Machine learning algorithm to predict mortality in patients undergoing continuous renal replacement therapy

Table 3 In-hospital mortality prediction models in the test set

Models	AUC (95% CI)	P value*	P value^†	P value^‡	Accuracy	F1 score
APACHE II	0.593 (0.563–0.622)				0.586	0.654
SOFA	0.664 (0.636–0.691)				0.603	0.645
MOSAIC	0.690 (0.641–0.740)				0.633	0.656
κ-Nearest neighbor	0.721 (0.675–0.767)	< 0.001	0.037	0.379	0.673	0.776
Support vector machine	0.755 (0.711–0.799)	< 0.001	< 0.001	0.054	0.686	0.782
Multivariate adaptive regression splines	0.756 (0.713–0.799)	< 0.001	< 0.001	0.050	0.694	0.781
Random forest	0.768 (0.726–0.810)	< 0.001	< 0.001	0.019	0.700	0.757
Extreme gradient boost	0.754 (0.709–0.798)	< 0.001	< 0.001	0.062	0.711	0.790
Artificial neural network	0.762 (0.719–0.806)	< 0.001	< 0.001	0.032	0.707	0.790

Abbreviations: AUC area under the curve, CI confidence interval, APACHE Acute Physiology and Chronic Health Evaluation, SOFA Sequential Organ Failure Assessment, MOSAIC Mortality Scoring system for AKI with CRRT
*Compared with the APACHE II model
^†Compared with the SOFA model
^‡Compared with the MOSAIC model

ISSN: 1364-8535