentiation of documentation for advanced care planning. We also taught the practical steps of conducting an EOL planning visit with attention to patient goals, preferences, and autonomy. The second session consisted of interactive practice for EOL planning conversations including real-time feedback from resident team leaders. Pre and post-teaching surveys were utilized to determine efficacy; all residents completed a pre-teaching survey before didactics, and took post- teaching surveys between two and four months after completion of both sessions.

RESULTS: Prior to our intervention, 41% of residents had completed an EOL planning visit and 27% had received formal palliative care education. Ninety-five percent of residents expressed desire for more teaching about how to incorporate goals of care and end of life discussions into primary care visits. By post-intervention, 68% had completed a visit and 100% of residents had formal didactics in EOL planning. Resident ability to differentiate medical power of attorney (MPOA) and medical surrogate increased from 36% to 86% and ability to distinguish advanced directive (AD) and Out of Hospital Do Not Resuscitate (DNR) increased from 59% to 95%. Residents were also asked to evaluate comfort level with risk stratifying patients and with communicating about EOL issues on a five point Likert scale. Comfort with identifying appropriate patients increased from 2.82 to 4.00 and comfort with discussion increased from 2.86 to 4.09. Overall, 100% of residents found the curriculum useful in teaching the skills needed for EOL planning in primary care.

CONCLUSIONS: A focused intervention for teaching residents EOL planning in the context of primary care increased knowledge, skills, and resident comfort. Given the efficacy of a two-hour educational intervention, this teaching could be widely adopted with minimal curricular adaptation.

EXAMINING EMERGENCY DEPARTMENT LENGTH OF STAY AMONG PERSONS LIVING WITH DEMENTIA

Stephanie Nothelle¹; Eric Slade³; Cynthia Boyd¹; Halima Amjad¹; Phillip Magidson¹; Tanya Chotrani⁴; Sarah Szanton³; Jennifer Wolff²

¹Medicine, Johns Hopkins University, Baltimore, MD; ²Johns Hopkins University Bloomberg School of Public Health, Baltimore, MD; ³Johns Hopkins University School of Nursing, Baltimore, MD; ⁴Boston Strategic Partners Inc, Boston, MA. (Control ID #3875083)

BACKGROUND: General internists are caring for a growing number of persons living with dementia (PLWD). PLWD have frequent emergency department (ED) visits. Given the challenges of caring for PLWD in the ED, PLWD may be at risk for prolonged ED length of stay (LOS), which is associated with risk of medication errors, delirium and mortality. We examined the differences in ED LOS by dementia status and factors associated with prolonged ED LOS for PLWD.

METHODS: We used data from the 2014-2018 Healthcare Cost and Utilization Project State ED Database, which includes ED visits that did not result in hospital admission statewide. We focus on data from 4 states: Arkansas (AR), Arizona (AZ), Florida (FL) and Massachusetts (MA) which report LOS. We limited analysis to visits by persons >64 years and matched visits by PLWD to visits by persons without dementia on age, race, sex, reason for visit and state. We define prolonged LOS as >90th percentile. We used mixed multilevel logistic regression to examine differences in risk of prolonged ED LOS for PLWD.

RESULTS: We included ED visits by 541,841 PLWD who were matched to 544,444 persons without dementia. Mean age was 76 years, 9% of visits were by persons of black race, 13% by persons of Hispanic ethnicity. PLWD had a mean LOS of 12.2 hours compared to 7.8 hours for persons without dementia. Twenty percent of PLWD had a prolonged ED LOS (i.e. LOS >22 hours), compared to 10% of persons without dementia. The risk of prolonged LOS varied by state (aORs: 6.9 FL, 2.9 AZ, 1.1 MA versus AR reference group, p<0.001), and was greater in EDs in metropolitan (aOR 1.8) and micropolitan (aOR 1.5) than rural areas (p<0.001), for persons from minority backgrounds (aORs: Hispanic 1.8, African American 1.5, Other 1.4 versus white non- Hispanic, p<0.001), persons with Medicaid and Medicare sources of payment (aORs: Medicaid 1.2, private 0.66 versus Medicare, p<0.001), persons with more chronic conditions (aOR: 1.3, p<0.001) and persons discharged to destinations other than home (e.g. nursing facility) (aOR 1.1, p<0.001).

CONCLUSIONS: In this study, PLWD had longer ED LOS and greater risk of prolonged ED LOS than matched older adults with similar reasons for ED visit. Personal and regional factors were associated with risk of prolonged ED LOS, highlighting a need for a deeper understanding of the causes of disparities in the prolonged ED LOS among PLWD.

