An evaluation of Ontario trauma outcomes and the development of regional norms for Trauma and Injury Severity Score (TRISS) analysis.

Abstract

Outcomes analysis of patient care programs has become increasingly necessary for a variety of reasons in recent years. This has been particularly true for trauma programs. The Trauma and Injury Severity Score (TRISS) methodology was developed for this purpose in the context of the Major Trauma Outcome Study (MTOS). It provides an estimate of the probability of survival for individual patients, based on anatomic, physiologic, and etiologic factors. In addition, it allows hospitals and groups of hospitals to compare survival rates with other hospitals submitting data to the data base. However, the published coefficients for TRISS analysis have been derived from the MTOS data base. Patterns of practice, time to treatment, and other variables may be significantly different in other jurisdictions. To compare outcomes among similar hospitals within the province of Ontario, Canada, a regression analysis was performed to develop TRISS coefficients specific to the province. Data were obtained from the 12 trauma centers in the province treating the most severely injured patients (Injury Severity Score > 12). A total of 3,880 cases were eligible for TRISS analysis, over a 3-year period. Of these, 3,672 were patients with blunt trauma, and 208 were victims of penetrating injury. Standard TRISS analysis of the patients with blunt trauma revealed z scores ranging from -10.260 to +1.849, with a mean of -6.648. Four centers had negative z scores that were significant (an absolute value of > 1.96 is considered statistically significant). Using Ontario TRISS coefficients, z scores ranged from -4.125 to +2.782, with a mean of 0.000. Four scores were significant with the Ontario coefficients, only one of which had been significant using the MTOS norms. The other three z scores were all positive, indicating more deaths than would have been predicted, but they were not significant when compared to the MTOS norms. The mean was also, of course, no longer significant. The area under the receiver operating characteristic curve analysis was strongly positive, and the Hosmer-Lemeshow Goodness-of-Fit analysis indicated good calibration. The new coefficients were subsequently validated by applying them to a subsequent year's data from patient records that did not form part of the original data set. This resulted in slightly improved z scores overall, and in most of the hospitals. This use of regional norms allows comparison with outcomes of patients cared for in hospitals within the same jurisdiction that are more similar to one another than to those in the MTOS, and helps to identify unexpected outcomes and outliers.