Skip to main content

Table 2 Multi-variate analysis of unanticipated ICU admissions and in-hospital mortality

From: Predictive modeling of perioperative patient deterioration: combining unanticipated ICU admissions and mortality for improved risk prediction

Method

AUROC

AUPRC

AUKC

Precision

Sensitivity

Kappa

F1 score

End-point: Unanticipated ICU admissions

     LR

83.4% (2.7%)

5.1% (1.9%)

2.1% (0.3%)

8.9% (4.4%)

12.5% (6.9%)

0.10 (0.05)

10.3% (5.2%)

     LR (Combined)a

82.1% (3.1%)

5.1% (1.8%)

2.1% (0.3%)

9.0% (3.4%)

12.5% (5.3%)

0.10 (0.04)

10.3% (4.0%)

End-point: In-hospital mortality

     LR

91.8% (2.9%)

6.4% (3.4%)

1.6% (0.4%)

7.0% (3.1%)

26.8% (11.8%)

0.11 (0.05)

11.0% (4.7%)

     LR (Combined)a

89.5% (3.2%)

4.9% (4.3%)

1.3% (0.3%)

5.9% (2.8%)

22.4% (11.0%)

0.09 (0.04)

9.2% (4.3%)

End-point: Unanticipated ICU admissions & in-hospital mortality

     LR

83.5% (2.8%)

7.2% (2.8%)

2.6% (0.3%)

12.2% (4.2%)

13.5% (5.4%)

0.12 (0.05)

12.8% (4.6%)

     LR (Ensemble)b

84.5% (2.8%)

7.0% (2.6%)

2.7% (0.3%)

11.4% (4.2%)

12.7% (5.8%)

0.11 (0.05)

11.9% (5.2%)

     SVM

82.8% (2.9%)

5.5% (2.2%)

2.4% (0.3%)

8.4% (4.4%)

8.8% (4.7%)

0.08 (0.04)

8.4% (4.3%)

     XGB

79.7% (3.9%)

5.7% (2.6%)

2.2% (0.4%)

10.9% (5.5%)

8.5% (4.6%)

0.09 (0.05)

9.4% (4.8%)

     RF

81.4% (2.8%)

5.1% (1.4%)

2.3% (0.3%)

9.5% (7.0%)

4.0% (3.5%)

0.05 (0.04)

5.1% (3.7%)

  1. Area Under the Receiver Operating Characteristic curve (AUROC), Area Under the Precision-Recall curve (AUPRC), Area Under the Kappa curve (AUKC)
  2. aThe combined LR is trained on the combined outcome of unanticipated ICU admissions and mortality
  3. bThe ensemble LR is an ensemble model of the LR trained on unanticipated ICU admissions and the LR trained on mortality