An In-hospital Mortality Risk Model for Patients Undergoing Coronary Artery Bypass Grafting based on machine learning: Cohort Study (Preprint)

BACKGROUND Ischemic heart disease remains the leading cause of death worldwide, coronary artery bypass grafting (CABG) remains the primary surgical treatment. A series of prediction models for operative mortality risk applicable to CABG have been established, but several problems still exist in poor discrimination, accuracy and applicability. OBJECTIVE To develop an innovative mortality risk prediction system for patients undergoing CABG based on multiple machine learning algorithms. METHODS From January 2017 to December 2020, patients underwent CABG in the Chinese Cardiac Surgery Registry (CCSR) were included. We developed and tested machine learning models to predict in-hospital mortality risk, comparing them to the EuroSCORE Ⅱ. End point was in-hospital mortality. RESULTS A total of 21443 patients were included. Overall, in-hospital mortality was 2.1%. extreme gradient boosting (XGBoost) model has the best discrimination and calibration in both training and test cohorts which were superior to the EuroSCORE Ⅱ. CONCLUSIONS XGBoost prediction model can predict in-hospital mortality risk for patients after CABG in China.