Machine learning approaches were used to identify predictors of complication risk in T2DM. A study in Diabetes, Metabolic Syndrome and Obesity evaluated ML-based prediction models to support community management of diabetes and related complications.

The retrospective analysis included 4,916 adults with T2DM from community settings in Shanghai. Prediction Model I was developed using least absolute shrinkage and selection operator (Lasso) regression, support vector machine (SVM), decision tree (DT), and logistic regression (LR). A Bayesian Network (BN) model was constructed to explore potential causal relationships. Model performance was evaluated with receiver operating characteristic (ROC) curves, AUC, calibration curves, and decision curve analysis (DCA).

Five consistent predictors emerged: disease course, diastolic blood pressure, HbA1c, urinary creatinine, and urinary microalbumin. Model I achieved AUC values of 0.695 in the training set and 0.676 in the validation set. Risk thresholds on DCA ranged from 12% to 92% and from 20% to 92%, respectively. The tree-augmented BN model achieved an AUC of 0.755, accuracy of 0.733, specificity of 0.802, and sensitivity of 0.519.

These findings identify individuals with longer disease duration, chronic comorbidities, and higher urinary microalbumin as important targets for community management. Strengthening health education, dietary self-management, and family and community support may improve complication prevention in T2DM.