Student Perceptions of a Custom Artificial Intelligence Clinical Case Companion

Abstract

INTRODUCTION: Artificial intelligence tools show promise in supplementing traditional physician assistant education, particularly in developing clinical reasoning skills. However, limited research exists on custom Generative Pretrained Transformer (GPT) applications in physician assistant (PA) education. This study evaluated student experiences and perceptions of a custom GPT-based clinical reasoning tool. METHODS: A mixed-methods study was conducted with first-year PA students (n = 72) at Butler University in April 2025. Students engaged in a custom GPT-4-Turbo Butler University PA Clinical Case Companion, designed to deliver interactive clinical cases using Socratic dialogue across hematology and nephrology specialties. Postengagement surveys assessed students' perceptions of the tool's helpfulness, cognitive challenge, and clinical skill development across 7 clinical domains. Surveys used Likert scales and open-ended questions. Researchers applied descriptive analysis on quantitative data and thematic analysis to qualitative responses. RESULTS: Fifty-nine students completed surveys (81.9% response rate). Ninety percent of respondents rated the tool moderately to extremely helpful. Students reported significant improvement in developing differential diagnoses (79.7%), ordering diagnostic studies (81.4%), and interpreting tests (78.0%). Qualitative analysis revealed 3 primary themes: appreciation for immediate feedback (36%), detailed real-time explanations (36%), and receiving GPT-initiated student prompting (12%). Students suggested earlier curriculum integration and expressed concerns about the accuracy of case content. DISCUSSION: Custom GPT-based clinical reasoning tools can serve as an adjunct to traditional PA educational methods by offering personalized, on-demand learning experiences. Students perceived substantial benefits in developing clinical reasoning skills, but noted limitations in history-taking skills. Implementation should include faculty oversight and artificial intelligence literacy training to address accuracy concerns while maximizing educational benefits.

Ähnliche Arbeiten