Skip to main content

Table 2 Overall performance characteristics

From: Prospective evaluation of an automated method to identify patients with severe sepsis or septic shock in the emergency department

Alert

Cutoff

sens

fpr

auc

spec

ppv

npv

FP/Day

Inc. Sepsis TP

Nurse

Other

Triangle Nurse

1.000

54.3 %

3.1 %

0.756

96.9 %

6.1 %

99.8 %

6.44

  

SIRS Derived

1.000

21.6 %

0.4 %

0.606

99.6 %

17.2 %

99.7 %

0.80

187

4

Cut05

0.050

76.4 %

4.7 %

0.859

95.3 %

5.8 %

99.9 %

9.64

22

100

Cut04

0.040

79.0 %

5.4 %

0.868

94.6 %

5.2 %

99.9 %

11.16

18

105

Sens80

0.036

80.1 %

5.8 %

0.871

94.2 %

4.9 %

99.9 %

12.03

17

108

FPC15

0.026

85.2 %

7.3 %

0.890

92.7 %

4.2 %

99.9 %

15.00

13

122

MaxAUC

0.017

89.2 %

9.9 %

0.897

90.1 %

3.3 %

100.0 %

20.33 %

12

135

NrsFP

0.097

70.2 %

3.1 %

0.835

96.9 %

7.8 %

99.9 %

6.44

32

88

NrsSens

0.330

54.3 %

1.6 %

0.763

98.4 %

11.4 %

99.8 %

3.27

64

64

  1. SIRS systemic inflammatory response syndrome, Cut05/Cut04 Bayesian cutpoint threshold of 0.05/0.04, Sens80 threshold for sensitivity = 80 %, FPC15 threshold of 15 false positive alerts per day, MaxAUC threshold maximizing area under the receiver operating characteristic curve, NrsFP nurse-driven false positive threshold, NrsSens nurse-driven sensitivity threshold. Boldface indicates the primary cutpoint used