Peer Review of “Performance Drift in Machine Learning Models for Cardiac Surgery Risk Prediction: Retrospective Analysis”

This manuscript [1] presents an interesting study that explores temporal trends in various performance metrics for different types of prediction models used in the prediction of in-hospital mortality after cardiac surgery in the United Kingdom from 2012 to 2019.The data set was divided into 2 periods: from 2012 to 2016 for model training and internal validation and from 2017 to 2019 for external validation.The study evaluated 5 prediction models: logistic regression, support vector machine (SVM), random forest, extreme gradient boosting (XGBoost), neural network, and European System for Cardiac Operative Risk Evaluation (EuroSCORE) II.The authors aimed to assess the model performance on 5 metrics (1 -expected calibration error [ECE], area under the curve [AUC], 1 -Brier score, F 1 -score, and net benefit) and proposed a composite metric, the clinical effectiveness metric (CEM), calculated as the geometric mean of the 5 mentioned metrics, as the primary metric.