
The study developed and validated machine learning (ML) models for self-identifying microvascular complication risks in type 1 diabetes (T1D). Using data from the Chinese Type 1 Diabetes Comprehensive Care Pathway program, 911 patients were analyzed, and models were built with the XGBoost algorithm using 15 self-reported variables. External validation included 157 participants from an online T1D community. The models achieved AUROC values of 0.889 (DR), 0.844 (DN), and 0.839 (DPN) in internal validation and 0.762, 0.718, and 0.721, respectively in external validation. Shapley Additive ExPlanation (SHAP) analysis provided insights into its importance. These models enable T1D patients to self-assess their risks for diabetic retinopathy, nephropathy, and peripheral neuropathy, encouraging timely healthcare engagement and improving management of complications outside clinical settings.
Like
Save
Share