Table 3.

Model prediction comparisons with the test data set.

Model		Sensitivity	Specificity	Accuracy	Balanced accuracy	AUROCa
921 features (including ICD-10b)
	AdaBoostc	0.9739	0.9888	0.9887	0.9813	0.9974
	XGBoostd	0.9689	0.9888	0.9888	0.9789	0.9972
	LightGBMe	0.9701	0.9887	0.9887	0.9794	0.9973
	GBMf	0.7954	0.9506	0.9504	0.8730	0.9350
	ERTg	0.9010	0.9270	0.9270	0.9140	0.9580
	LRh	0.9584	0.9922	0.9922	0.9753	0.9952
	RFi	0.9433	0.9537	0.9538	0.9476	0.9848
	DNNj	0.9694	0.9845	0.9845	0.9770	0.9931
	AdaBoost+XGBoost	0.9700	0.9889	0.9889	0.9794	0.9973
	AdaBoost+LightGBM	0.9702	0.9888	0.9888	0.9795	0.9974
	XGBoost+LigtGBM	0.9694	0.9888	0.9888	0.9791	0.9973
	AdaBoost+XGBoost+LightGBM	0.9698	0.9888	0.9888	0.9793	0.9973
	In-hospital mortality AIk [5]	0.9468	0.9761	0.9760	0.9614	0.9929
878 features (ICD-10 only)
	AdaBoost	0.8313	0.9476	0.9475	0.8895	0.9485
	XGBoost	0.6933	0.9751	0.9748	0.8342	0.8943
	LightGBM	0.6995	0.9745	0.9742	0.8370	0.8953
	GBM	0.9730	0.9774	0.9774	0.9752	0.9958
	ERT	0.9739	0.9810	0.9810	0.9775	0.9938
	LR	0.7623	0.9558	0.9557	0.8591	0.9437
	RF	0.6583	0.9740	0.9737	0.8161	0.9313
	DNN	0.8809	0.8959	0.8959	0.8884	0.9438
	AdaBoost+XGBoost	0.7046	0.9747	0.9744	0.8397	0.9452
	AdaBoost+LightGBM	0.7057	0.9744	0.9741	0.8400	0.9452
	XGBoost+LigtGBM	0.6950	0.9748	0.9745	0.8349	0.8954
	AdaBoost+XGBoost+LightGBM	0.7024	0.9746	0.9743	0.8385	0.9450
	In-hospital mortality AI [5]	0.2838	0.9751	0.9751	0.6298	0.7675
43 features (excluding ICD-10)
	AdaBoost	0.9743	0.9862	0.9862	0.9802	0.9965
	XGBoost	0.9684	0.9886	0.9886	0.9785	0.9965
	LightGBM	0.9691	0.9884	0.9884	0.9788	0.9966
	GBM	0.7954	0.9506	0.9504	0.8730	0.9350
	ERT	0.9010	0.9270	0.9270	0.9140	0.9580
	LR	0.9519	0.9906	0.9906	0.9713	0.9935
	RF	0.9040	0.9492	0.9492	0.9266	0.9806
	DNN	0.9345	0.9784	0.9783	0.9565	0.9860
	AdaBoost+XGBoost	0.9687	0.9887	0.9887	0.9787	0.9966
	AdaBoost+LightGBM	0.9693	0.9886	0.9885	0.9789	0.9966
	XGBoost+LigtGBM	0.9690	0.9885	0.9885	0.9788	0.9965
	AdaBoost+XGBoost+LightGBM	0.96914	0.9886	0.9885	0.9788	0.9966
	In-hospital mortality AI [5]	0.9441	0.9855	0.9855	0.9648	0.9923
Traditional methods
	Inclusive SRRl	0.8716	0.9389	0.9388	0.9052	0.9328
	Exclusive SRR	0.9070	0.9227	0.9227	0.9149	0.9567
	KTASm	0.9465	0.9777	0.9777	0.9621	0.9405

^aAUROC: area under the receiver operating characteristic curve.

^bICD-10: International Classification of Disease 10th revision.

^cAdaBoost: adaptive boosting.

^dXGBoost: extreme gradient boosting.

^eLightGBM: light gradient boosting machine.

^fGBM: gradient boosting machine.

^gERT: extremely random trees.

^hLR: logistic regression.

ⁱRF: random forest.

^jDNN: deep neural network.

^kAI: artificial intelligence.

^lSRR: survival risk ratio.

^mKTAS: Korean Triage and Acuity Scale.