Predicting low density lipoprotein cholesterol target attainment using machine learning in patients with coronary...

Area of Science:

• Cardiology
• Medical Informatics
• Machine Learning in Healthcare

Background:

• Low-density lipoprotein cholesterol (LDL-C) is a critical factor in cardiovascular disease (CVD) development.
• Effective LDL-C management is essential for cardiovascular health, often involving statin therapy.
• Patient-specific factors and side effects necessitate careful consideration in high-dose statin recommendations.

Purpose of the Study:

• To develop and evaluate machine learning (ML) models for predicting target LDL-C achievement in coronary artery disease (CAD) patients on moderate-dose statins.
• To assess the predictive performance of Extreme Gradient Boosting (XGBoost), Random Forest, and Logistic Regression models.
• To identify key patient characteristics influencing LDL-C target achievement through SHAP analysis.

Main Methods:

• Utilized electronic medical records from Asan Medical Center, Seoul.
• Developed and compared three ML models: XGBoost, Random Forest, and Logistic Regression.
• Evaluated model performance using six metrics and performed feature reduction and SHAP analysis.

Main Results:

• All three ML models demonstrated predictive performance with an average AUROC of 0.695.
• Significant feature reduction (over 43%) was achieved without compromising predictive ability.
• SHAP analysis identified key patient features associated with achieving LDL-C targets.

Conclusions:

• ML models can effectively estimate the likelihood of achieving target LDL-C levels with moderate-dose statins.
• These models offer potential for personalized treatment strategies in cardiovascular disease management.
• ML-based insights can support clinical decision-making for optimizing LDL-C reduction.