Machine learning predictions of the adverse events of different treatments in patients with ischemic left ventricular...

Area of Science:

• Cardiology
• Machine Learning
• Medical Informatics

Background:

• Ischemic left ventricular systolic dysfunction (LVSD) poses a significant risk for major adverse cardiac events (MACE).
• Predicting MACE is crucial for guiding treatment strategies, including percutaneous coronary intervention (PCI) and optimal medical therapy (OMT).
• Existing predictive models may not fully capture the complexity of MACE risk in LVSD patients.

Purpose of the Study:

• To develop and validate novel machine learning models for predicting MACE in LVSD patients.
• To identify key predictors of MACE using statistical and machine learning techniques.
• To compare the performance of different machine learning models in predicting MACE.

Main Methods:

• Retrospective analysis of 329 LVSD patients, randomly assigned to training and validation cohorts.
• Least Absolute Shrinkage and Selection Operator (LASSO) regression to identify MACE-associated variables.
• Development of multiple machine learning models including Random Survival Forest (RSF), Gradient Boosting Machine (GBM), XGBoost, Cox, and DeepSurv.
• Model performance evaluation using concordance index (C-index), Area Under the Receiver Operating Characteristic Curve (AUC), and Brier Score (BS).

Main Results:

• LASSO regression identified five key factors for MACE prediction.
• The RSF model demonstrated superior performance with high C-indices (0.888 in development, 0.955 in validation).
• RSF achieved the highest AUC (0.991) and low Brier Scores (<0.25 at 12, 18, 24 months), indicating excellent predictive accuracy and calibration.

Conclusions:

• A novel predictive model based on RSF effectively predicts MACE in LVSD patients.
• The RSF model offers a robust tool for risk stratification in patients receiving PCI or OMT.
• This study highlights the potential of machine learning in improving cardiovascular event prediction.