Implementation of an AI-simulated consultation platform for undergraduate clinical communication training: a service evaluation

Résumé fourni par la source

Abstract Background AI-simulated consultation platforms are increasingly used to support clinical communication training in medical education, with systematic reviews reporting comparable or improved skills outcomes across digital simulation modalities [1, 2]. However, much of the existing evidence derives from controlled trials focused on educational efficacy, with limited reporting on effectiveness in naturalistic settings or the routine integration of these tools into wider curricula [3, 4]. This service evaluation examined engagement and consultation performance during the implementation of an AI-simulated virtual patient platform within an undergraduate medical programme. Methods A service evaluation was undertaken during an 8-week open-access pilot of an AI-simulated consultation platform (MedAscend) at a UK medical school (March–May 2025). Platform usage and performance data were captured automatically for all completed consultations. Descriptive analyses summarised uptake, engagement, and consultation performance by attempt number. Within-student change in performance was described using first-versus-last consultation score differences, with a sensitivity analysis comparing each student’s median score across their first two and last two consultations among those completing at least four consultations. Analyses were descriptive and intended to characterise feasibility and usage patterns rather than determine effectiveness. Results A total of 292 students completed 2,215 AI-simulated consultations across 24 scenarios. The median number of consultations per student was 6.5 (IQR 3.0–12.0), with substantial variation in individual engagement. Median consultation performance increased from 70.0% (IQR 61.9–76.7) at attempt 1 to 77.4% (71.2–84.1) at attempt 2 and reached 84.1% (75.8–88.9) by attempt 6, before fluctuating modestly thereafter. The majority of students demonstrated higher consultation scores at their final consultation compared with their first. Higher engagement bands showed larger median changes in performance, although variability was observed across all engagement levels. Findings were consistent in sensitivity analyses using a more conservative definition of performance change. Conclusions This service evaluation demonstrates the feasibility of deploying an AI-simulated consultation platform as an optional, self-directed educational resource within undergraduate medical education. The findings describe patterns of engagement and consultation performance during routine use and support further structured evaluation of AI-simulated consultation platforms alongside existing approaches to clinical communication training.

Ce résumé expose les affirmations des auteurs. BNTIC ne l’interprète pas comme une validation indépendante des résultats.