Table 2.

Sensitivity and specificity of prediction models for in-hospital and out-of-hospital mortality in the training set and verification set

Subgroup	Model type	Training Cohort			Verification Cohort
		Number	Sen(95%CI)	Spe(95%CI)	Number	Sen(95%CI)	Spe(95%CI)
Death in hospital	Newly Developed Machine Learning Models
	ANN	7	0.78(0.57,0.91)	0.92(0.87,0.95)	1	0.84	0.93
	DT	6	0.83(0.67,0.92)	0.93(0.78,0.98)	2	0.44-1	0.98–0.98
	KNN	1	0.91
	LR	9	0.68(0.58,0.77)	0.91(0.85,0.95)	2	0.59–0.66	0.92–0.94
	RF	7	0.86(0.73,0.93)	0.92(0.76,0.98)
	SVM	12	0.86(0.80,0.91)	0.87(0.74,0.94)	1	0.66	0.95
	NB	6	0.82(0.74,0.87)	0.89(0.70,0.96)	1	0.59	0.89
	LDA	2	0.66–0.85	0.48–0.81
	GBM	3	0.76–0.94	0.63–0.81
	LASSO	1	0.95	0.48
	Ridge	1	0.88	0.71
	AdaBoost	1	0.34	0.97
	xgboost	1	0.79	0.81
	Other	7	0.67(0.50,0.80)	0.88(0.65,0.97)
	Overall	64	0.79(0.75,0.82)	0.89(0.86,0.92)	7	0.64(0.53,0.75)	0.95(0.92,0.97)
	Mature Clinically Recommended Tools
	CRASH core	1	0.73	0.90
	CRASH CT	2	0.72–0.84	0.72–0.90
	CRASH base	2	0.72–0.78	0.80–0.82
	Helsinki	1	0.83	0.59	1	0.66	0.84
	Marshall	1	0.75	0.98	1	0.50	0.99
	Total GCS	2	0.50–0.51	0.90–0.90
	Rotterdam	1	0.58	0.77	1	0.66	0.64
	IMPACT ct	1	0.80	0.78
	Overall	11	0.71(0.64,0.78)	0.84(0.78,0.89)
Death within 1 month	Newly Developed Machine Learning Models
	DT	1	0.71	0.77
	LR	2	0.70–0.77	0.61–0.75
	RF	1	0.80	0.84
	SVM
	LASSO/Rige
	Overall	4	0.74(0.67,0.81)	0.75(0.66,0.82)
	Mature Clinically Recommended Tools
	CRASH core	1	0.77	0.89
	CRASH CT	1	0.87	0.88
	Total GCS	1	0.60	0.70
	APACHE II	1	0.75	0.75
	IMPACT core	1	0.70	0.74
	IMPACT extended	1	0.76	0.74
	IMPACT lab	1	0.72	0.76
	Overall	7	0.75(0.68,0.80)	0.79(0.73,0.84)