The prognostic value of stress hyperglycemia ratio (SHR) in cardiac metabolic syndrome: dual-cohort evidence and machine learning modeling

Objective: ​​ This study aimed to evaluate the prognostic value of the stress hyperglycemia ratio (SHR) in predicting all-cause and cardiovascular disease (CVD) mortality among patients with cardiometabolic syndrome (CMS), leveraging large-scale cohort data and machine learning models. ​​ Methods:​​ This study analyzed data from the US NHANES (2001–2018; n = 7,750) and Chinese CHARLS (2011; n = 4,279) cohorts. SHR was calculated as fasting plasma glucose/(1.59×HbA1c−2.59). Multivariate Cox regression, restricted cubic spline (RCS), and threshold effect analyses were employed to assess nonlinear associations. Eleven machine learning algorithms (e.g., XGBoost, LightGBM) were developed and validated to predict mortality risk, with SHAP values interpreting model outputs. ​​ Results:​​ Elevated SHR (tertile Q3) independently increased risks of all-cause mortality (HR: 1.38, 95% CI: 1.18–1.63) and CVD mortality (HR: 1.49, 95% CI: 1.14–1.94). RCS revealed U-shaped relationships, with inflection points at SHR = 0.87 (all-cause mortality) and 0.86 (CVD mortality). XGBoost achieved the highest predictive performance (AUC = 0.827), with SHR ranked as the third most influential feature after age and race. Subgroup analyses highlighted stronger associations in younger populations and gender-specific effects in older adults. ​​ Conclusion:​​ SHR is a robust, nonlinear predictor of mortality in CMS patients, underscoring its utility in risk stratification. Machine learning models integrating SHR enhance prognostic accuracy, supporting personalized clinical decision-making.