Comparison between TRISS and ASCOT methods in controlling for injury severity

Abstract

ASCOT was developed by Champion et al. to address known limitations to TRISS. The present research attempted to validate ASCOT using an independent trauma registry. Data were collected by the Institute for Trauma and Emergency Care (ITEC), New York Medical College, between July 1, 1987 and June 30, 1989; 5685 trauma patients admitted to three level I trauma centers or five non-trauma center hospitals were included. Information was gathered by trained nurse-abstractors using all available prehospital and hospital records. ASCOT and TRISS were compared using sensitivity, disparity, misclassification rates, and the Hosmer-Lemeshow goodness-of-fit statistics. Disparity and sensitivity rates were relatively low for both indexes, particularly among blunt injury patients. Total numbers of patients misclassified by TRISS and by ASCOT were similar; most misclassifications were made by both TRISS and ASCOT and involved nonsurvivors. Each method had advantages in predicting the outcomes of particular subgroups of patients; ASCOT with regard to predicting outcomes among patients with head injuries and in correctly classifying blunt injured patients; TRISS in correctly classifying survivors. We conclude (1) the relatively small gain in predictive accuracy by ASCOT over TRISS is largely offset by its complexity and increased computer processing requirements; (2) Hosmer-Lemeshow tests indicate that neither index provides good statistical agreement between predicted and actual outcomes among either blunt or penetrating injury patients. Future models should include additional variables, stratify patients by several injury causes, and use decision rules to select variables and variable weights.