Faculty Bibliography

BACKGROUND AND OBJECTIVES/OBJECTIVE:Patients with epilepsy (PWE), especially temporal lobe epilepsy (TLE), experience impaired memory for personally experienced events. However, current assessments of episodic memory are limited in their ecological validity with a potential to miss detection of subtle cognitive decline. We conducted an exploratory study to determine whether a naturalistic film-viewing task with open-ended spoken recall could detect memory differences between TLE patients and healthy controls (HCs). METHODS:TLE patients (ages 18-60, fluent in English, not legally blind) were recruited from a Level 4 Epilepsy Center (2018-2024). TLE diagnosis was based on seizure semiology, MRI Brain, and EEG. TLE patients scored 22/30 on the Montreal Cognitive Assessment (MOCA); HCs scored 26/30. Subjects watched 6 short films and then freely recalled film details. Spoken responses were recorded, transcribed, segmented, and scored for film- and event-level recall. Recall order was assessed using the Damerau-Levenshtein distance. Semantic and causal centrality were quantified using sentence embeddings and rater-identified causal links, respectively. Beta regression with cluster-robust standard errors assessed group and centrality effects on recall probability. Beta regression evaluated the influence of age, MOCA, and testing platform on sequence recall error. RESULTS:We recruited 51 subjects (27 TLEs; 24 HCs, 70.1% F, mean 29.9 ±8.3 years). TLE patients and HCs showed similar recall of films (HC 89% ±11% vs TLE 88% ±18%, p = 0.54), coarse-grained events (HC 50% ±16% vs TLE 44% ±18%, p = 0.19) and fine-grained events (HC 25%±10% vs. TLE 22%±12%, p=0.17). Both groups recalled high causal centrality events better. However, TLE patients showed significantly greater fine-grained event sequence deviations at recall than HCs (HC 15% ±13% vs TLE 23% ±18%, p = 0.02, Hedges' g = 0.85, Cliff's δ = 0.51), with RTLE demonstrating more sequence deviations than HCs (15%±13 vs. 29%±21% p = 0.021) Age, education, MOCA, and performance on standard verbal and visual memory tasks were unrelated to film, event, and sequence recall performance. DISCUSSION/CONCLUSIONS:We demonstrate that a short film task with spontaneous spoken recall can identify group level differences in episodic memory. TLE patients demonstrate impaired sequence memory despite intact film- and event-level recall compared to healthy controls. Sequence memory may represent a subtle manifestation of memory impairment that is not detected by standard cognitive testing.

OBJECTIVES/OBJECTIVE:Generative artificial intelligence (AI) holds promise for creating patient education materials (PEM). However, AI-produced content is often difficult to read. Prompts for AI to write at a lower reading level either fail to meet recommended targets (≤6th grade) or demonstrate inconsistent results. We aimed to determine whether a new AI prompt for plain language would create PEMs that are easier to read, understand, and act upon. METHODS:We analyzed PEMs from two websites (HealthyChildren.org, UpToDate) and two generative AI programs (ChatGPT, Gemini). For each AI program, we used a basic prompt and an advanced prompt to generate separate documents. From these sources, we collected six sets of PEMs about five acute pediatric conditions, evaluated them using a panel of readability indices, and used ANOVA to compare average readability scores between sources. Two pediatricians assessed understandability and actionability using the Patient Education Materials Assessment Tool (PEMAT). We calculated mean PEMAT scores, with values > 70% considered understandable/actionable. RESULTS:ChatGPT (advanced prompt) created PEM with the lowest mean grade reading level (5.8; SD 0.84) and highest PEMAT scores (91% understandable, 84% actionable). Documents from generative AI using the advanced prompt and UpToDate had mean readability scores ≤ 6th grade, with significantly better readability than AI-generated documents using the basic prompt or HealthyChildren.org. All sources except HealthyChildren.org had mean understandability > 70%; AIs using the advanced prompt and UpToDate exceeded the actionability threshold. CONCLUSIONS:A plain-language prompt for generative AI created PEMs that were readable, understandable, and actionable, with comparable results across two AI programs. PRACTICE IMPLICATIONS/CONCLUSIONS:Healthcare providers can use this new plain-language prompt for generative AI programs to quickly create free, patient-friendly PEM about acute healthcare topics. As these AI programs were not trained to provide medical guidance, medical professionals should continue to review PEM prior to sharing with patients.

