Development and Validation of an XGBoost-Based Machine Learning Model With Nomogram for Predicting Diabetic...

Area of Science:

• Computational medicine and bioinformatics
• Endocrinology and metabolic diseases
• Machine learning applications in healthcare

Background:

• Diabetic peripheral neuropathy (DPN) is a common complication of type 2 diabetes mellitus (T2DM).
• Accurate risk stratification is crucial for early intervention and management of DPN.
• Existing methods may not fully capture the complex interplay of risk factors for DPN.

Purpose of the Study:

• To develop and validate machine learning models for DPN risk stratification in T2DM patients.
• To identify the optimal predictive model and key risk factors for DPN.
• To create a visual nomogram as a clinical decision-support tool for early DPN risk identification.

Main Methods:

• Retrospective analysis of 180 T2DM patients (88 with DPN, 92 without).
• Development of five predictive models: Logistic Regression (LR), Random Forest (RF), Extreme Gradient Boosting (XGBoost), Support Vector Machine (SVM), and Decision Tree (DT).
• Internal validation using ROC curves, calibration curves, DCA, and SHAP analysis; nomogram construction based on the best model.

Main Results:

• Six independent predictors for DPN identified: diabetes duration, HbA1c, MAU, LDL-C, NEUT%, and BMI.
• XGBoost model achieved the highest predictive performance (AUC=0.903) in internal validation, outperforming other models.
• SHAP analysis highlighted diabetes duration, HbA1c, and MAU as most influential; nomogram showed good calibration and clinical utility (AUC=0.85).

Conclusions:

• The XGBoost-based model demonstrates strong preliminary performance for DPN risk stratification in T2DM.
• The developed nomogram, incorporating SHAP interpretability, serves as an exploratory clinical tool for early risk identification.
• Further external validation is necessary prior to widespread clinical application.