Faculty Bibliography

Acute coronary syndrome (ACS) requires prompt treatment, yet management delays are difficult to identify. In this study, we developed a large language model (LLM) system to identify ACS management delays and characterized delay cases. Admissions to internal medicine residents at NYU July 2022-June 2025 (n=4,642) were included. Prompts were validated to determine if the resident admission note documented initiation of ACS management and if the initial cardiology consult note documented initiation of ACS management (ground truth) (n=161 for each). Discordant cases were reviewed by three physicians using a validated tool to confirm management delays. Demographics and key clinical findings of patients with and without delays were compared. The LLM identified management delays with a 52% positive predictive value (n=35/67). Patients who were older, females, and with preferred language other than English or Spanish were more likely to have a management delay (73.4 ± 15.3 vs 68.5 ± 12.6 years-old, p=0.036, 56.8% vs 34% females, p=0.014, and 27.0% vs 15.5% other preferred language, p=0.046, in management delay vs non-management delay cases). The management delay group had longer average time in hours to receiving heparin, aspirin, and cardiac catheterization (56.91 ± 56.78 vs 18.97 ± 13.76, p<0.001, 13.94 ± 16.64 vs 8.23 ± 9.82, p=0.005, and 65.12 ± 51.65 vs 39.51 ± 44.19, p=0.006, respectively in management delay vs non-management delay cases). In conclusion, the LLM-based system we developed to identify ACS management delays can detect cases at scale to inform individual and systems-level interventions to improve quality of ACS care.

UNLABELLED:Student feedback is essential to curriculum evaluation. While methods for analyzing quantitative feedback data are readily available and easy to implement, methods for analyzing text-based, qualitative feedback data are less widely available, requiring more time, effort, and expertise. And yet, students' responses to open-ended questions hold great value for curriculum refinement because narrative comments can identify un-anticipated areas of concern that closed-ended rating scales might miss and often provide specific suggestions for improvement. In this paper, we describe efforts to analyze the feasibility and accuracy of using a Large Language Model (ChatGPT 4o) to analyze medical student comments in response to a question asking them to identify basic science topics they found challenging. ChatGPT 4o was used to categorize and summarize students' identification of and explanations for these challenging topics. We describe the specific prompts used to generate and refine results and then conducted a series of experiments to explore consistency, accuracy, and meaningfulness: (1) reviewing the consistency of 10 replications of the ChatGPT 4o request; (2) comparing "expert" human ratings of topic categories with ChatGPT's categorization; and (3) comparing "expert" human analyses of the explanations for a challenging topic with those generated by ChatGPT. Overall, we found the LLM output to be useful, fairly closely aligned with human experts, and easy to implement. However, results were not perfectly replicated across multiple trials and we found some differences between human and LLM analyses. Our use case is well suited to the current capabilities of genAI models in that summaries can be rapidly and easily generated with sufficient (but not perfect) consistency and accuracy to support continuous quality improvement of basic science curriculum. SUPPLEMENTARY INFORMATION/UNASSIGNED:The online version contains supplementary material available at 10.1007/s40670-025-02615-1.

Over the last decade there has been tremendous growth in the development of accelerated MD pathways that allow medical students to graduate in three years. Developing an accelerated pathway program requires commitment from students and faculty with intensive re-thinking and altering of the curriculum to ensure adequate content to achieve competency in an accelerated timeline. A re-visioning of assessment and advising must follow and the application of AI and new technologies can be added to support teaching and learning. We describe the curricular revision to an accelerated pathway at NYU Grossman School of Medicine highlighting our thought process, conceptual framework, assessment methods and outcomes over the last ten years.

BACKGROUND:The association between coronavirus disease 2019 (COVID-19) and hypercoagulability has been extensively described, and pulmonary embolism is a recognized complication of COVID-19. Currently, the need for computed tomography pulmonary angiogram (CTPA) relies on the Wells score and serum D-dimer levels. However, because COVID-19 patients have a different thrombotic and inflammatory milieu, the usefulness of the Wells score deserves further exploration for this patient population. We aimed to explore the ability of the Wells score to predict pulmonary embolism in patients with COVID-19. METHODS:In this retrospective study, patients found to have a CTPA and a COVID-19 diagnosis during the same admission were selected for analysis. Age and sex, CTPA results, and associated D-dimer levels were entered in a database. The Wells score sensitivity and specificity were calculated at different values, and the area under the curve of the receiver operating characteristic curve measured. RESULTS:Of 459 patients with COVID-19, 64 had a CTPA and 12 (19%) had evidence of pulmonary embolism. Previous or current evidence of deep vein thrombosis, a Wells score above 4 points, and serum D-dimer levels 5 times above age-adjusted upper normal values were associated with pulmonary embolism. However, only 33% of patients with pulmonary embolism had a Wells score of 4 points or higher. The area under the curve of the receiver operating characteristic showed non-discriminating values (0.54) CONCLUSIONS: Although a Wells score of 4 or more points predicted pulmonary embolism in our cohort, the outcome can be present even with lower scores.

Dual anti-platelet therapy denotes a regimen of aspirin plus a P2Y₁₂ receptor inhibitor, clopidogrel, prasugrel, or ticagrelor. Such therapy is a cornerstone of medical management following acute coronary syndromes and is imperative following percutaneous coronary interventions. While there is uncertainty about the optimal duration of dual antiplatelet therapy following percutaneous coronary intervention, the new 2016 American College of Cardiology/American Heart Association Guidelines suggest that for patients with stable ischemic heart disease at least six months of such therapy following a drug eluting stent and one month following a bare metal stent should be implemented. In patients with acute coronary syndrome including non-ST elevation and ST elevation myocardial infarction it is recommended to extend dual antiplatelet therapy treatment to one year in both drug eluting stent and bare metal stent groups. There may be latitude for earlier discontinuation in appropriately selected patients; extended dual antiplatelet therapy beyond one year may be beneficial in others. Herein, we describe current guidelines and evidence supporting if and when dual antiplatelet therapy should be interrupted for surgery for patients who have undergone percutaneous coronary intervention.

Gene conversion, the non-reciprocal exchange of genetic information, is one of the potential products of meiotic recombination. It can shape genome structure by acting on repetitive DNA elements, influence allele frequencies at the population level, and is known to be implicated in human disease. But gene conversion is hard to detect directly except in organisms, like fungi, that group their gametes following meiosis. We have developed a novel visual assay that enables us to detect gene conversion events directly in the gametes of the flowering plant Arabidopsis thaliana. Using this assay we measured gene conversion events across the genome of more than one million meioses and determined that the genome-wide average frequency is 3.5×10(-4) conversions per locus per meiosis. We also detected significant locus-to-locus variation in conversion frequency but no intra-locus variation. Significantly, we found one locus on the short arm of chromosome 4 that experienced 3-fold to 6-fold more gene conversions than the other loci tested. Finally, we demonstrated that we could modulate conversion frequency by varying experimental conditions.