Development and Interpretable Machine Learning-Based Prediction of Cardiovascular Disease Risk in Chinese COPD...

Area of Science:

• Cardiology
• Pulmonology
• Data Science

Background:

• Cardiovascular disease (CVD) significantly impacts the quality of life for individuals with chronic obstructive pulmonary disease (COPD).
• A predictive framework is needed to evaluate CVD risk in COPD patients.

Purpose of the Study:

• To develop and compare machine learning (ML) models for predicting CVD risk in COPD patients.
• To identify key predictors associated with CVD in this population.

Main Methods:

• Analysis of data from 1070 COPD patients (China Health and Retirement Longitudinal Study, 2015).
• Feature selection using LASSO regression and Boruta algorithm.
• Comparison of six ML models (Logistic Regression, Random Forest, SVM, Gradient Boosting, XGBoost, MLP) with SMOTE-NC for class imbalance.
• Risk assessment tool development using SHAP for interpretability.

Main Results:

• 305 participants (28.50%) had CVD.
• The Support Vector Machine (SVM) model demonstrated superior performance with a training AUROC of 0.819 and test AUROC of 0.719.
• Seven key predictors identified: sex, body weight, hypertension, dyslipidemia, disability, self-rated health, and vision status.

Conclusions:

• Seven variables significantly predict cardiovascular risk in Chinese COPD patients.
• The SVM model shows promise but requires further validation due to moderate predictive capacity and data limitations.
• Future prospective studies and external validation are crucial for clinical decision-support systems.