PURPOSE: Study Objectives: Specific features of different populations may influence the results of prognostic indexes. Literature shows differences in SMR (Standartized Mortality Rate), calibration, and discrimination of such indexes. The present study intends to evaluate the accuracy of UNICAMP II Model on predicting survival and to compare it to APACHE II in a general intensive care unit (ICU).Design: Retrospective analysis of a prospectively collected general ICU database.Interventions: None.
METHODS: A total of 2,941 patients were admitted to the ICU from January 1, 2002 to March 31, 2006; 1,029 were excluded, according to the following criteria: 1 –coronary insufficiency as the principal cause of admission: 454 (15.44%); 2 –cardiac surgery: 26 (.88%); 3 –incomplete or uncertain data: 318 (10.81%); 4 –readmissions: 231 (7.85%). In the 1,912 patients included in the study we evaluated SMR, calibration through the Hosmer-Lemeshow Hg test, and discrimination by the under the ROC curve area. The evaluated outcome was death or hospital discharge. We also evaluated the performance of the two indexes for the subgroup of clinical and surgical patients. Calculations were proceeded by using EXCEL 2000 Microsoft Corporation program and Medcalc Version 184.108.40.206 Frank Schoonjans.
RESULTS: The calculated UNICAMP II Model SMR approached to one (.9629) in the global evaluation of patients, and in clinical patients (1.0360), but overestimated mortality in surgical patients (.7758). The calculated APACHE II SMR approached to one in surgical patients (1.0834), but underestimated mortality in the global evaluation of patients (1.4227) and in clinical patients (1.5661). Both indexes showed excellent discrimination in all evaluations (under the ROC curve area > .8), whereas calibration was inadequate (P < .05), except for the evaluation of clinical patients through UNICAMP II Model (P = .255).
CONCLUSION: UNICAMP II Model had a better accuracy on measuring ICU clinical patients outcome in this study.
CLINICAL IMPLICATIONS: UNICAMP II Model might be considered to measure outcome of ICU clinical patients in Brazil.
DISCLOSURE: Carlos Alves, None.