Injury severity scores: a review of current designs and the development of a new objectively based score.

The present array of injury severity scores was reviewed for their value in aiding clinical decisions concerning trauma victims. Four scores, the Glasgow Coma Score, the Trauma Score, the CRMS Scale, and the Injury Severity Score (ISS), were evaluate using LDS Hospital's Trauma Registry. The Trauma Registry is a computer database containing information on 1824 trauma patients. The component variables (i.e., blood pressure, respiratory expansion, etc.) used by the four scores, along with additional information routinely obtainable within 24 hours of hospital admission were screened to determine which had the greatest correlation with patient mortality. The screening was performed using a two step process. Initially all 39 of the component variables were subjected to a forward stepping discriminant analysis using BMDP 7M. The 15 variables selected by this first method were than further screened by forward stepping maximum likelihood, an integral part of BMDP LR. The final 5 variables were then used in two different modeling approaches, sequential Bayes and logistic regression, to create a new objectively derived injury severity algorithm. The results obtained by the logistic model were found to superior to the Baysian approach, yelling a sensitivity of 0.66, specificity of 0.99, misclassification rate of 0.02, and a positive predictive value of 0.84.