Machine learning prediction models for prognosis of critically ill patients after open-heart surgery

Abstract

We aimed to build up multiple machine learning models to predict 30-days mortality, and 3 complications including septic shock, thrombocytopenia, and liver dysfunction after open-heart surgery. Patients who underwent coronary artery bypass surgery, aortic valve replacement, or other heart-related surgeries between 2001 and 2012 were extracted from MIMIC-III databases. Extreme gradient boosting, random forest, artificial neural network, and logistic regression were employed to build models by utilizing fivefold cross-validation and grid search. Receiver operating characteristic curve, area under curve (AUC), decision curve analysis, test accuracy, F1 score, precision, and recall were applied to access the performance. Among 6844 patients enrolled in this study, 215 patients (3.1%) died within 30 days after surgery, part of patients appeared liver dysfunction (248; 3.6%), septic shock (32; 0.5%), and thrombocytopenia (202; 2.9%). XGBoost, selected to be our final model, achieved the best performance with highest AUC and F1 score. AUC and F1 score of XGBoost for 4 outcomes: 0.88 and 0.58 for 30-days mortality, 0.98 and 0.70 for septic shock, 0.88 and 0.55 for thrombocytopenia, 0.89 and 0.40 for liver dysfunction. We developed a promising model, presented as software, to realize monitoring for patients in ICU and to improve prognosis.