Skip to main content

Table 2 Performance metrics of the gradient-boosted decision trees models with 33 and five features (33F and Top5F, respectively), the Emergency Heart Failure Mortality Risk Grade (EHMRG) model, a single decision tree (DT) with the same five features as Top5F, and a logistic regression (LR) model for the prediction of seven-day mortality

From: Prediction of short-term mortality in acute heart failure patients using minimal electronic health record data

 

33F

Top5F

EHMRG

DT

LR

AUROC

0.843

0.830

0.776

0.589

0.685

Sensitivity

0.895

0.895

0.684

0.211

0.579

Specificity

0.749

0.618

0.642

0.967

0.722

LR+

3.558

2.344

1.911

6.469

2.082

LR-

0.141

0.170

0.492

0.816

0.583

DOR

25.300

13.771

3.886

7.927

3.569