This study examines mental health disparities among transgender, nonbinary, and gender-expansive (hereinafter "trans") patients within a large urban health care system in the Bronx, New York-a relatively trans-protective legislative environment-amid escalating federal anti-trans legislation. Using electronic health record data (2015-2023) from a large safety-net health care system, we conducted a retrospective cohort study comparing mental health outcomes and services among trans (n = 2,898) and cisgender (n = 29,818) adult patients. We applied the Institute of Medicine rank-and-replace method to isolate disparities attributable to systemic inequities independent of clinical need. Despite greater engagement, trans patients faced disproportionately poor outcomes. Trans patients had more than twice the prevalence of mental health concerns compared with cisgender patients (all ps < .001), including higher rates of mood disorders (10.9% vs. 2.6%) and anxiety disorders (9.7% vs. 3.1%). Trans patients had higher predicted means of primary care utilization (40.92% vs. 38.45%, disparity: 2.48 percentage points, p < .001), minimally recommended mental health treatment (21.18% vs. 19.44%, disparity: 1.74 percentage points, p < .001), suicide-related visits (2.3% vs. 1.2%, disparity: 1.1 percentage points, p < .001), emergency department mental health visits (8.6% vs. 6.4%, disparity: 2.2 percentage points, p < .001), and inpatient mental health stays (0.13 vs. 0.09 stays per year, p < .001). Intersectional analyses revealed compounded disparities for Black and Hispanic patients. These findings highlight a persistent "care paradox," wherein greater health care utilization fails to mitigate adverse outcomes. Results underscore the need for integrated, trans-affirming interventions at the health care system level, broader structural reforms, and community-driven efforts to bolster trans mental health. (PsycInfo Database Record (c) 2026 APA, all rights reserved).

BACKGROUND:Non-English-speaking individuals have noted disparities in health outcomes in the United States, but the role of language in obstetric settings is not well characterized. OBJECTIVE:We examined the association between primary spoken (preferred) language and the risk of nulliparous, term, singleton vertex cesarean delivery. STUDY DESIGN/METHODS:We conducted a retrospective cohort study of nulliparous, term, singleton vertex deliveries in California (2016-2021) using birth certificates linked to hospital discharge records. The primary outcome was the risk of cesarean delivery. Poisson log-linear regression was used to estimate the association between preferred language and cesarean delivery overall, adjusted for demographic, hospital, and obstetric factors. Analyses were stratified by labor type (pre-labor versus intrapartum), risk characteristics (standard versus high risk), and hospital characteristics (hospital type [academic versus nonacademic], setting [urban versus non-urban], and annual hospital cesarean delivery volume by sample tertiles). RESULTS:Among 721,263 birthing people in California, 87,538 (12.1%) indicated that their preferred language was not English and 359,988 (49.9%) had no measured medical comorbidities (termed standard risk). Non-English speaking patients had a decreased risk of cesarean delivery (adjusted relative risk 0.94, 95% confidence interval 0.92-0.96), including among Chinese- (adjusted relative risk 0.83, 95% confidence interval 0.79-0.86) and Spanish-speaking patients (adjusted relative risk 0.94, 95% confidence interval 0.93-0.97), although non-English speaking birthing people who preferred languages other than Spanish or Chinese had slightly higher risk of cesarean delivery (adjusted relative risk 1.06, 95% confidence interval 1.02-1.10). There was a decreased cesarean risk among non-English speaking patients across labor types (spontaneous labor or prelabor rupture of membranes: adjusted relative risk 0.94, 95% confidence interval 0.89-1.00; induced labor: adjusted relative risk 0.90, 95% confidence interval 0.87-0.93; no indication of either: adjusted relative risk 0.97, 95% confidence interval 0.95-1.00), among standard risk patients (adjusted relative risk 0.95, 95% confidence interval 0.92-0.97), and across hospital type, setting, and annual cesarean delivery volume tertile. A small decrease in cesarean risk was also observed in teaching (adjusted relative risk 0.93, 95% confidence interval 0.91-0.96) and non-teaching hospitals (adjusted relative risk 0.95, 95% confidence interval 0.92-0.97), urban settings (adjusted relative risk 0.93, 95% confidence interval 0.92-0.95), and highest volume tertile (adjusted relative risk 0.93, 95% confidence interval 0.91-0.95), though there was no increase risk in non-urban settings (adjusted relative risk 1.02, 95% confidence interval 0.95-1.11) and across lowest and middle volume tertiles (adjusted relative risk 1.01, 95% confidence interval 0.92-1.11; and adjusted relative risk 0.95, 95% confidence interval 0.91-0.98, respectively). CONCLUSION/CONCLUSIONS:Non-English language preference was associated with a slight decreased risk for cesarean delivery among nulliparous, term, singleton, and vertex presenting patients, and an increased risk among people who spoke other languages (i.e., not Chinese or Spanish). Our findings suggest that populations that speak less common non-English languages in California may be at elevated risk of disparate cesarean delivery outcomes.

