Development of machine learning models for mortality risk prediction after cardiac surgery

Area of Science:

• Cardiology
• Medical Informatics
• Machine Learning

Background:

• Cardiac surgery carries significant postoperative mortality risks.
• Accurate prediction of mortality is crucial for patient management and surgical planning.
• Existing risk scores have limitations in predicting outcomes.

Purpose of the Study:

• To develop and evaluate machine learning models for predicting mortality after cardiac surgery.
• To compare the performance of machine learning models against established clinical risk scores.
• To identify key predictors of postoperative mortality using explainable AI.

Main Methods:

• Development of machine learning models including random forest, neural network, support vector machine, and gradient boosting.
• Comparison with EuroSCORE I, EuroSCORE II, Society of Thoracic Surgeons (STS) risk scores, and logistic regression.
• Utilized Shapley's additive explanations for model interpretability.

Main Results:

• Machine learning models demonstrated superior predictive performance (AUC: 0.87-0.82) compared to traditional scores (AUC: 0.70-0.74).
• Random forest achieved the highest AUC of 0.87 for predicting operative mortality.
• Shapley's analysis identified top predictors and provided individual-level risk explanations.

Conclusions:

• Machine learning models show potential to outperform clinical scoring tools in predicting cardiac surgery mortality.
• Explainable AI can provide personalized risk profiles, aiding clinical decision-making.
• Further multicenter studies are needed to validate the clinical utility of these models.