Assessing the Performance of Two Clinical Severity Scoring Systems in the ICU of a Tertiary Respiratory Disease Center


1 Tracheal Disease Research Center

2 Department of Pulmonary Medicine

3 Tobacco Prevention and Control Research Center

4 Lung Transplantation Research Center, NRITLD, Shahid Beheshti University M.C., TEHRAN-IRAN.

5 Department of Pulmonary Medicine,


Background: The aim of this study is to compare the performance of five applied general severity scoring systems and their ability to predict mortality rate for the intensive care unit patients: Simplified Acute Physiology Score II (SAPS II), Mortality Probability Model II at admission (MPM II0), at 24 hours (MPM II24), at 48 hours (MPM II48) and over time (MPM IIovertime). These scoring systems have been developed in response to an increased emphasis on the evaluation and monitoring of health care services; and also making cost-effective decisions. Materials and Methods: In this historical cohort study, all of the scoring systems were applied to 114 patients and the predicted mortality rate and the Standardized Mortality Ratio (SMR) were calculated for them. Calibration of each model and discriminative powers were evaluated by using Hosmer-Lemeshow goodness of fit test and ROC curve analysis, respectively. Results: The predicted mortalities were not significantly deviated from the main systems (SMR for SAPS II: 0.79, MPM II0: 1.10, MPM II24: 1.32, MPM II48: 1.08 and MPMOvertime: 1.02). The Hosmer-Lemeshow statistics had the least value for MPM II48 (C=2.922, p-value=0.939); and the discrimination was best for MPM II24 (AUC=0.927) followed by SAPS II (AUC=0.903), MPM II0 (AUC=0.899), MPM II48 (AUC=0.848) and MPM IIovertime(AUC=0.861). Conclusion: All five general ICU morality predictors showed accurate standardized mortality ratio. MPM II24 had the best discrimination, MPM II0 had the best SMR before 24 hours and MPMovertime had the best SMR after 24 hours. Performance of MPM II and its ease of use make it an efficient model for mortality prediction in our study. (Tanaffos2010; 9(3): 58-64)