Assessment and comparison of SAPS II and MPM24 II scoring systems in predicting hospital mortality in intensive care units

Abstract

Objective: To evaluate and compare the performance and validity of the Simplified Acute Physiology Score II (SAPS II) and the Mortality Prediction Model at 24 hours II (MPM24 II) scoring systems in predicting hospital mortality in Thai ICU patients. Design: Prospective, observational study. Materials and methods: Data on 639 patients consecutively admit to the ICUs of Nopparat-Rajathanee Hospital, a 580-bed university-affiliated tertiary care referring hospital, during May 1, 2008, to March 15, 2009 were used to calculate predicted mortality for SAPS II and MPM24 II using the original formulas. Standardized mortality ratio (SMR) was computed. Discrimination of both scoring systems was evaluated by the area under the ROC curves. Calibration was assessed by Hosmer-Lemeshow statistics and calibration curves. Comparison between areas under the ROC curves of both systems was undertaken. Results There were 218 in-hospital deaths (34.1%). SMR for SAPS II = 0.83 (0.74 – 0.92), MPM24 II = 0.89 (0.78–1.01). Both systems exhibited excellent discrimination (area under ROC curve was 0.88 for SAPS II and 0.91 for MPM24II). The discrimination of MPM24 II was better than SAPS II (p = 0.007). MPM24 II showed good calibration (Hosmer-Lemeshow statistics 14.45, p=0.07), while SAPS II was poorly calibrated (Hosmer-Lemeshow statistics 20.65, p = 0.008). Conclusions SAPS II showed excellent discrimination, but poor calibration. MPM 24 II provided better discrimination than SAPS II, and showed good calibration. The local performance of MPM24 II in addition to its easier use suggested that it might be possible to use this scoring system for mortality prediction in Thai ICUs.