REFINeAI-Supported Diagnostic Accuracy Improvement in Nephrology

This study is all about improving the diagnosis of kidney-related diseases, also known as nephrology cases. It explores whether providing doctors with real-time diagnostic suggestions from a high-performing AI model, specifically GPT-5, can enhance their diagnostic accuracy, confidence, and efficiency. The study is important because it could potentially revolutionize the way nephrology diagnoses are made, leading to improved patient care and addressing current challenges in this field. The study involves medical students, residents, fellows, and practicing physicians. Participants are divided into two groups: one group receives AI support, while the other group does not. Participants in both groups are presented with 10 nephrology cases to solve, and they can complete these cases over multiple sessions. In the AI-supported group, participants first provide their initial diagnoses and confidence ratings without AI assistance. Then, they are shown GPT-5's diagnostic suggestions, after which they can revise their diagnoses once. In the control group, participants view each case and provide up to three diagnoses followed by a confidence rating. The study measures initial and final diagnoses, confidence ratings, completion times, and participant demographic variables. The primary outcome is the diagnostic accuracy with and without AI support.