Skip to main content
. 2001 Jan 5;5(1):31–36. doi: 10.1186/cc975

Table 5.

Discrimination and calibration of APACHE III before and after customization for various subgroups of the validation sample

Subgroup Prediction n ROC area H C SMR (95% CI)
Medical Original 590 0.861 22.01 24.90 1.15 (1.01-1.29)
Customized 0.861 15.83* 18.27* 0.93 (0.81-1.06)
Surgical Original 299 0.842 8.51 13.18 1.28 (0.96-1.60)
Customized 0.842 4.75 7.62 0.98 (0.71-1.25)
< 65 years Original 464 0.894 16.62* 15.59* 1.19 (0.95-1.42)
Customized 0.894 21.24 10.19 0.94 (0.74-1.14)
≥ 65 years Original 425 0.812 13.63 15.74* 1.17 (1.02-1.33)
Customized 0.812 12.93 9.42 0.95 (0.81-1.08)
Direct admission Original 614 0.874 19.41* 23.46 1.29 (1.10-1.48)
Customized 0.874 9.07 11.67 1.02 (0.86-1.18)
Transferral Original 275 0.817 9.54 12.18 1.05 (0.87-1.23)
Customized 0.817 15.80* 15.18* 0.86 (0.70-1.01)
Cardiovascular Original 317 0.890 16.72* 19.21* 1.11 (0.92-1.35)
Customized 0.890 18.47* 15.61* 0.90 (0.74-1.07)
Gastrointestinal Original 296 0.887 8.17 5.56 1.08 (0.81-1.34)
Customized 0.887 7.80 5.26 0.85 (0.62-1.08)

Direct admission is defined as admission from emergency room, operating theatre, or recovery room; and transferral is defined as admission from other hospital, other ICU, or floor. C, Hosmer-Lemeshow goodness-of-fit C test; H, Hosmer-Lemeshow goodness-of-fit H test; CI, confidence interval. *P < 0.05, P = 0.01.