Development and Validation of a Machine Learning-Based Predictive Model for Peripheral Neuropathy Risk in Elderly...

Area of Science:

• Medical Informatics
• Machine Learning in Healthcare
• Diabetology

Background:

• Diabetic peripheral neuropathy (DPN) is common in elderly type 2 diabetes patients.
• Current DPN identification methods lack performance and specificity.
• There is a need for improved early detection models.

Purpose of the Study:

• To develop and validate a machine learning (ML) model for early DPN risk identification in elderly type 2 diabetes patients.
• To enhance the specificity and performance of DPN prediction.
• To provide a tool for timely clinical intervention.

Main Methods:

• Retrospective data collection from 1450 elderly type 2 diabetes patients.
• Feature selection and preprocessing followed by model construction using logistic regression, naïve Bayes, random forest, and XGBoost.
• Model performance evaluation using AUC, accuracy, precision, recall, F1-score, calibration curves, DCA, and SHAP analysis.

Main Results:

• DPN prevalence was 42.9%.
• Nine independent predictors identified: diabetes duration, HbA1c, sleep quality, Charlson Comorbidity Index, sugar-sweetened beverage intake, peripheral arterial disease, sedentary behavior, smoking, and hypertension.
• XGBoost model achieved the highest performance (AUC: 0.951, accuracy: 0.878), with diabetes duration and HbA1c as key predictors.

Conclusions:

• The XGBoost-based model demonstrates strong predictive performance and clinical utility for DPN risk in elderly type 2 diabetes patients.
• The model facilitates early identification of high-risk individuals.
• This approach supports targeted clinical interventions for DPN management.