Development and validation of a machine learning-based predictive model for new-onset atrial fibrillation after CABG

Area of Science:

• Cardiology
• Medical Informatics
• Machine Learning

Background:

• New-onset atrial fibrillation (NOAF) is a frequent complication following coronary artery bypass grafting (CABG).
• NOAF is linked to adverse post-operative outcomes in cardiac surgery patients.
• Predictive tools for NOAF after CABG are crucial for risk stratification.

Purpose of the Study:

• To develop and evaluate machine learning (ML) models for predicting NOAF post-CABG.
• To compare the performance of ML models against conventional risk scores.
• To identify key predictors of NOAF using ML interpretability techniques.

Main Methods:

• A cohort of 925 patients undergoing CABG was analyzed.
• Data were split into training (70%) and validation (30%) sets.
• Eight ML models were developed and evaluated using AUC, accuracy, sensitivity, specificity, F1 score, and Youden index. SHAP values were used for feature interpretation.

Main Results:

• The incidence of NOAF was 19%.
• The Gradient Boosting model achieved the highest AUC of 0.842, outperforming logistic regression (0.790).
• All developed ML models demonstrated superior performance compared to CHA2DS2-VASc and HATCH risk scores.

Conclusions:

• Machine learning models accurately predict NOAF after CABG.
• These models can aid in risk stratification and hypothesis generation for NOAF.
• External validation is necessary prior to clinical application of these predictive models.