Mortality prediction by SOFA score in ICU-patients after cardiac surgery; comparison with traditional prognostic–models

Table 3 Performance of the models for predicting ICU mortality; N = 36,632 patients

Models	AUC (CI 95%) *	Brier score (CI 95%)	Ĉ-statistic (CI)*	Ĉ-statistic p-value
APACHE IV – model	0.906 (0.904–0.906)	0.011 (0.011–0.011)	38.5 (30.2–54.8)	< 0.001
APACHE II – model	0.892 (0.891–0.893)	0.011 (0.011–0.011)	27.1 (14.1–35.4)	0.001
SOFA - model	0.865 (0.864–0.866)	0.012 (0.012–0.012)	16.4 (12.9–25.1)	0.030
SAPS-II - model	0.919 (0.917–0.919)	0.011 (0.011–0.012)	9.7 (5.7–21.3)	0.215
MPM₂₄-II - model	0.862 (0.860–0.863)	0.012 (0.012–0.012)	7.2 (2.8–15.3)	0.462

*AUC: area under the receiver operating characteristic curve; Ĉ-statistic: Hosmer and Lemeshow goodness-of-fit Ĉ-statistic; CI: 95% confidence interval

ISSN: 1471-2253