HEALTH CARE CONTACT DAYS AS A PATIENT-CENTERED CARE MEASURE FOR MEDICARE BENEFICIARIES

Ishani Ganguli¹; E. John Orav¹; Emma D. Chant¹; Joyce Lii²; Ateev Mehrotra²; Christine S. Ritchie³

¹Division of General Medicine and Primary Care, Brigham and Women's Hospital, Chestnut Hill, MA; ²Harvard Medical School Department of Health Care Policy, Boston, MA; ³Medicine, Massachusetts General Hospital, Boston, MA. (Control ID #3876498)

BACKGROUND: Days spent obtaining health care outside the home represent access to care but also substantial time, effort, and cost, especially for older adults and their caregivers. These “health care contact days” are a patient-centered measure that should inform care decisions yet has not been characterized. We assessed variation, content, and drivers of contact days among older adults in Traditional Medicare (TM).

METHODS: We studied 2019 claims data linked to nationally representative Medicare Current Beneficiary Survey data for community-dwelling adults who were ≥65-years-old, without end-stage renal disease, and continuously enrolled in TM for the year or until death. Outcomes were ambulatory contact days (days with a primary care or specialty care visit, test, imaging, procedure, or treatment) and total contact days (ambulatory days plus days in an emergency department, hospital, skilled nursing facility, or inpatient hospice facility).

We described variation in median contact days among hospital referral regions (HRRs) with ≥10 respondents. To identify patient factors (survey and claims-derived sociodemographic and clinical factors and care-seeking behaviors) associated with contact days, we used weighted multivariable Poisson regression with offset for days alive and adjustment for overdispersion and HRR random effects.

RESULTS: In 2019, 6385 older adults (weighted: 28,631,117) had a median[IQR] of 10.8[3.5, 22.3] (range 0-179) ambulatory contact days and 11.7[4.0-24.7] (range 0-256) total contact days. Across HRRs, there was also wide variation in ambulatory contact days (13.0[11.0-16.0] (range 4-23.5)) and total contact days (15.0[12.0-18.0] (range 6.0-32.0)).

Older adults spent most contact days on ambulatory care including primary care visits (1.4[0-3.8]), specialty care visits (2.3[0-5.9]), tests (3.1[0.55-6.8]), imaging (1.2[0-3.4]), procedures (0.43[0-2.6]), and treatments (0.89[0-4.8]). Adults spent 4.2[0.70-10.9] (39%) of their ambulatory contact days coming in for a test, imaging, procedure, or treatment alone; 46.5% of test days and 45.4% of imaging days were not on the same day as a visit.

In multivariable analysis, factors associated with more ambulatory contact days included younger age, female sex, white race, non-Hispanic ethnicity, urban residence, and number of chronic conditions. Those reporting they “worry about their health more than other people their age” or “go to the doctor…as soon as they start to feel bad” had more contact days; those who “will do just about anything to avoid going to the doctor” had fewer. There was no association with self-rated health.

CONCLUSIONS: Contact days were mostly ambulatory and showed wide variation associated not only with multimorbidity but also with age, sex, race, ethnicity, care-seeking behaviors, and geography. These results point to factors beyond clinical need that may drive over- and underuse of contact days and to the importance of right-sizing this patient-centered measure, eg, via telemedicine and care coordination.

IDENTIFYING HOSPITALIZED PATIENTS WHO MAY BENEFIT FROM SERIOUS ILLNESS CONVERSATIONS

Myrna K. Serna¹; Katrina Grace Sadang²; Hanna Vollbrecht³; Cathy Yoon⁴; Julie Fiskio⁴; Joshua Lakin⁵; Anuj Dalal⁴; Jeffrey L. Schnipper⁴

¹Internal Medicine, The University of Texas Medical Branch at Galveston Department of Internal Medicine, Galveston, TX; ²University of California San Francisco, San Francisco, CA; ³Internal Medicine, Brigham and Women's Hospital, Boston, MA; ⁴Hospital Medicine Unit, Brigham and Women's Hospital, Boston, MA; ⁵Psyschosocial Oncology and Palliative Care, Dana-Farber Cancer Institute, Boston, MA. (Control ID #3876023)

BACKGROUND: Identifying patients who may benefit from a Serious Illness Conversation (SIC) in the hospital is an important step in increasing SIC timeliness. Epic’s Readmission Risk Score (RRS) is an electronic health record integrated composite score (0-100%) that predicts unplanned, 30-day readmission. The aim of this study is to assess if RRS combined with clinician response to the “surprise question” aids in identification of hospitalized patients who may benefit from SICs.

