Artificial intelligence based clinical exam training significantly improved medical students’ standardized clinical exam performance
1. Lavigne and colleagues evaluated whether an AI-powered training platform could improve Objective Structured Clinical Examination (OSCE) performance.
2. Students with access to AI-powered OSCE training achieved higher OSCE scores and greater emotional readiness as compared to controls.
Evidence Rating Level: 1 (Excellent)
Study Rundown: OSCEs are commonly used to teach and assess clinical competence among medical students, but their reliance on trained examiners and standardized patients makes frequent practice resource-intensive. Lavigne and colleagues assessed an artificial intelligence (AI) platform’s ability to improve students’ performance on faculty-administered OSCEs. Participants were early-clerkship students who received either access to 16 text-based, GPT-4-powered simulated patient scenarios with automated feedback or standard training alone (control). The primary outcome was the median faculty score across clinical OSCE stations. Secondary outcomes included emotional readiness, stress, perceived realism, satisfaction, and platform engagement. Students assigned to the AI training platform achieved higher median OSCE scores than controls (median 11.4 versus 10.7; p = 0.02) and reported greater readiness and lower stress. Additionally, the majority of the students in the AI-training group felt interactions with AI were realistic. This study demonstrated that AI-based OSCE training offered a scalable adjunct to conventional OSCE preparation.
Click here to read the study in NEJM AI
Relevant Reading: Artificial Intelligence and Objective Structured Clinical Examinations: Using ChatGPT to Revolutionize Clinical Skills Assessment in Medical Education
In-Depth [randomized controlled trial]: 251 fourth-year medical students were enrolled in the study and were randomized 1:1 to the AI-Standardized Clinical Examination (ASCE) group or the control group. All students were given 2.5 months to prepare for the OSCE. ASCE users could complete 16 text-based simulated encounters and receive GPT-4-generated scores and written feedback. During the faculty-administered OSCE, all students were assigned to one of two sessions containing two 8-minute clinical stations. However, everyone completed an otoscopy physical exam station as a control. Evaluators, investigators, and data analysts were blinded to the allocation. The primary outcome was the median score across OSCE stations, graded out of 20 using standardized faculty checklists and ability scales. Secondary outcomes were measured using five-point Likert surveys and in-platform feedback. Participants in the ASCE group achieved significantly higher OSCE scores than those in the control group (median 11.4 vs 10.7, 95% confidence interval [CI] 0-1.2). No differences between the two groups were found for the otoscopy station. After the OSCE, emotional readiness scores were higher in the intervention than in the control group (median 3 vs 2, respectively; 95% CI 1-2). Stress level scores were significantly lower in the intervention group than in the control group (median 4 vs 5, respectively; 95% CI −1-0). This study was limited by possible crossover, as control group students may have accessed AI tools. Nonetheless, this study provided promising evidence that AI software can improve OSCE performance among medical students.
Image: PD
©2026 2 Minute Medicine, Inc. All rights reserved. No works may be reproduced without expressed written consent from 2 Minute Medicine, Inc. Inquire about licensing here. No article should be construed as medical advice and is not intended as such by the authors or by 2 Minute Medicine, Inc.
