Prognostic performance and customization of the SAPS II: results of a multicenter Austrian study

Abstract

Objectives: To evaluate the prognostic performance of the original Simplified Acute Physiology Score (SAPS) II in Austrian intensive care patients and to evaluate the impact of customization. Design: Analysis of the database of a multicenter study. Setting: Nine adult medical, surgical, and mixed intensive care units (ICUs) in Austria. Patients: A total of 1733 patients consecutively admitted to the ICUs. Measurements and results: The database included admission data, SAPS II, length of stay, and hospital mortality. The Hosmer–Lemeshow goodness-of-fit test for the SAPS II showed a lack of uniformity of fit (H = 89.1, 10 df, p < 0.0001; C = 91.8, 10 df, p < 0.0001). Subgroup analysis showed good performance in patients with cardiovascular (medical and surgical) diseases as the primary reasons for admission. A new predictive equation was derived by means of the logistic regression. Goodness-of-fit was excellent for the customized model (SAPS II_AM) (H = 11.2, 9 df, p = 0.33, C = 11.6, 9 df, p = 0.24). The mean standardized mortality ratio (SMR) changed from 0.81 ± 0.26 to 0.93 ± 0.29 with customization. Conclusions: SAPS II was not well calibrated when applied to all patients. However, it performed well for patients with cardiovascular diseases as the primary reason for admission and may thus be applied to these patients. Standardized mortality ratios that are calculated from scoring systems without known calibration must be viewed with skepticism.