Comparison of the predictive performance of the BIG, TRISS, and PS09 score in anadult trauma population derived from multiple international trauma registries

Table 4 Quality criteria for all scores compared.

	BIG score	TRISS	PS09 score
Discrimination
AUROC curve (95% CI) All patients (n = 12206)	0.892 (0.879 - 0.906)	0.922 (0.913 - 0.932)	0.925 (0.915 - 0.934)
AUROC curve (95% CI) Blunt trauma (n = 6540)	0.876 (0.859 - 0.892)	0.917 (0.906 - 0.928)	0.921 (0.911 - 0.932)
AUROC curve (95% CI) Penetrating trauma (n = 5666)	0.920 (0.898 - 0.942)	0.929 (0.912 - 0.947)	0.921 (0.902 - 0.939)
Precision
All patients
Predicted Mortality (%)	4.8	6.6	7.9
Observed Mortality (%; 95% CI)	4.8 (4.4-5.2)
Blunt Trauma
Predicted Mortality (%)	5.5	9.6	11.9
Observed Mortality (%; 95% CI)	6.7 (6.1-7.3)
Penetrating Trauma
Predicted Mortality (%)	4.0	3.2	3.4
Observed Mortality (%; 95% CI)	2.7 (2.3-3.1)

Quality criteria for all scores on all data (n = 12,206), on all blunt(n = 6,540) and penetrating (n = 5,666) trauma patients.Data are presented as AUROC (Area under receiver operating characteristiccurve) together with a 95% Confidence Interval (CI).

ISSN: 1364-8535