Table 4.

AUROC and time points of mortality prediction studies

Study	Model	Algorithm	AUROC	Time
Taylro, 2015	Logistic regression	Logistic regression	0.755	28 days
	CART	Classification and regression tree	0.693
	Random forest	Random forest	0.860
	MEDS score	NR	0.705
	CURB-65 score	NR	0.734
	REMS score	NR	0.717
Perng et al. (2019)	KNN	KNN	0.84	28 days
	SoftMax	SoftMax	0.88
	PCA + SoftMax	PCA + SoftMax	0.91
	AE + SoftMax	AE + SoftMax	0.90
	CNN + SoftMax	CNN + SoftMax	0.92
Kwon and Baek (2020)	qSOFA scores	NR	0.78	3 days
	qSOFA-based machine-learning models	Extreme gradient boosting, light gradient boosting machine, and random forest	0.86
Hou et al. (2020)	XGBoost	eXtreme Gradient Boosting	0.857	30 days
	logistic regression	logistic regression	0.819
	SAPS-II scores	Simplified acute physiology score-II	0.797
Kong et al. (2020)	LASSO	least absolute shrinkage and selection operator	0.829	In hospital
	RF	random forest	0.829
	GBM	gradient boosting machine	0.845
	LR	logistic regression	0.833
	SAPS II	Simplified acute physiology score-II	0.77
Li et al. (2021)	GBDT	GBDT	0.992	In hospital
	LR	Logistic regression	0.876
	KNN	k-nearest neighbor	0.877
	RF	Random forest	0.980
	SVM	Support vector machine	0.898
Qi et al. (2021)	XGBoost	Extreme gradient boosting	0.848	In hospital
	SAPSII	The simplified acute physiology score	0.777
	SOFA	Sequential organ failure assessment score	0.704
	SIRS	Systemic inflammatory response syndrome	0.609
	qSOFA	Quick sequential organ failure assessment	0.580

Abbreviation: CART: classification and regression tree; MEDS: mortality in emergency department sepsis score; KNN: K nearest neighbor; REMS: rapid emergency medicine score; CURB-65 score: the confusion, urea nitrogen, respiratory rate, blood pressure, 65 years of age and older; PCA: principal component analysis; AE: Autoencoder; CNN: Convolutional Neural Network; qSOFA: quick Sequential Organ Failure Assessment.