Related Experiment Videos
A Randomized Controlled Pilot Study Comparing ChatGPT to Live Coaching Deliberate Practice to Teach Difficult
Meron Amariw1, Mitchell L Phillips1, Clara S Miller1
1The following authors are at Northwestern University, Feinberg School of Medicine in Chicago, IL: and are medical students and is a Professor of Medicine and Medical Education and American Board of Internal Medicine Certified in the Department of Medicine and Medical Education. is an Assistant Professor of Anesthesiology with a Fellowship in Pediatric Anesthesiology and American Board of Anesthesiology (ABA) and Pediatric Anesthesiology Certified in the Department of Pediatric Anesthesiology, Nemours Children's Hospital, Orlando, FL. is a Professor of Anesthesiology with a Fellowship in Cardiothoracic Anesthesiology, ABA, and National Board of Echocardiography Certified in the Department of Anesthesiology, Henry Ford Health, Detroit, MI. is an Associate Professor of Anesthesiology with a Fellowship in Pediatric Anesthesiology and ABA and Pediatric Anesthesiology Certified in the Department of Pediatric Anesthesiology and Medical Education, Ann & Robert H. Lurie Children's Hospital of Chicago, IL.
Background:
Poor performance during difficult conversations is associated with decreased patient satisfaction and increased clinician burnout. Communication skills can be taught using simulated patients (SPs), but this approach is resource intensive. Artificial intelligence (AI) has emerged as a potential adjunct for communication training. However, its role within anesthesiology education and structured communication curricula remains unclear.
Methods:
We conducted a randomized controlled pilot study involving anesthesiology trainees, including anesthesiology residents and student registered nurse anesthetists, at a tertiary pediatric hospital. All participants completed a didactic session on the SPIKES framework, followed by deliberate practice using either OpenAI's ChatGPT or an SP with live instructor feedback. Participants completed pretraining and posttraining assessments using a validated 16-item communication skills checklist with an SP. Self-reported knowledge, confidence, and ability to apply communication skills were assessed using Likert-scale surveys.
Results:
Twenty-five trainees were enrolled (ChatGPT n = 13; SP n = 12). Both groups demonstrated significant improvements in checklist scores following training. Self-reported knowledge, confidence, and application of communication skills improved in both groups.
Conclusions:
A curriculum combining didactic instruction with deliberate practice using either ChatGPT or SPs was associated with improved simulated communication performance among anesthesiology trainees. This pilot study demonstrates the feasibility of incorporating AI-based tools into communication training.
