SHM Converge Scientific Abstract ePoster Gallery-Enhancing Diagnostic Reasoning in Third Year Medical Students Through Ai-Augmented Mystery Case Learning

Pdf Summary

This document describes a medical education study evaluating whether large language models (LLMs) can help assess and improve third-year medical students’ diagnostic reasoning through AI-augmented “mystery case” learning.<br /><br />The project involved presenting students with patient-style vignettes and asking them to generate a differential diagnosis list of 10 specific diseases, ranked from most to least likely. The prompts were intentionally designed to be clinically challenging and to encourage structured reasoning.<br /><br />The study used three LLMs: ChatGPT, OpenEvidence, and Gemini. These models were queried with illustrative patient statements, including:<br /><br />- A 51-year-old woman with chronic diarrhea, decreased appetite, concentric left and right ventricular hypertrophy, and nephrotic-range proteinuria.<br />- A 40-year-old man with hypertension, asthma, moderate alcohol use, altered mental status, acute liver failure, markedly elevated AST/ALT and alkaline phosphatase, pancytopenia, and hyperbilirubinemia.<br />- A 19-year-old man with shortness of breath, cough, a right-sided exudative pleural effusion, normal white blood cell count, and negative initial infectious workup.<br /><br />The title and framing suggest the main conclusion is that large language models may be effective tools for evaluating growth in medical students’ diagnostic reasoning during structured educational exercises. The study appears to explore both the educational value of these models and their ability to generate clinically plausible differential diagnoses.<br /><br />Overall, the document highlights an emerging use case for LLMs in medical education: supporting case-based learning, prompting deeper diagnostic thinking, and potentially serving as scalable tools for assessment and feedback.

Asset Subtitle

Chris Snyder

Meta Tag

Author List Chris Snyder, Keegan Schuchart, Ryder Cuppett

Category Innovations

Concept AI-augmented Mystery Case Learning

Concept Diagnostic Reasoning

Concept Differential Diagnosis

Concept Clinical Findings

Concept Medical Students

Distinguished Non-Finalist

Presenter Organization University of Nebraska Medical Center

Presenting Author Chris Snyder

Track Education

Keywords

medical education

large language models

diagnostic reasoning

differential diagnosis

third-year medical students

mystery case learning

AI-augmented learning

ChatGPT

OpenEvidence

Gemini

AI-augmented Mystery Case Learning

Diagnostic Reasoning

Differential Diagnosis

Clinical Findings

Medical Students