Temporal and spatial associations between health financing composition and crude death rate in mainland China, 2012–2023: a provincial ecological panel study

The declining share of out-of-pocket health expenditure (OOP) has changed the relative contributions of social health expenditure (SHE) and government health expenditure (GHE). Evidence on health financing composition and crude death rate (CDR) in a province-year framework remains limited. We described their distributions and examined temporal and spatial associations in mainland China from 2012 to 2023. We analyzed a balanced panel of 31 provinces (372 province-year observations). Health financing composition was represented by the SHE, GHE, and OOP shares and the SHE/OOP and GHE/OOP ratios. Maps and regional summaries described distributions; Moran scatterplots and annual Global Moran’s I characterized spatial structure. Share-based and ratio-based two-way fixed-effects (TWFE) models estimated conditional within-province temporal associations with ln(CDR). A spatial Durbin model (SDM) using a row-standardized Queen matrix estimated within-province, cross-province, and total spatial associations. CDR increased in most provinces and rose most in Northeast China, from 5.98 to 9.57 deaths per 1,000 population (59.98%). Across all four regions, the SHE share increased and the OOP share decreased, while changes in the GHE share varied. Annual Global Moran’s I for ln(CDR) increased from 0.098 ( p = 0.133) in 2012 to 0.336 ( p = 0.001) in 2023; Moran’s I values for both log ratios were positive throughout. In TWFE models, the SHE- and GHE-share estimates relative to OOP were positive and attenuated after adjustment. Estimates for ln(SHE/OOP) were negative and non-significant and changed direction in sensitivity analyses. Estimates for ln(GHE/OOP) were positive but attenuated after full adjustment. In the SDM, ln(SHE/OOP) had a small, non-significant within-province spatial association with ln(CDR) (0.0058) and negative cross-province (− 0.1544) and total (− 0.1486) associations. Corresponding associations for ln(GHE/OOP) were positive (0.0833, 0.1609, and 0.2442; all p < 0.01). Health financing composition and CDR varied across provinces and regions and showed distinct spatial structures. Conditional within-province temporal associations between health financing composition and ln(CDR) varied by indicator form, covariate adjustment, and study period. The negative total spatial association of ln(SHE/OOP) was accounted for mainly by the cross-province component, whereas ln(GHE/OOP) showed positive within-province, cross-province, and total spatial associations. These provincial ecological findings may inform joint monitoring and help identify provinces and regions warranting further assessment.