Faculty Bibliography

We developed and implemented a PrEP navigation program ("SNAPS") in a NYC safety-net hospital with the objectives to co-locate navigation, clinical PrEP services, and payment assistance. Adherence and retention to PrEP-related care were assessed by mean medication possession ratios (MPRs) and number of appointments over 12 months. Compared to the pre-SNAPS cohort, the post-SNAPS cohort was less likely to be cisgender male (64.8% vs. 84.2%), White (6.5% vs. 23%) and to speak English (33.3% vs. 80.6%) (all p < 0.001). Mean MPR was lower for post-SNAPS (0.68, SD = 0.33) compared to pre-SNAPS (0.89, SD = 0.22) (p = 0.001). Among post-SNAPS patients, cisgender men and MSM were more likely to be retained in PrEP care compared to cisgender women (p < 0.05). Although SNAPS linked diverse patients to PrEP-care, mean MPR was lower post-SNAPS compared to the pre-SNAPS. Continued investments to strengthen later stages of the PrEP cascade model for all populations vulnerable to HIV are needed.

Early identification of inpatient discharges to skilled nursing facilities (SNFs) facilitates care transition planning. Predictive information in admission history and physical notes (H&Ps) is dispersed across long documents. Language models adeptly predict clinical outcomes from text but have limitations: token length constraints, noisy inputs, and opaque outputs. Therefore, we developed extraction-based language model classification (ELC): generative language models distill H&Ps into task-relevant categories ("Structured Extracted Data") before summarizing them into a concise narrative ("AI Risk Snapshot"). We hypothesized that language models utilizing AI Risk Snapshots to predict SNF discharges would perform the best. In this retrospective observational study, nine language models predicted SNF discharges from unstructured predictors (raw H&P text, truncated assessment and plan) and ELC-derived predictors (Structured Extracted Data, AI Risk Snapshots). ELC substantially reduced input length (AI Risk Snapshot median 141 tokens vs raw H&P median 2,120 tokens) and improved average AUROC and AUPRC across models. The best performance was achieved by Bio+Clinical BERT fine-tuned on AI Risk Snapshots (AUROC = .851). AI Risk Snapshots enhanced interpretability by aligning with nurse case managers' risk assessments and facilitating prompt design. Structuring and summarizing H&Ps via ELC thus mitigates the practical limitations of language models and improves SNF discharge prediction.

BACKGROUND:Chinese immigrants face numerous social determinants of health (SDOH) challenges that limit access to evidence-based diabetes self-management education and support programs (DSMES). To address these challenges, our team developed the LINK-IT intervention. This manuscript presents the study protocol for the LINK-IT trial. METHODS:The LINK-IT trial is a 12-month, 3-arm randomized controlled trial aiming to enroll 405 Chinese immigrants with T2D (HbA1c≥7%) from multiple community and clinical settings in New York City. A total of 405 participants will be randomly allocated to one of three groups (n = 135 per group): (1) video-based DSMES plus community health worker (CHW) support (VIDEO+CHW), (2) video-based DSMES only (VIDEO), or (3) wait-list control (CONTROL). The VIDEO+CHW group will receive 24 culturally and linguistically tailored DSMES videos (one per week for 24 weeks) delivered via text message links, along with biweekly (every other week) phone calls from trained CHWs to review video content, support goal setting, and address SDOH barriers. The VIDEO group will receive the same video intervention without CHW support. The CONTROL group will receive usual care and will be offered access to the videos upon study completion. The primary outcome is the change in HbA1c at 6 months. Secondary outcomes include changes in HbA1c at 12 months, self-efficacy for diabetes, dietary intake, physical activity, medication adherence and emotional support at 6 and 12 months. Data will be analyzed using an intention-to-treat approach with linear mixed-effects models. ETHICS AND DISSEMINATION/BACKGROUND:This study protocol has been approved by the Institutional Review Board of the NYU Grossman School of Medicine (S23-01274). All study procedures will adhere to the ethical principles outlined in the Declaration of Helsinki. Written or verbal informed consent will be obtained from all participants. Study results will be disseminated through peer-reviewed publications, presentations at scientific conferences, and community events. TRIAL REGISTRATION/BACKGROUND:The LINK-IT trial was registered on March 20, 2024, on ClinicalTrials.gov under the identifier NCT06319716; https://clinicaltrials.gov/study/NCT06319716.

