Table 4.

Performance of the candidate models and baseline models on the validation set. The best-performing parameters are indicated in italic text.

Model	Optimal cutoff	Specificity	Precision	Recall	F1 measure	AUCa
Logistic regression	0.157	0.642	0.857	0.729	0.773	0.741
Naïve Bayes	0.220	0.666	0.855	0.685	0.740	0.720
Alternating decision tree	0.298	0.662	0.857	0.705	0.755	0.732
Random forest	0.122	0.747	0.862	0.611	0.680	0.726
XGBoostb	0.175	0.611	0.856	0.759	0.794	0.743
Neural network	0.125	0.686	0.858	0.681	0.737	0.735
HOSPITAL score	4	0.564	0.838	0.694	0.745	0.688
LACE index	11	0.469	0.830	0.745	0.779	0.675
LACE-rt index	7	0.542	0.833	0.688	0.740	0.668

^aAUC: area under the curve.

^bXGBoost: extreme gradient boosting.