Fig. 2: ROC curves for PCE and ML model performance.
ROC curves are outlined for PCE versus model ML performance in the PCE-eligible cohort (in a cross-validation and b held-out test data) and ML model performance on the full cohort (in c cross-validation and d held-out test data). Legend entries denote the AUC for each method with 95% confidence intervals. ROC receiver-operating characteristic, PCE pooled cohort equations, ML machine learning, AUC area under receiver-operating characteristic curve, LRL2 logistic regression with an L2 penalty, LRLasso logistic regression with an L1 (lasso) penalty, RF random forest, GBM gradient boosting machine, XGBoost extreme gradient boosting.
