Factors associated with mortality in trauma: re-evaluation of the TRISS method using the National Trauma Data Bank

摘要: Background: TRISS remains a standard method for predicting survival and correcting severity in outcome analysis. The National Trauma Data Bank (NTDB) is emerging as major source of trauma data that will be used both primary research bench-marking. We NTDB data, to determine whether still an accurate predictor coefficients the ability predict could improved by updating or building predictive models include information on co-morbidities. Methods: To compare utility different methods calculation we identified records 72,517 patients (62,103 blunt 10,414 penetrating trauma) who had complete all covariates considered Multiple logistic regression was recalculate using original which also included variables co-morbidities potentially affect survival. Model discrimination evaluated calculating area under receiver operating characteristic curves (AUC), model calibration with Hosmer-Lemeshow Goodness-of-Fit Statistic (H-L). Results: For equation good discriminative (AUC=0.98), but poorly calibrated (H-L=267.04). When generate revised coefficients, unchanged, (H-L=38.66). only co-morbid factor significantly associated after acute alcohol consumption, increased (p < 0.0001). However, its inclusion did not improve discrimination, somewhat (AUC = 0.98; H-L=19.95). less 0.84; H-L= 10,720.7). covariates, predictions 0.94; H-L=25.45). With trauma, consumption prior hypertension were survival, male gender, congestive failure, cirrhosis, myocardial infarction decreased these H-L=34.83). Conclusions: In traditional limited trauma. Accuracy prediction recalculating further improvements seen about