Table 6.

Comparison of predictive performance between training set and validation set

	AUC (95% CI, Training set)	AUC (95% CI, Validation set)	P value
ANN	0.873 (0.846–0.900)	0.811 (0.792–0.830)	< 0.001
Logistic regression	0.720 (0.684–0.756)	0.752 (0.731–0.773)	0.067
APACHEII	0.629 (0.607–0.651)	0.607 (0.584–0.630)	0.174
SOFA	0.619 (0.596–0.641)	0.628 (0.605–0.651)	0.350

ANN artificial neural networks, SOFA sequential organ failure assessment, APACHE acute physiology, and chronic health evaluation, AUC area under the ROC curve, CI confidential interval