Mortality predictive indexes in non-critical inpatients.

BACKGROUND Mortality predictive indexes have not been applied to patients in general wards out of the ICU. METHODS Retrolective study aimed to evaluate the value of mortality prediction indexes in a cohort of 944 non-critical patients. Three indexes were evaluated according to their calibration and discriminative power: the Mortality Probability Model II (MPMII), the Simplified Acute Physiology System II (SAPS II) and the Logistic Organ Dysfunction System (LODS). The bivariate calculation of relative risk (RR) to die was performed relative to the group of patients that had an expected probability to die > 10%, calculated by an index. To evaluate the calibration, data were arranged in descending order using the chi2 goodness-of-fit model. To evaluate discrimination power, ROC curves were used. RESULTS SAPS II, MPM II and LODS predicted significant risks at levels of P < 0.005, (RR = 6.56, 4.03 and 3.44, respectively). Regarding the calibration, the null hypothesis was accepted only by using SAPS II (P = 0.664). CONCLUSIONS The three evaluated indexes each had a good discriminative capacity to detect non-critical inpatients with high risk to die. SAPS II was the best index to predict mortality, as determined by both the bivariate and the calibration analysis. There is no reason for not using mortality predictive indexes for non-critical inpatients.