Regional variation in public perceptions of China’s resident medical insurance: evidence from 340,627 Douyin comments
Background Social media discourse can provide complementary evidence on how health insurance policies are publicly evaluated. This study examined provincial variation in Douyin discussions framed around “Xin Nong He” and explored their associations with provincial socio-economic conditions. Methods We analyzed 340,627 comments posted during 2024 under 156 highly discussed Douyin videos. Comments were classified into five substantive dimensions: Economic Affordability and Cost Burden, Benefit Effectiveness and Quality of Service, Governance Quality and Implementation, Institutional Equity and Structural Design, and Institutional Trust and Behavioral Intent. A simple random validation sample ( $$N=500$$ ) yielded an accuracy of 81.2%, Cohen’s $$\kappa =0.760$$ , and a weighted F1-score of 0.822. Provincial Perception Intensity Indices were standardized by mobile internet access population. Spatial and exploratory province-level regression analyses ( $$n=31$$ ) were supplemented by influence diagnostics, leave-one-province-out estimation, province-level bootstrap analysis, and SHAP decomposition. Results Institutional Trust and Behavioral Intent constituted the largest model-assigned category (34.71%). Affordability-related discourse intensity was comparatively high in Guangxi and Guizhou, whereas equity-related discourse was comparatively high in Zhejiang and Jiangsu. Gender Ratio was the only association that met the FDR-adjusted threshold ( $$\beta =0.522$$ , 95% CI [0.240, 0.804], $$q=0.015$$ ). The coefficient remained positive in all 31 leave-one-province-out models and in 99.8% of 1,000 bootstrap replications, although FDR-adjusted significance was retained in 28 leave-one-province-out models. SHAP decomposed the contribution of Gender Ratio to fitted OLS predictions rather than estimating province-specific effects. Conclusions Xin-Nong-He-related discourse on Douyin exhibited multidimensional and geographically differentiated patterns. Social media data may complement conventional health-policy monitoring, but the findings should not be interpreted as population-representative public opinion. The Gender Ratio association was directionally consistent but remains an exploratory ecological finding.
Authors
- Pan Hui (ORCID: https://orcid.org/0000-0001-6026-1083)
- Pu Zhang (ORCID: https://orcid.org/0009-0000-3360-9937)
- Na Jiang
Institutions
- University of Helsinki (FI)
- Hong Kong University of Science and Technology (HK)
- The Hong Kong University of Science and Technology (Guangzhou) (CN)
Publication Details
- Journal
- BMC Public Health
- Published
- 2026-11-21
- DOI
- https://doi.org/10.1186/s12889-026-30312-0
- Primary Topic
- Healthcare Systems and Reforms
- Type
- article
- Field-Weighted Citation Impact
- 0.00