Machine Learning Model Improves Mortality Risk Prediction In Cardiac Surgery

This article discusses a machine learning-based model that was developed to more accurately predict mortality after cardiac surgery than population-based models. The model was developed from a cohort of 6,392 patients who underwent cardiac surgery between 2011 and 2016. The researchers used 4,016 features and randomly assigned them to the training/development cohort and the test/evaluation cohort. The best-performing predictor was the eXtreme Gradient Boosting (XGBoost) algorithm, which achieved an area under the receiver operating characteristic curve of 0.845.