ChatGPT for Patients: A Comprehensive Study on Atrial Fibrillation Awareness.

Abstract

Due to the intricate nature of atrial fibrillation (AF), the diagnostic process often gives rise to a spectrum of concerns and inquiries. A 20-question survey on AF, covering general concerns, diagnosis, treatment, and post-diagnosis inquiries, was conducted via Google Forms (Google LLC, Mountain View, CA, USA). The questions were input into the Chat Generative Pre-trained Transformer (ChatGPT) system (OpenAI LP, San Francisco, CA, USA) in November 2023, and the responses were meticulously collated within the same Google Forms. The survey, involving 30 experienced physicians, including 22 cardiologists and 8 hospitalists, practicing for an average of 18 years, assessed artificial intelligence (AI)-generated responses to 20 medical queries. Out of 600 evaluations, "excellent" responses were most common (29.50%), followed by "very good" (26%), "good" (19.50%), and "fair" (17.3%). The least common response was "poor" (7.67%). Questions were categorized into "general concerns," "diagnosis-related," "treatment-related," and "post-diagnosis general questions." Across all categories, >50% of experts rated responses as "excellent" or "very good," indicating the potential for improvement in the AI's clinical response methodology. This study highlights the efficacy of ChatGPT as an AF informational resource, with expert-rated responses comparable to those of clinicians. While proficient, concerns include infrequent updates and ethical considerations. Nevertheless, it underscores the growing role of AI in health care information access.