Social determinants of multimorbidity diabetes- Cardiovascular disease

This study uses nationally representative data from the Health Survey for England to explore the role of Social Determinants Of Health (SDoH) in shaping the comorbidity of type 2 diabetes (T2D) and cardiovascular disease (CVD), and to compare the predictive performance and interpretability of machine learning models.Weighted datasets were analyzed using logistic regression, Xgboost, Multilayer Perceptron (MLP), and a stacking ensemble.Model performance was evaluated with ROC-AUC, PR-AUC, and F1 scores, alongside Brier score and reliability curves to assess calibration.To enhance transparency, interpretability techniques, including SHAP, permutation feature importance (PFI), and partial dependence plots (PDP), were applied.The results indicate that, in the full specification combining SDoH and core health variables, XGBoost achieved the highest discrimination on the test set (AUROC≈0.866;AUPRC≈0.108;F1≈0.145), with the stacked model performing similarly.Explanatory analysis consistently identified age, education, economic activity, housing, and income disparities as key drivers of comorbidity risk, while clinical covariates such as BMI, blood pressure, and HDL played a secondary role.These findings suggest that structural social inequalities are central to the development of T2D-CVD comorbidity and highlight the importance of incorporating SDoH into both predictive modeling and public health policy.The study demonstrates that explainable machine learning can provide robust and interpretable evidence to inform population-level screening strategies and guide more equitable allocation of health resources.