Socioenvironmental disasters such as wildfires, floods, extreme heat events, and zoonotic epidemics have been increasing in magnitude and frequency. These events are sometimes are followed by increases in overdose mortality. This paper applies an adaptation of the Big Events framework, originally developed to understand macrosocial changes and their effects on HIV, to guide research into the pathways through which socioenvironmental disasters may cause increases in overdose mortality. These pathways include economic disruption, health and social service disruption, population displacement, changes in social networks, and social movements. They also include processes and conditions more causally proximal to overdose events such as housing insecurity and changes in drug markets and drug-use behaviors. The paper also discusses relevant existing measures and variables that must be developed to conduct research on the effects of socioenvironmental disasters on overdose The end goal of this application of the Big Events framework is to encourage scholarship that identifies potential interventions for people who use drugs, public and private service agencies and governments to mitigate the harms that socioenvironmental disasters can cause for people who use drugs.

OBJECTIVE/UNASSIGNED:This study examined decade-long trends and differences in sodium-glucose cotransporter 2 inhibitor (SGLT2i) and glucagon-like peptide-1 receptor agonist (GLP-1 RA) prescriptions among adults with type 2 diabetes using real-world data from Epic Cosmos. RESEARCH DESIGN AND METHODS/UNASSIGNED:We analyzed electronic health records of 1 517 594 adults with type 2 diabetes without end-stage renal disease from 2014 to 2024. Annual prescribing trends were evaluated by patient race and insurance type using negative binomial regression. Medication exposure was defined using active prescriptions/orders recorded in Epic Cosmos during the calendar year. In pooled descriptive analyses, we also characterized patients prescribed these medications by clinical characteristics, neighborhood-level social vulnerability, and prescriber specialty. RESULTS/UNASSIGNED:From 2014 to 2024, SGLT2i use rose from 0.5% to 12.1% and GLP-1 RA use increased from 0.9% to 15.9%. Black patients had consistently lower prescription rates than White patients across insurance groups. Primary care physicians prescribed about one-third of these medications. In pooled descriptive analyses, endocrinology was associated with higher observed prescribing rates than primary care or cardiology. Patients from neighborhoods with lower social vulnerability were more likely to receive these therapies. CONCLUSIONS/UNASSIGNED:Use of SGLT2i and GLP-1 RA increased substantially over the past decade, significant racial, socioeconomic, and insurance-related differences persist in prescribing these therapies.

OBJECTIVES/OBJECTIVE:To estimate past-year prevalence and correlates of major depressive episode (MDE), characterize mental health treatment utilization, and identify barriers to care among community-dwelling US adults aged ≥50. DESIGN/METHODS:Cross-sectional analysis of a nationally representative survey. SETTING/METHODS:Noninstitutionalized US population. PARTICIPANTS/METHODS:20,580 respondents aged ≥50 from 2022 to 2023 National Survey on Drug Use and Health. MEASUREMENTS/METHODS:We estimated past-year MDE prevalence and delineated demographic and past-year substance use correlates; examined mental health treatment utilization among individuals with MDE; and identified barriers to care among untreated individuals. RESULTS:An estimated 4.6% (95% confidence interval: 4.2-5.2) of adults aged ≥50 experienced past-year MDE. In multivariable analyses, higher odds were estimated among females, divorced/separated or widowed individuals, college-educated individuals, and those with alcohol use disorder, cannabis use (with or without use disorder), and prescription opioid or tranquilizer/sedative misuse. Lower odds were estimated among adults aged ≥65, Non-Hispanic Black individuals, and those with income ≥$50,000. Among those with MDE, 66.2% received any mental health treatment. In bivariable comparisons, adults aged ≥65 were less likely than those aged 50-64 to receive outpatient mental health services (38.7% versus 50.1%, p = 0.032) and to report perceived unmet need for treatment (8.8% versus 23.3%, p = 0.019). Among untreated individuals with perceived unmet need, barriers to treatment were attitudinal (86.2%; e.g., thought could handle without treatment), access-related (73.1%), and financial (48.6%). CONCLUSIONS:MDE is prevalent among older adults with heterogeneity across demographic and substance use factors. Treatment disparities and high barrier rates highlight the need for targeted interventions in this population.

