Table 4.

Quality parameter of the different models (n = 594, validation data set)

Model	P	A	P+A	B1	P+A+B1	B2	P+A+B1+B2	TRISS	RISC
Quality criteria
Discrimination
Prognosis survivors (S)	12.5	11.5	10.7	10.7	9.6	10.5	9	11.9	10
Prognosis Non-survivors	16.1	26.9	31	31	38.1	32.2	43.5	48.8	48
(N-S)
Difference between S	3.6	15.4	20.3	20.3	28.5	21.7	34.5	36.9	38
and N-S
Sensitivity	0.0	16.0	26.7	24.0	33.3	16.0	36.0	44.6	44.0
Specificity	100.0	98.5	97.1	99.4	97.1	97.7	97.7	92.4	97.1
Overall accuracy	87.4	88.0	88.2	89.9	89.1	87.4	89.9	87.7	90.4
Nagelkerkes R2 [24]	-	-	-	-	-	-	-	-	-
AUC of ROC	0.61	0.75	0.79	0.77	0.82	0.82	0.87	0.84	0.89
95% CI	0.54-0.68	0.68-0.81	0.73-0.85	0.71-0.83	0.76-0.88	0.77-0.87	0.82-0.91	0.79-0.90	0.85-0.93
Precision
Observed mortality (%)	12.6	12.6	12.6	12.6	12.6	12.6	12.6	12.6	12.6
Score prognosis (%)	13.0	13.4	13.3	13.3	13.2	13.3	13.4	16.5	14.8
Calibration
H-L [25]	n.a.	3.96	4.68	4.66	6.92	11.04	6.01	35.60	8.34
p-value	n.a.	0.138	0.322	0.200	0.230	0.137	0.420	< 0.001	0.400

P Patients basic data; A Prehospital phase; B1 early trauma room phase; B2 late trauma room phase; TRISS trauma and injury severity score, RISC revised injury severity classification score; S survivors; N-S non-survivors; AUC area under the curve. ROC receiver operating characteristic; CI confidence interval; H-L Goodness of fit statistics according to Hosmer and Lemeshow [25]; n.a. not applicable