Skip to main content

Table 2 The performance of the SAPS 3 model and the ANN model for different primary ICU diagnoses based on the test set (n = 36,214)

From: Artificial neural networks improve and simplify intensive care mortality prognostication: a national cohort study of 217,289 first-time intensive care unit admissions

 

Number of patients

AUC of SAPS 3

AUC of ANN

p value

Test set

36,214

0.850 (0.846–0.855)

0.889 (0.885–0.893)

<10−15

Cardiac arrest

1,651

0.858 (0.835–0.881)

0.893 (0.875–0.912)

<10−7

Septic shock

1,481

0.846 (0.821–0.870)

0.889 (0.869–0.909)

<10−8

Respiratory failure

1,467

0.830 (0.804–0.856)

0.878 (0.855–0.900)

<10−8

Gastrointestinal haemorrhage

1,324

0.878 (0.858–0.900)

0.910 (0.892–0.927)

<10−5

SIRS

1,320

0.836 (0.811–0.862)

0.884 (0.863–0.906)

<10−8

Trauma

1,301

0.844 (0.820–0.869)

0.882 (0.860–0.903)

<10−5

Bacterial pneumonia

1,173

0.856 (0.830–0.882)

0.895 (0.874–0.916)

<10−7

Seizures

797

0.847 (0.814–0.880)

0.892 (0.865–0.918)

<10−4

Head injury

760

0.833 (0.796–0.869)

0.888 (0.860–0.916)

<10−5

  1. Mean, 95% confidence intervals, and p values were obtained using the method of DeLong [18] SIRS Systemic Inflammatory Response Syndrome