Evaluation of standard and modified severity of illness scores in the obstetric patient

Abstract Purpose To test discrimination and calibration of APACHE-II and SAPS-II risk prediction scores in a cohort of obstetric patients, and to evaluate the effect of modifying these scores for the physiological changes in pregnancy. Materials and Methods A retrospective review of obstetric patients, 12 weeks gestation to 48 hours postpartum, admitted to the ICU for more than 24 hours. APACHE-II and SAPS-II, and versions modified for the physiological changes of pregnancy, were evaluated by receiver operating characteristic (ROC) curves and standardized mortality ratios (SMR). Multivariable analysis identified other parameters associated with mortality. Results Data were obtained from 332 patients from 5 countries, with a mortality rate of 12%. Mean (± SD) APACHE-II score was 16.8 ± 6.1 and SAPS-II score 26.5 ± 15.8. Good discrimination was demonstrated with area under the ROC curves of 0.82 and 0.78 respectively, with no improvement after modification for altered maternal physiology. APACHE-II overestimated mortality, with an SMR of 0.43 (0.52 after including diagnostic weighting) compared with 0.89 for SAPS-II. Bilirubin, albumin and Glasgow Coma Scale were independently associated with mortality. Conclusion APACHE-II and SAPS-II are good discriminators of illness severity and may be valuable for comparing obstetric cohorts, but APACHE-II significantly over-estimates mortality.