Fig. 2From: Risk factor analysis of device-related infections: value of re-sampling method on the real-world imbalanced datasetComparison in ROC curves and AUCs between before (left) and after (right) re-sampling in prediction of CIEDI with similar variables. As shown in Figure, whether before or after re-sampling, replacement procedure, gender, and age all had large AUC. This suggests potential roles of them in prediction of patients at risks for CIEDIBack to article page