METHODS: In this retrospective study, patient encounters >18 y.o. with an admission to a general medicine service at an academic medical center from January 2019 to October 2021 and an RRS >28% were included and randomly selected electronically. Three clinicians (a hospitalist, a 3^rd year internal medicine resident and a 4^th year medical student) independently performed chart reviews to answer the following questions: 1) Would you be surprised if the patient died in the next 12 months? 2) Can the patient meaningfully participate in an SIC? 3) If the patient cannot participate in an SIC, why not? Each chart was reviewed by 2 clinicians with disagreements resolved by consensus-based discussion. Patient outcomes including standardized documentation of an SIC and post-discharge, 3-month mortality are reported, and Fisher Exact Test used to assess statistically significant differences.

RESULTS: A total of 202 patient encounter charts were reviewed. Clinicians felt they would not be surprised if 156 patients (77%) died in the next 12 months (i.e., answered “no” to the surprise question). Of these patients, 119 (59%) would be able to meaningfully participate in an SIC. Primary reasons for inability to participate included encephalopathy and dementia. Patients where clinicians answered “no” to the surprise question were more likely to have a documented SIC (9% vs 0%, p = 0.04), palliative care consultation (23% vs 2%, p < 0.01), change in code status from full to do not resuscitate/ do not intubate (17% vs 0%, p < 0.01), and post-discharge, 3-month mortality (8% vs 0%, p = 0.04) compared to patients where clinicians answered “yes” to the surprise question.

CONCLUSIONS: In a medical record review of patients with an RRS > 28%, clinicians felt they would not be surprised if 77% of patients died in the next 12 months, and these patients were more likely to receive advance care planning and to have higher post-discharge 3-month mortality, although rates of advance care planning were low in both groups. Most of these patients were determined to be able to meaningfully participate in an SIC. RRS, coupled with clinician acumen, may be a practical though imperfect means to identify patients who may benefit from an SIC during hospitalization. Encephalopathy and dementia were primary reasons for inability to participate in an SIC, suggesting need to perform SICs earlier in the life- course.

IMPLEMENTATION OF A GERIATRIC ASSESSMENT SMARTPHRASE: A MULTI- INSTITUTION PILOT STUDY

Eva Szymanski¹; Jessica Zuo²; Emily Fessler³; Ryan Chippendale⁴; Jennifer Ouellet²; Leah Schecter²; Richard Marottoli²; Rachel Miller¹

¹Division of Geriatrics, University of Pennsylvania, Philadelphia, PA; ²Section of Geriatrics, Yale University, New Haven, CT; ³Division of Geriatrics & Palliative Care, Weill Cornell Medicine, New York, NY; ⁴Section of Geriatrics, Boston University, Boston, MA. (Control ID #3871872)

BACKGROUND: The geriatric 5Ms are an increasingly utilized framework to communicate the complexity of the geriatric assessment (GA) to medical trainees and non-geriatricians. As the population ages, it will be important for internists to understand and perform a GA to help meet the promise of tomorrow in caring for complex older patients. The objective of this study is to pilot an electronic medical record (EMR)-based SmartPhrase using the 5Ms (Mind, Mobility, Medications, Multicomplexity, and Matters Most) to teach the GA to medical students and internal medicine/family medicine residents.

METHODS: We developed a concise Epic SmartPhrase organizing the GA within the 5Ms framework using iterative feedback from learners and educators. We incorporated the SmartPhrase into geriatrics teaching

across 3 academic institutions, including two inpatient Acute Care for Elders rotations (medical students and residents) and one geriatrics outpatient rotation (internal medicine residents only). Pre- and post-rotation surveys were used to evaluate utilization of the SmartPhrase, effects on comfort performing a GA, and influence on patient care.

RESULTS: Survey data collection is ongoing and planned to end in April 2023. Preliminary data demonstrated that 39% of learners (n=12/31) felt comfortable performing a GA prior to their geriatrics rotation and only 30% had a framework for the GA. Of the 18 learners who completed post-rotation surveys, 100% reported feeling comfortable performing a GA and 94% (n=17/18) confirmed having a framework for the GA. Out of 17 respondents (94%) who reported using the 5Ms SmartPhrase, 82% agreed that it was helpful when caring for older adults. Use of the SmartPhrase also affected clinical practice by improving interdisciplinary teamwork and aiding in identification and management of delirium, cognitive impairment, and polypharmacy.

CONCLUSIONS: An EMR-based SmartPhrase is an effective educational tool to disseminate geriatric principles and increase comfort with the GA among medical trainees. It can be used across multiple sites of care to enhance geriatrics education and patient care.