Lavigne E, Lopez A, Frandon J, Blaizot G, Gabellier L, Adham S, Ursic Bedoya J, Charriot J, Thieblemont T, Badreddine M, Guenou E, Carbonnel F, Duflos Claire, Morin D, Laffont I, Pers Y, Yauy K. AI-Standardized Clinical Examination Training on OSCE Performance. NEJM AI. 2025 Jul. doi: 10.1056/AIoa2500066


Abstract

Background: Objective structured clinical examinations (OSCEs) are a critical but resource-intensive means of training medical students and assessing their clinical competence. The potential of artificial intelligence (AI)–driven training methods in the field of OSCEs remains underexplored. We evaluate the impact of an AI-standardized clinical examination (ASCE) methodology, powered by Generative Pretrained Transformer 4 (DocSimulator platform), on medical students’ OSCE performance. Methods: In this single-blind, two-site randomized controlled trial (from November 2023 to January 2024), we recruited 247 early-clerkship students. Participants were randomly assigned to receive 2.5 months of either ASCE training (n = 125) or standard training (n = 122). The primary outcome was the faculty OSCE score. Secondary outcomes included emotional readiness and stress levels prior to the examination, perceived realism of the simulation on the platform, and student satisfaction. Results: ASCE users scored significantly higher on the OSCE than the control group (median 11.4 vs. 10.7; P = 0.02; 95% confidence interval [CI], 0 to 1.2). Absenteeism was prevented in the intervention group (0.8% vs. 4.9%). Postintervention five-point Likert surveys (72 ASCE users, 81 control) revealed that ASCE improved emotional readiness (P < 0.001; 95% CI, 1 to 2) and reduced stress (P = 0.02; 95% CI, −1 to 0). In-session feedback from 459 participants indicated that 74% of ASCE users felt they were interacting with real patients and 67% felt there were being evaluated by teachers, despite AI-driven interactions and assessments. Conclusions: ASCE training enhances clinical skill development while addressing OSCE resource challenges. Its scalability and ability to simulate diverse scenarios position it as a transformative tool in medical education and practice. (Funded by the University of Montpellier I-SITE Excellence Program and Booster Innovation Montpellier. Research Ethics Committee of Montpellier University; trial registration number, UM 2023-039bis.) Link to paper : https://ai.nejm.org/doi/full/10.1056/AIoa2500066