Performance of APACHE II and SAPS 3

Regional adaptation in a population of critically Ill patients in Colombia

Abstract

The simplified acute physiological score (SAPS 3) has been recently developed with samples from the five continents, but has not been validated in patients in intensive care units (ICU) in Colombia. In this work the performance of SAPS 3 with APACHE II in Colombian ICU patients was compared.

Methods: in a population of a historical cohort of critically ill patients admitted to an ICU of a university hospital between January 1°, 2006 to June 30, 2011, the performance of the mortality models was measured. The probability of hospital death was calculated with APACHE II and SAPS 3. To improve the accuracy of the models, a first order adaptation was realized, using the logistic regression of the original score for the APACHE II, and for SAPS 3 the calculation of the probability of death was done with the formulas for Eastern Europe, Central and South America and Australasia.

Results: the study included 2523 patients. Hospital mortality was 27%. Discrimination was acceptable for all models, being lower for APACHE II, which improves with adaptation from 0.74 to 0.78. For all forms of SAPS 3, the area under the ROC curve was 0.78. Calibration, measured with the Hosmer-Lemeshow statistic was poor for the APACHE II, adjusted APACHE II, SAPS 3 adapted to Central South America and Eastern Europe, but it was good for the global SAPS 3 and for the one adapted to Australasia.

Conclusion: in a population of Colombian critically ill patients, adaptation manages to improve discrimination of APACHE II but not its calibration. Calibration is appropriate only for the global SAPS 3 and the one adapted to Australasia