Figure 1.

Comparison of model discrimination between the different models in both the external and temporal validation cohorts. As shown, with point estimates and 95% CIs for the area under the receiver operating characteristic curve (AUC), the eXtreme Gradient Boosting (XGBoost) model had the highest discrimination in both validation datasets. KNN = K-nearest neighbors, NEWS = National Early Warning Score, SCMI = Study of the Treatment and Outcomes in Critically Ill Patients With Coronavirus Disease 2019 Mortality Index, SOFA = Sequential Organ Failure Assessment, SVM = support vector machine.