Skip to main content

Table 3 The 1-year mortality prediction performance of logistic regression models and random forest models for the training sample and test sample

From: Death after discharge: prognostic model of 1-year mortality in traumatic brain injury patients undergoing decompressive craniectomy

   Cutoff Accuracy Sensitivity Specificity AUC
Logistic regression Train data 0.5 0.744 0.346 0.917 0.770
Optimal (0.307) 0.750 0.731 0.758
Test data 0.5 0.655 0.167 0.875 0.765
Optimal (0.195) 0.672 1.000 0.525
SMOTE logistic regression Train data 0.5 0.686 0.692 0.683 0.760
Optimal (0.641) 0.756 0.615 0.817
Test data 0.5 0.741 0.722 0.750 0.843
Optimal (0.405) 0.741 0.889 0.675
SMOTE random forest Train data 0.5 0.959 1.000 0.942 0.998
Optimal (0.596) 0.983 1.000 0.975
Test data 0.5 0.828 0.778 0.850 0.830
Optimal (0.487) 0.810 0.833 0.800