BACKGROUND:The COVID-19 pandemic had a significant impact on healthcare delivery. Older adults with multimorbidities were at risk of healthcare disruptions for the management of their chronic conditions. OBJECTIVE:To characterize healthcare disruptions during the COVID-19 healthcare shutdown and recovery period (March 7, 2020-October 6, 2020) and their effects on disease management among older adults with multimorbidities who were patients of NYC Health + Hospitals (H + H), the largest municipal safety-net system in the United States. DESIGN/METHODS:Observational. PATIENTS/METHODS:Patients aged 50 + with hypertension or diabetes and at least one other comorbidity, at least one H + H ambulatory visit in the six months before COVID-19 pandemic onset (March 6, 2020), and at least one visit in the post-acute shutdown period (October 7, 2020 to December 31, 2023). MAIN MEASURES/METHODS:We characterized disruption in care (defined as no ambulatory or telehealth visits during the acute shutdown) and estimated the effect of disruption on blood pressure control, hemoglobin A1c (HbA1c), and low-density lipoprotein (LDL) cholesterol using difference-in-differences models. KEY RESULTS/RESULTS:Out of 73,889 individuals in the study population, 12.5% (n = 9,202) received no ambulatory or telehealth care at H + H during the acute shutdown. Low pre-pandemic healthcare utilization, Medicaid insurance, and self-pay were independent predictors of care disruption. In adjusted analyses, the disruption group had a 3.0-percentage point (95% CI: 1.2-4.8) greater decrease in blood pressure control compared to those who received care. Disruption did not have a significant impact on mean HbA1c or LDL. CONCLUSIONS:Care disruption was associated with declines in blood pressure control, which while clinically modest, could impact risk of cardiovascular outcomes if sustained. Disruption did not affect HbA1c or LDL. Telehealth mitigated impacts of the pandemic on care disruption and subsequent disease management. Targeted outreach to those at risk of care disruption is needed during future crises.

OBJECTIVES/OBJECTIVE:Documentation of gender identity (GI) and anatomy data in the electronic health record (EHR) is a proposed standard of care for transgender populations. However, there is limited research on implementation of proposed best practices, particularly anatomy data collection. This study aims to characterize factors that influence patient preferences and comfort around the collection and documentation of GI and anatomy in EHRs. MATERIALS AND METHODS/METHODS:From November 2023 to January 2024, 17 one-on-one, semi-structured virtual interviews were conducted with transgender adults residing in the Metropolitan New York area. Transcriptions were analyzed using inductive thematic analysis. RESULTS:Themes clustered around comfort and preferences for data collection processes and outcomes. Factors that influenced preferences and comfort around anatomy data were distinct from those impacting GI documentation preferences and comfort. The tension between the categories of GI and sex assigned at birth impacted anatomy data documentation preferences. Clinical context emerged as a consistent factor that impacts both preferences and comfort of GI and anatomy data documentation. DISCUSSION AND CONCLUSION/CONCLUSIONS:GI data collection efforts in clinical settings must consider the implication of anatomy data collection when determining data collection best practice methodologies. Anticipated and experienced stigma remain significant hurdles to patient comfort and willingness to collect GI and anatomy data, and their impact on actual data collection should be further elucidated among diverse gender identities. Clinical data collection methods, tools, and education warrant ongoing research investment to further elucidate best practices.

Consumption of a western diet and high body mass index (BMI) are risk factors for progression from pre-malignant phenotypes to multiple myeloma, a hematologic cancer. In the NUTRIVENTION trial (NCT04920084), we administered a high-fiber, plant-based diet (meals for 12 weeks, coaching for 24 weeks) to 23 participants with myeloma precursor states and elevated BMI. The intervention was feasible, improved quality of life and modifiable risk factors: metabolic (BMI, insulin resistance), microbiome (diversity, composition), and immune (inflammation, monocyte subsets). Disease-progression trajectory improved (n=2) or was stable. Findings were translated to Vk*MYC mice modeling the myeloma-precursor state, in which a high-fiber diet delayed disease progression through improved metabolism and microbiome composition leading to increased short-chain fatty acid production that reinvigorated anti-tumor immunity and inhibited tumor growth. These effects from fiber consumption were independent of calorie restriction and weight loss. A high-fiber diet is a low-risk intervention that may delay progression to myeloma.

OBJECTIVE:Continuous glucose monitoring (CGM) is essential in diabetes care and research; however, extracting key data (e.g., time above, in, or below range) from CGM reports is manual, time-consuming, and inefficient. Natural language processing (NLP) can extract data from unstructured sources (e.g., images), but its application in CGM remains unexplored. We aimed to evaluate the accuracy of extracting CGM data using NLP. RESEARCH DESIGN AND METHODS/METHODS:We analyzed CGM reports stored as PDF files from the electronic health record at New York University Langone Health. The steps of our algorithm pipeline consist of 1) performing optical character recognition (OCR) to obtain glucose matrix data from CGM reports, 2) determining the type of CGM documents based on keywords in OCR results, 3) extracting variables of glucose based on CGM document type, and 4) storing the extracted glucose data in a structured database. Two experts with experience in CGM research and clinical practice conducted an independent manual review of 1% of the documents (n = 226). We calculated accuracy (correct extraction of CGM data) by comparing the algorithm's results with the manual review. RESULTS:Of the documents analyzed, 36.8% were Freestyle Libre and 63.2% were Dexcom. For information extraction, the agreement in evaluating Libre results between two experts was 99.93%. When comparing algorithm accuracy with manual review, the accuracy for Libre was 99.87% and, for Dexcom, 100.00%. CONCLUSIONS:Using an NLP approach to extract valuable glucose data from CGM PDF files is feasible and accurate, which can benefit clinical practice and diabetes research.