Answers are grounded in this paper and cite their sources.
Clinical history-taking and documentation are foundational competencies in undergraduate medical training, yet scalable and individualized training tools remain limited. Large language model (LLM)-based systems provide automated practice opportunities but often lack structured medical reasoning, closed-loop instructional design, and integration into formal curricula. As part of a curriculum reform initiative, we constructed a clinical case repository, developed an artificial intelligence (AI)-integrated faculty Plan–Do–Check–Act (PDCA) training framework supported by dual chain-of-thought (CoT)-driven LLMs, and evaluated the reliability of AI-based scoring against faculty assessment. We then conducted a retrospective observational analysis of routinely collected educational performance data from 135 undergraduate medical students enrolled in one of three existing instructional configurations: traditional instruction, AI-assisted training, or AI + PDCA training. Educational outcomes were assessed using standardized Objective Structured Clinical Examination (OSCE) assessments. The AI + PDCA configuration was associated with the highest OSCE performance in both history-taking and clinical documentation. Among the two AI-supported configurations, longitudinal mixed-effects modeling showed steeper learning trajectories in the AI + PDCA configuration than in the AI-assisted configuration. These findings suggest that integrating AI-derived learning analytics with faculty-guided instructional intervention may support a closed-loop educational model associated with favorable competency acquisition while maintaining human oversight. This framework offers a scalable approach for implementing trustworthy, competency-based AI-enhanced medical education in routine academic practice.
No takes yet. Share an insight, caveat, or question.
Chen et al. (2026) studied this question.
Synapse has enriched 10 related papers. Consider them for comparative context: