TABLE 3.

Prediction models of postoperative major bleeding in the test set.

Models	AUC (95%CI)	P-value*	P-value†	P-value‡
TRUST	0.629 (0.517–0.741)
WILL-BLEED	0.557 (0.449–0.665)
Logistic regression	0.702 (0.577–0.827)	0.347	0.103
Support vector machine	0.792 (0.678–0.907)	0.016	0.002	0.031
Xgboost	0.802 (0.691–0.913)	0.028	<0.001	0.120
Random forest	0.810 (0.719–0.902)	0.005	<0.001	0.084
Conditional inference random forest	0.831 (0.732–0.930)	0.002	<0.001	0.027
Stochastic gradient boosting	0.811 (0.739–0.883)	<0.001	<0.001	0.073
Naïve Bayes	0.687 (0.561–0.813)	0.468	0.059	0.842
Bagged CART	0.791 (0.706–0.876)	0.008	<0.001	0.098
Boosted classification trees	0.794 (0.675–0.913)	0.007	0.003	0.117

AUC, the area under the receiver operating characteristic curve; CART, classification and regression tree.

*Compared with TRUST.

^†Compared with WILL-BLEED.

^‡Compared with the logistic regression model.