Comparison between the SAPS 3 and APACHE II score in a general intensive care unit in Brazil
© BioMed Central Ltd 2007
Published: 19 June 2007
Specific features of different populations may influence prognostic index results. The literature shows differences in the standardized mortality rate (SMR), calibration, and discrimination of such indexes. This study intends to evaluate the capacity of SAPS 3 on predicting ICU patient outcome, using two of its equations – global and Central/South American – and to compare it with the APACHE II score in a general ICU in Brazil.
We analyzed prospectively collected data from 1 January to 31 August 2006. From the 544 admitted patients, 42 (7.72%) were excluded due to readmission. In the remaining 502 patients we analyzed the SMR, calibration through the Hosmer–Lemeshow C test, and discrimination through the area under the ROC curve (aROC). The evaluated endpoint was death or hospital discharge. We also evaluated the index performance through the SMR for subgroups of patients – clinical, surgical and according to APACHE II diagnostic categories. Calculation was performed using EXCEL 2000 software (Microsoft Corporation) and MedCalc Version 126.96.36.199 (Frank Schoonjans).
The SMR approached one when calculated through the global SAPS 3 equation and the APACHE II score (1.0644 and 1.0765, respectively). The SAPS 3 Central/South American equation overestimated mortality (SMR = 0.8182). Calibration was adequate for the SAPS 3 global and Central/South American equations (13.02, P = 0.1110 and 15.07, P = 0.0578) and was inadequate for the APACHE II score (19.53, P = 0.0123). Both indexes showed excellent discrimination (aROC > 0.8). In the evaluated subgroups there was great SMR variation (0.5419–1.5754).
The global SAPS 3 equation showed the best performance in this group of patients.