In search of benchmarking for mortality following multiple trauma : a Swiss trauma center experience

BACKGROUND: The manifestations associated with non-survival after multiple trauma may vary importantly between countries and institutions. The aim of the present study was to assess the quality of performance by comparing actual mortality rates to the literature. METHODS: The study involved evaluation of a prospective consecutive multiple trauma cohort (injury severity score, ISS < 16) primarily admitted to a university hospital. Univariate and multivariate testing of routine parameters and scores, such as the Trauma and Injury Severity Score (TRISS), was used to determine their predictive powers for mortality. RESULTS: The 30-day mortality of 22.8% (n = 54) exactly matched predicted TRISS versions of Champion or the Major Trauma Outcome Study for our 237 multiple trauma patients (42.8 +/- 20.9 years; ISS 29.5 +/- 11.5). Univariate analysis revealed significant differences between survivors and non-survivors when compared for age, ISS, Glasgow coma scale (GCS), pulse oximeter saturation (SapO2), hemoglobin, prothrombin time, and lactate. In multivariate analysis, age, ISS, and GCS (P > 0.001 each) functioned as major independent prognostic parameters of both 24 h and 30-day mortality. Various TRISS versions hardly differed in their precision (area under the curve [AUC] 0.83-0.84), but they did differ considerably in their level of requirement, with the TRISS using newer National Trauma Data Bank coefficients (NTDB-TRISS) offering the highest target benchmark (predicted mortality 13%, Z value -5.7) in the prediction of 30-day mortality. CONCLUSIONS: Because of the current lack of a single, internationally accepted scoring system for the prediction of mortality after multiple trauma, the comparison of outcomes between medical centers remains unreliable. To achieve effective quality control, a practical benchmarking model, such as the TRISS-NTDB, should be used worldwide.