Machine learning algorithms for predicting mortality after coronary artery bypass grafting.

Area of Science:

• Cardiovascular Surgery
• Medical Informatics
• Machine Learning in Healthcare

Background:

• Big data analytics is increasingly utilized in healthcare.
• Machine learning (ML) offers potential for predicting clinical outcomes.
• Coronary artery bypass grafting (CABG) outcomes require accurate prediction models.

Purpose of the Study:

• To evaluate the predictive performance of various ML models for mortality after CABG.
• To identify key predictors of mortality in CABG patients.
• To compare the efficacy of different ML algorithms in predicting CABG mortality.

Main Methods:

• Utilized a CABG data registry with baseline and follow-up features.
• Selected key variables using the random forest method.
• Developed and assessed prediction models using Logistic Regression (LR), Support Vector Machine (SVM), Naïve Bayes (NB), K-Nearest Neighbors (KNN), Extreme Gradient Boosting (XGBoost), and Random Forest (RF) algorithms, evaluating performance with Area Under the Curve (AUC).

Main Results:

• Included 16,850 patients undergoing isolated CABG; 468 deaths occurred within one year.
• Total ventilation hours and left ventricular ejection fraction were significant predictors of mortality.
• All ML models demonstrated acceptable performance (AUC > 0.7) for one-year mortality prediction.
• Logistic Regression (LR) achieved the highest AUC (0.811), followed closely by XGBoost (0.792).
• LR also showed the highest predictive ability for two-to-five-year mortality.

Conclusions:

• Multiple machine learning models exhibit acceptable performance in predicting CABG-related mortality.
• Logistic Regression (LR) demonstrated superior predictive capability compared to other evaluated ML models.
• These ML models can assist clinicians in risk stratification and decision-making for patients undergoing CABG.