Virtual patients and standardized patients combined training is associated with improved clinical reasoning among...

To develop an artificial intelligence (AI)-driven virtual standardized patients (VSPs) system, and to evaluate its educational effectiveness when combined with traditional standardized patients (SPs) training.

Leveraging natural language processing and a Chinese large language model, we built an AI-powered VSPs application. A total of 80 medical students at Peking University Third Hospital were randomized into two groups: experimental (n = 40, VSPs and SPs combined training) and control (n = 40, traditional SPs training). The 4-week intervention included assessments on clinical reasoning, core competencies, OSCE performance, and learning experience.

The experimental group showed greater improvement than the control group in clinical reasoning scores (1.3 ± 0.7 vs. 0.3 ± 0.5, 95% CI: 0.73-1.27, P < 0.01), core competencies total scores (9.9 ± 4.1 vs. 3.8 ± 2.7, 95% CI: 4.56-7.64, P < 0.01), and OSCE performance (11.9 ± 5.2 vs. 3.2 ± 5.0, 95% CI: 6.43-10.97, P < 0.01). The improvements were observed in communication, patient care, and internal medicine knowledge. Learner-reported satisfaction was higher (4.6 ± 0.5 vs. 3.2 ± 0.8, 95% CI: 1.10-1.70, P < 0.01). Subgroup analysis revealed greater benefits for students with lower baseline scores.

Combined VSPs and SPs training showed better performance than the SPs-only approach used in this study.