Table 4.

AUROCs of models with various features.

	All variables		Preoperative variables
Features included in the model	AUROC (95% CI) in the test set	AUROC (95% CI) in the validation cohort	AUROC (95% CI) in the test set	AUROC (95% CI) in the validation cohort
Top 40	0.852 (0.838–0.865)	0.840 (0.820–0.860)	0.830 (0.815–0.845)	0.814 (0.791–0.837)
Top 30	0.849 (0.835–0.863)	0.833 (0.812–0.854)	0.825 (0.810–0.841)	0.809 (0.786–0.833)
Top 20	0.848 (0.834–0.862)	0.827 (0.806–0.849)	0.826 (0.810–0.842)	0.807 (0.784–0.831)
Top 10	0.839 (0.824–0.853)	0.807 (0.783–0.83)	0.818 (0.802–0.834)	0.800 (0.776–0.825)

The models were developed using the gradient boosting decision tree (GBDT) algorithm.

AUROC, area under the receiver operating characteristic curve.