Supplementary dataset for comparative evaluation of ChatGPT and Gemini in explaining social determinants of hypertension

This dataset contains supplementary materials for the manuscript entitled “Comparative evaluation of large language models in explaining social determinants of hypertension: implications for public health communication.” The files include the anonymized evaluation matrix, structured prompt catalog, primary text outputs generated by ChatGPT and Gemini, reviewer scoring materials, and statistical analysis outputs used to compare model performance across seven evaluation criteria: content relevance, clarity and structure, comprehensiveness, audience appropriateness, scientific language use, internal coherence, and reproducibility. The study evaluated 30 structured questions covering six social determinant domains: socioeconomic factors, education and health literacy, healthcare access and quality, lifestyle and behavioural determinants, environmental determinants, and public health policy/systemic factors. Responses were assessed by three expert reviewers using a standardized rubric, and paired comparisons between ChatGPT and Gemini were conducted using the Wilcoxon signed-rank test. These files are provided to support transparency, reproducibility, and verification of the manuscript’s statistical and methodological reporting.