A Clinical Evaluation of Cardiovascular Emergencies: A Comparison of Responses from ChatGPT, Emergency Physicians, and Cardiologists

Abstract

<b>Background:</b> Artificial intelligence (AI) tools, like ChatGPT, are gaining attention for their potential in supporting clinical decisions. This study evaluates the performance of ChatGPT-4o in acute cardiological cases compared to cardiologists and emergency physicians. <b>Methods:</b> Twenty acute cardiological scenarios were used to compare the responses of ChatGPT-4o, cardiologists, and emergency physicians in terms of accuracy, completeness, and response time. Statistical analyses included the Kruskal-Wallis H test and post hoc comparisons using the Mann-Whitney U test with Bonferroni correction. <b>Results:</b> ChatGPT-4o and cardiologists both achieved 100% correct response rates, while emergency physicians showed lower accuracy. ChatGPT-4o provided the fastest responses and obtained the highest accuracy and completeness scores. Statistically significant differences were found between ChatGPT-4o and emergency physicians (<i>p</i> < 0.001), and between cardiologists and emergency physicians (<i>p</i> < 0.001). A Cohen's kappa value of 0.92 indicated a high level of inter-rater agreement. <b>Conclusions:</b> ChatGPT-4o outperformed human clinicians in accuracy, completeness, and response time, highlighting its potential as a clinical decision support tool. However, human oversight remains essential to ensure safe AI integration in healthcare settings.

Ähnliche Arbeiten