Table 5

Measures of discrimination and calibration for APACHE II in obstetric and non-obstetric admissions


All obstetric admissions (n = 1565)
Direct obstetric admissions (n = 1216)
Indirect or coincidental obstetric admissions (n = 349)
Female nonobstetric admissions aged 16–50 years (n = 18,450)

Discrimination (AUC [95% CI])
     APACHE II scorea
0.839 (0.820 to 0.857)
0.819 (0.795 to 0.840)
0.839 (0.797 to 0.876)
0.812 (0.807 to 0.818)
     APACHE II probabilitya
0.810 (0.790 to 0.829)
0.737 (0.711 to 0.761)
0.806 (0.760 to 0.845)
0.855 (0.850 to 0.860)
Calibration
     Mortality ratio (95% CI)
0.245 (0.167 to 0.346)
0.214 (0.120 to 0.352)
0.282 (0.163 to 0.453)
0.907 (0.878 to 0.937)
     Hosmer–Lemeshow
          C statistic (P value)
166 (< 0.001)
101 (<0.001)
47 (<0.001)
69 (<0.001)
     Cox's regression calibration
          Slope (95% CI)
0.728 (0.491 to 0.966)
0.570 (0.277 to 0.863)
0.937 (0.509 to 1.37)
1.05 (1.01 to 1.08)
          Intercept (95% CI)
-2.09 (-2.60 to -1.59)
-2.57 (-3.39 to -1.76)
-1.69 (-2.33 to -1.05)
-0.120 (-0.179 to -0.061)
          χ2(2) (P value)a
140 (<0.001)
87 (<0.001)
55 (<0.001)
57 (<0.001)

aχ2 statistic with two degrees of freedom (and associated P value) from a likelihood ratio test of slope 1 and intercept 0 in Cox's regression calibration model. APACHE, Acute Physiology and Chronic Health Evaluation; AUC, area under the receiver operating characteristic curve; CI, confidence interval.

Harrison et al. Critical Care 2005 9(Suppl 3):S25   doi:10.1186/cc3542