Assessing the Capabilities of Artificial Intelligence (AI) Tools in Community Medicine: A Comparative Study of...

Background and objective Artificial intelligence (AI) is being increasingly integrated into healthcare, offering opportunities to enhance decision-making, education, and patient engagement. Community medicine often involves interpreting clinico-social case studies that combine clinical, social, and environmental dimensions. However, there is limited research evaluating how AI tools perform in analyzing such community-based clinico-social cases. This study aimed to address that gap in the literature. Methods A comparative cross-sectional study was conducted using 30 standardized clinico-social case studies covering communicable and non-communicable diseases, maternal and child health, adolescent health, and social pathology. Three conversational AI models -ChatGPT (OpenAI, San Francisco, CA), Microsoft Bing AI (Microsoft Corp., Redmond, WA), Gemini (Google, Mountain View, CA) - were provided with identical prompts to interpret cases. Their responses were assessed by a panel of five community medicine experts using a 100-point rubric across five domains: diagnosis, intervention, recognition of social determinants, ethical reasoning, and public health appropriateness. Descriptive statistics and Spearman's rank correlation were used for analysis. Results All three AI tools demonstrated near-ceiling performance. Gemini achieved the highest total score (97.00 ± 1.74), excelling in diagnosis and public health appropriateness. ChatGPT (96.00 ± 2.98) performed best in intervention suggestions, while Bing AI (95.70 ± 3.64) showed slightly lower but comparable scores. Correlation analysis revealed weak-to-moderate alignment, with limited statistically significant associations across domains. Conclusions ChatGPT, Gemini, and Bing exhibit broadly similar capabilities in interpreting clinico-social cases, with domain-specific strengths. Their complementary nature suggests that they can aid medical education and public health practice, but should supplement rather than replace human expertise to minimize bias and errors.