Comparison of Predicted Outcomes of Trauma and Injury Severity Score (TRISS), Acute Physiology and Chronic Health Evaluation (APACHE) II, and Simplified Acute Physiology Score (SAPS) II Scoring Systems in Intensive-care-unit Trauma Patients.
- Author:
Kwang Won CHO
1
;
Seong Youn HWANG
;
Eun Seok HONG
Author Information
1. Department of Emergency Medicine, Sungkyunkwan University School of Medicine, Masan Samsung Hospital, Masan, Korea. syellow@unitel.co.kr
- Publication Type:Original Article
- Keywords:
Injury serverity score;
APACHE II;
Intensive care units
- MeSH:
APACHE*;
Calibration;
Emergencies;
Hospital Mortality;
Humans;
Injury Severity Score*;
Intensive Care Units;
Medical Records;
Models, Statistical;
Mortality;
Physiology*;
Retrospective Studies;
ROC Curve
- From:Journal of the Korean Society of Emergency Medicine
2002;13(4):434-443
- CountryRepublic of Korea
- Language:Korean
-
Abstract:
PURPOSE: Several statistical models, such as the TRISS, the APACHE II and the SAPS II scoring systems, have been utilized over the recent decades to accurately predict outcomes in Intensive Care Unit (ICU) trauma patients. This study was performed to evaluate the ability of these three statistical models to predict hospital mortality and to compare the performance of these three statistical models in ICU trauma patients. METHODS: Seven hundred forty-seven trauma patients were admitted to the ICU via the emergency center, Masan Samsung Hospital, from March 1999 to February 2001. Of them, 684 patients were included in this study, and their medical records were retrospectively analyzed. The probability of death was calculated for each patient based on the TRISS, the APACHE II, and the SAPS II equations. RESULTS: The values of the Hosmer and Lemeshow X2 for TRISS, APACHE II, and SAPS II were 41.32, 43.03, and 17.64, respectively, and all of them underestimated mortality (p<0.05). For two-by-two decision matrices with a decision criterion of 0.5, the specificities and percentages correctly classified of APACHE II and SAPS II were higher than those of TRISS (p<0.001). For the ROC curve analysis, the areas under the curves (+/-SEM) of TRISS, APACHE II, and SAPS II were 0.922 +/- 0.011, 0.951 +/- 0.011, and 0.957 +/- 0.009, respectively (95% confidence interval). The areas under the curves of APACHE II and SAPS II were larger than that of TRISS (p<0.05 and p<0.01, respecively). CONCLUSION: All of these three statistical models had good discriminative power, with APACHE II and SAPS II performing better than TRISS. However, all of them showed poor calibration and underestimated mortality. The authors conclude that a new statistical model is needed to accurately predict hospital mortality in severely injured patients.
