A Pilot Test of an AI Voice-Driven Simulation With Feedback for Medical Students to Practice Discussing Diagnostic...

Introduction Effective communication when delivering sensitive results or bad news to patients is necessary for medical professionals, warranting greater attention in medical education. The opportunity to practice communication skills with human standardized patients has limitations, including not being readily available, prompting researchers to develop virtual patients (VPs) to supplement training with a focus on routine tasks. This study capitalized on artificial intelligence (AI) advances to develop a VP for more complex dialogs where learners practice discussing uncertain screening results with patients. Methods We conducted a single-arm pilot study with 10 medical students to evaluate the feasibility and acceptability of a simulated telephone call between a physician (learner) and a VP to discuss the patient's abnormal mammogram results that require breast biopsy examination. Objectives culminate in the VP agreeing to schedule a biopsy. Afterward, an AI agent displays feedback for the learner. Results Study completion demonstrated feasibility, with quantitative measures indicating acceptability. The average System Usability Scale score of 91 was high (range, 78-100). For Technology Acceptance Model questions, most students "agreed" or "strongly agreed" that VP responses "seemed reasonable", AI agent-generated feedback was "helpful", and VP training would increase their "confidence communicating with patients". Qualitative feedback solidified these findings, highlighting the simulation's realism, detailed feedback, and perceived effectiveness in improving communication skills. Conclusion The study recommends further evaluation for including VPs in medical education, especially to help learners practice difficult conversations with patients. Future developments will expand the scenario and evaluate communication skills outcomes while continuing to refine the VP and feedback.