Development of heart failure risk prediction models based on a multi-marker approach using random forest algorithms.

Area of Science:

• Cardiology
• Biomarker Discovery
• Predictive Analytics

Background:

• Early identification of heart failure (HF) risk is vital for patient outcomes.
• Combining multiple biomarkers offers a more comprehensive approach to HF risk stratification.
• Individual cardiac biomarkers include creatine kinase MB isoenzyme (CK-MB), B-type natriuretic peptide (BNP), galectin-3 (Gal-3), and soluble suppression of tumorigenicity-2 (sST2).

Purpose of the Study:

• To assess the diagnostic importance of individual cardiac biomarkers (CK-MB, BNP, Gal-3, sST2) for HF.
• To evaluate the predictive performance of a multi-marker model for HF using random forest algorithms.

Main Methods:

• 193 participants (80 HF patients, 113 controls) were analyzed.
• Correlation and regression analyses were performed between biomarkers and echocardiographic parameters.
• Random forest algorithms were used to assess predictor accuracy and importance.

Main Results:

• HF patients showed significantly higher levels of all four biomarkers.
• BNP demonstrated strong independent predictive capacity for HF (AUC 0.956).
• A multi-marker model combining CK-MB, BNP, Gal-3, and sST2 improved predictive performance, achieving 91.5% sensitivity and 96.7% specificity.

Conclusions:

• The random forest algorithm is a robust tool for assessing biomarker accuracy and importance.
• A combination of CK-MB, BNP, Gal-3, and sST2 significantly enhances prediction accuracy for heart failure.