Decoupling analysis of medical service provision and population health outcomes: a hybrid evaluation and machine learning approach

Introduction Global healthcare faces a critical challenge where extensive medical resource expansion fails to translate into proportional population health outcomes due to diminishing marginal returns. The aim of this study was to explore the dynamic evolution, regional disparities, and non-linear associations between medical service provision (MSP) and population health outcomes (PHO) across Chinese provinces. Methods Based on panel data from 31 Chinese provinces (2012–2024), an integrated CRITIC-Entropy-TOPSIS approach was constructed to measure MSP and PHO levels. The Tapio model and Sankey diagrams characterized decoupling states and transitions, and XGBoost-SHAP framework quantified feature contributions and non-linear associations among the contextual variables. Results Nationally, the median MSP score rose from 0.295 to 0.373 with the spatial gap between eastern and western provinces expanded, but the median PHO score decreased from 0.739 to 0.696. The Tapio analysis revealed widespread non-coordination, with 27 provinces locked in negative decoupling (20 strong negative decoupling, 7 expansive negative decoupling). SHAP analysis identified Population Density (PD, mean absolute SHAP = 0.0174), Urbanization Rate (UR, 0.0126), and Government Health Expenditure (GHE, 0.0119) as the three most important predictors of the decoupling index. The dependence and interaction plots showed non-linear relationships within the provincial sample. Higher PD and UR generally corresponded to higher predicted decoupling values, but higher GHE and IPC corresponded to lower predicted values over some observed ranges. Discussion MSP–PHO decoupling was common across Chinese provinces. Increases in medical resources did not always coincide with proportional improvements in population health outcomes. Policymakers should improve the conversion of medical resources into health outcomes through fiscal support based on local needs, stronger healthcare cooperation across regions, and coordinated policies on demographic, socioeconomic and environmental conditions.