This randomized controlled trial evaluated a large language model-based Intelligent Simulated Patient System (ISPS) for gastroenterology history-taking education. The system integrates retrieval-augmented generation (RAG), electronic health record-based knowledge retrieval, virtual patient interaction, and automated scoring. Ninety medical students were randomly assigned to conventional training or ISPS-assisted training. Primary outcomes included history-taking completeness and diagnostic accuracy. Secondary outcomes included empathy performance and technology acceptance.
Inclusion Criteria:
Exclusion Criteria:
Participants and instructors were aware of group assignment because of the nature of the educational intervention. Outcome assessors who evaluated the OSCE performance were blinded to group allocation. Statistical analyses were also performed by an independent statistician blinded to treatment assignment.
Participants received four weeks of training using the Intelligent Simulated Patient System (ISPS), a large language model-based virtual patient platform for gastroenterology education. The system provides repeated history-taking practice, automated assessment, and individualized feedback.
Participants received four weeks of conventional clinical education, including bedside teaching, case-based learning, paper-based medical record review, and faculty-guided clinical discussions according to the standard curriculum.
Hangzhou, Zhejiang, China
Simulated Patient and AI-based Roleplay for History-taking
Simulation-assisted Teaching in Learning Gastroscopy
The Effects of a Large Language Model on Clinical Questioning Skills
AI-generated Feedback in Social Robotic Virtual Patients
Medical Students' Acceptance and Learning Efficacy With Physical and Virtual Standardized Patients
AI Virtual Patient Training for Dental Student History Taking
ChatGPT Helping Advance Training for Medical Students: A Study on Self-Directed Learning Enhancement
Virtual Patient Education From Real Cases