Impact of Artificial Intelligence-Assisted Feedback on Clinical Reasoning Skills among Postgraduate Medical Trainees
Keywords:
Artificial Intelligence, Clinical Reasoning, Feedback, Postgraduate Medical Education, Medical Trainees, Formative Assessment.Abstract
Background: Clinical reasoning is an important competency in postgraduate medical education, but traditional feedback systems can be slow and intermittent due to faculty workload and the time available for supervision. In case-based learning, feedback can be immediate, individualized and structured with the support of systems supported by artificial intelligence.
Objective: To determine the impact of artificial intelligence-assisted feedback on clinical reasoning skills among postgraduate medical trainees.
Methods: The study was a prospective quasi-experimental comparative study carried out at Rawal Institute of Health Sciences Islamabad, from January 2025 to June 2025. In total, there were 76 postgraduate medical trainees, half of whom received feedback via artificial intelligence (38 trainees) and the other half received it from the faculty (38 trainees). Both groups answered clinical cases and had pre- and post-test clinical reasoning assessments. The main outcome was the difference from the overall clinical reasoning score. Secondary outcomes were diagnostic accuracy, generation of differential diagnosis, selection of investigations, planning for management, diagnostic errors, time to complete the case, confidence and satisfaction. Analysis of the data was done by suitable descriptive and inferential statistical tests, and the p-value of < 0.05 was taken as statistically significant.
Results: Baseline clinical reasoning scores were comparable between the AI-assisted and conventional-feedback groups (61.8 ± 7.4 versus 62.3 ± 7.1; p = 0.766). After the intervention, the mean score was significantly higher in the AI-assisted group than in the conventional group (78.6 ± 6.8 versus 70.4 ± 7.3; p < 0.001). The mean improvement was 16.8 ± 6.1 points in the AI group compared with 8.1 ± 5.7 points in the conventional group. The AI-assisted group also demonstrated greater diagnostic accuracy, fewer diagnostic errors, shorter case-completion time, and higher confidence and satisfaction scores.
Conclusion: Artificial intelligence-assisted feedback was associated with greater improvement in clinical reasoning performance than conventional feedback. It may be used as a supplementary educational strategy under appropriate faculty supervision.
