A Stacked Ensemble Multi-Label Model for Predicting Co-Occurring Microvascular and Macrovascular Complications in...

Area of Science:

• Medical Informatics
• Machine Learning in Healthcare
• Diabetes Research

Background:

• Type 2 diabetes mellitus (T2DM) causes severe microvascular and macrovascular complications.
• Existing single-label prediction models inadequately address the co-occurrence of these complications.
• There is a need for integrated approaches to jointly predict diabetes-related complications.

Purpose of the Study:

• To develop a stacked ensemble multi-label framework for the joint prediction of T2DM complications.
• To integrate machine learning techniques into a clinically interpretable model.
• To identify distinct risk factors for co-occurring microvascular and macrovascular complications.

Main Methods:

• A retrospective study included 965 T2DM patients.
• Complications were categorized into microvascular and macrovascular groups.
• A class-weighted stacking ensemble model (Stacking-CC) was developed using Random Forest, Light GBM, and Cat Boost, with a logistic regression meta-learner, evaluated via cross-validation and SHAP analysis.

Main Results:

• The Stacking-CC model demonstrated superior performance with an F1-score of 0.752 ± 0.049 and AUC of 0.857 ± 0.032.
• SHAP analysis revealed distinct risk profiles: macrovascular complications linked to LDL cholesterol and diastolic blood pressure; microvascular complications linked to drug addiction, fasting blood sugar, and HDL.
• Conventional factors like age and BMI showed minimal importance.

Conclusions:

• The Stacking-CC framework accurately models co-occurring diabetic complications with high interpretability.
• The model's ability to delineate distinct risk hierarchies supports targeted, complication-specific management strategies.
• This approach enhances clinical decision-making for T2DM patients.