New Trauma and Injury Severity Score (TRISS) adjustments for survival prediction

BACKGROUND: The objective of this study is to propose three new adjustments to the Trauma and Injury Severity Score (TRISS) equation and compare their performances with the original TRISS as well as this index with coefficients adjusted for the study population. METHODS: This multicenter, retrospective study evaluated trauma victims admitted to two hospitals in São Paulo-Brazil and San Diego-EUA between January 1st, 2006, and December 31st, 2010. The proposed models included a New Trauma and Injury Severity Score (NTRISS)-like model that included Best Motor Response (BMR), systolic blood pressure (SBP), New Injury Severity Score (NISS), and age variables; a TRISS peripheral oxygen saturation (SpO(2)) model that included Glasgow Coma Scale (GCS), SBP, SpO(2), Injury Severity Score, and age variables; and a NTRISS-like SpO(2) model that included BMR, SBP, SpO(2), NISS, and age variables. All equations were adjusted for blunt and penetrating trauma coefficients. The model coefficients were established by logistic regression analysis. Receiver operating characteristic (ROC) curve analysis was used to evaluate the performance of the models. RESULTS: The original TRISS (area under the curve (AUC) = 0.90), TRISS with adjusted coefficients (AUC = 0.89), and the new proposals (NTRISS-like, TRISS SpO(2), and NTRISS-like SpO(2)) showed no difference in performance (AUC = 0.89, 0.89, and 0.90, respectively). CONCLUSIONS: The new models demonstrated good accuracy and similar performance to the original TRISS and TRISS adjusted for coefficients in the study population; therefore, the new proposals may be useful for the assessments of quality of care in trauma patients using variables that are routinely measured and recorded.