Comparative Evaluation of Chatbot Responses on Coronary Artery Disease

OBJECTIVE
Coronary artery disease (CAD) is the leading cause of morbidity and mortality globally. The growing interest in natural language processing chatbots (NLPCs) has driven their inevitable widespread adoption in healthcare. The purpose of this study was to evaluate the accuracy and reproducibility of responses provided by NLPCs, such as ChatGPT, Gemini, and Bing, to frequently asked questions about CAD.


METHODS
Fifty frequently asked questions about CAD were asked twice, with a one-week interval, on ChatGPT, Gemini, and Bing. Two cardiologists independently scored the answers into four categories: comprehensive/correct (1), incomplete/partially correct (2), a mix of accurate and inaccurate/misleading (3), and completely inaccurate/irrelevant (4). The accuracy and reproducibility of each NLPC’s responses were assessed.


RESULTS
ChatGPT’s responses were scored as 14% incomplete/partially correct and 86% comprehensive/correct. In contrast, Gemini provided 68% comprehensive/correct responses, 30% incomplete/partially correct responses, and 2% a mix of accurate and inaccurate/misleading information. Bing delivered 60% comprehensive/correct responses, 26% incomplete/partially correct responses, and 8% a mix of accurate and inaccurate/misleading information. Reproducibility scores were 88% for ChatGPT, 84% for Gemini, and 70% for Bing.


CONCLUSION
ChatGPT demonstrates significant potential to improve patient education about coronary artery disease by providing more sensitive and accurate answers compared to Bing and Gemini.