Performance of ChatGPT and Gemini Compared with Emergency Physicians in NSTEMI Cases: A Prospective Cross-sectional Study.

AI models significantly outperformed emergency physicians on NSTEMI clinical questions based on 2023 ESC guidelines.

• Both ChatGPT 4.0 and Gemini 2.5 correctly answered nine of ten questions.
• Emergency physicians achieved a mean score of 7.62±1.32 correct answers out of ten.
• The difference was statistically significant (P<0.001).
• AI models were queried using identical standardized prompts with temperature=0 and no web access on April 20, 2025.

Effect sizes between AI and physician performance varied by physician experience level.

• Effect sizes indicated a 'very large difference' for less experienced physicians compared to AI models.
• Effect sizes indicated a 'moderate difference' for specialists compared to AI models.
• AI performance exceeded even the most experienced physicians.

Physician performance on NSTEMI questions improved with clinical experience.

• Performance increased as experience level increased, though AI still exceeded even the most experienced group.
• The study included 1,106 emergency physicians in Turkey surveyed via an online survey.
• Participants from training and research hospitals scored higher than those from state hospitals.

The study was conducted as a prospective, cross-sectional online survey among emergency physicians in Turkey.

• Survey included ten NSTEMI-related multiple-choice questions based on the 2023 European Society of Cardiology guidelines.
• A total of 1,106 emergency physicians participated.
• The same ten questions were presented to both ChatGPT 4.0 and Gemini 2.5.
• Statistical analyses were performed using SPSS 26.0.
• The study acknowledges limitations including a non-proctored online setting and absence of real clinical context.

The authors highlight AI's potential to enhance medical education, clinical decision support, and patient care in emergency settings.

• ChatGPT and Gemini demonstrated superior performance over emergency physicians in NSTEMI clinical questions.
• The authors call for future research focusing on optimizing AI-clinician collaboration for safe and effective integration.
• The findings are described as limited by the non-proctored online setting and absence of real clinical context.