Gun violence is a pervasive and multifaceted public health problem. This study describes a novel application of group model building-a process that centers stakeholder voices to describe a complex problem-to investigate drivers of gun violence and design community-led interventions for prevention in New Haven, Connecticut. We recruited community leaders to participate in 2 group model building sessions in New Haven. Discussions and diagrams created during the sessions were used to produce a causal loop diagram illustrating proposed relationships between gun violence and other variables. Our resulting "iceberg model" illustrates how short-term and visible responses to gun violence "above the surface," such as incarceration and policing, may only transiently decrease gun violence. Long-term, downstream effects of "above the surface" responses may ultimately increase gun violence and are summarized in 3 layers "under the surface," which examine the effects of incarceration and lack of re-entry support, vulnerability to stress and social stigmatization, and community-police relations and community safety, respectively. This study demonstrates the value of using group model building to design community-driven interventions for gun violence prevention, as well as reiterates the importance of targeting structural causes of gun violence. This work has informed TRUE HAVEN, a multi-level intervention that addresses community gun violence in New Haven.

Adverse social determinants of health (SDOH) associate with greater heart failure (HF) risk. Among 9879 community-based participants of the Atherosclerosis Risk in Communities (ARIC) cohort study, we assessed 4955 plasma proteins using an aptamer-based platform (SomaLogic). We derived an SDOH factor comprising income, education, and area deprivation index (ADI) in Black and white participants. Proteins associated with this factor at Bonferroni significance were tested for associations with incident HF using multivariable Cox proportional hazard models. Among Black participants, 36 proteins were associated with both the SDOH factor and incident HF, 126 were among white participants, and 12 were shared between race strata. Eight of these demonstrated significant mediation effect in causal mediation models. Key results were replicated in 49,396 participants in UK Biobank. Mendelian randomization and colocalization suggested a potentially causal effect of C1q tumor necrosis factor-related protein 1 (C1QTNF1), an adiponectin paralog, on HF. We demonstrate protein biomarkers associated with SDOH burden and HF risk.

BACKGROUND/UNASSIGNED:Early childhood is a critical window for cardiac development, yet the impact of prenatal factors and early growth patterns on heart structure and function in children with structurally normal hearts remains incompletely described. This study investigates the roles of birth weight (BW), gestational age (GA), birthweight-for-gestational-age (BW/GA) z-score, and weight gain z-score in shaping cardiac dimensions. METHODS/UNASSIGNED:-scores for cardiac dimensions and volumes were used to standardize for body surface area (BSA). BW, BW/GA z-score, GA, and weight gain z-score were analyzed as predictors of these parameters. RESULTS/UNASSIGNED:Higher weight gain z-score was independently associated with smaller cardiac dimension z-scores, particularly in the aortic root, ascending aorta, left atrium diameter, left ventricular internal diameter at end-diastole, left ventricular internal diameter at end-systole, end-systolic volume, end-diastolic volume, and left ventricular mass. While birth characteristics influenced some cardiac parameters, postnatal weight gain emerged as a more significant predictor of cardiac size. Ejection fraction and fractional shortening were not influenced by the predictors in our analysis. CONCLUSIONS/UNASSIGNED:Our findings suggest that postnatal weight gain was more consistently associated with cardiac structure z-scores than anthropometric characteristics at birth. Specifically, contractile myocardial thicknesses were relatively stable across weight gain patterns, whereas vascular structures showed inverse scaling with higher weight gain.