Objective: The purpose of this research was to evaluate the discrimination and calibration of mortality prediction of Simplified Acute Physiology Score 2, Simplified Acute Physiology Score 3, Mortality Probability Model II, and Mortality Probability Model III in peripartum women.
Design: A retrospective cohort study.
Setting: Rafael Calvo Maternity Hospital, a large teaching hospital in Cartagena (Colombia).
Patients: All obstetric patients admitted to the ICU from 2006 to 2011.
Interventions: None.
Measurements and main results: Seven hundred twenty-six obstetric critical care patients were included. All scores showed good discrimination (area under the receiver operator characteristic curve > 0.86). Simplified Acute Physiology Score 2, Simplified Acute Physiology Score 3, and Mortality Probability Model III inaccurately estimated mortality. The only mortality prediction score that showed good calibration through mortality ratio and Hosmer-Lemeshow test was Mortality Probability Model II. Mortality ratio for Mortality Probability Model II was 0.88 (95% CI, 0.60-1.25). Hosmer-Lemeshow test was not significant (p = 0.571).
Conclusions: Simplified Acute Physiology Score 2 and Simplified Acute Physiology Score 3 overestimate mortality in obstetric critical care patients. Mortality Probability Model III was inadequately calibrated. Mortality Probability Model II showed good fit to predict mortality in a developing country setting. Future studies in developed and developing countries are needed to further confirm our findings.