A predictive model of early mortality in trauma patients

Abstract Background Rapid thrombelastography (rTEG) is a real-time whole-blood viscoelastic coagulation assay. We hypothesized that admission rTEG and clinical data are independent predictors of trauma-related mortality. Methods Prospective observational data (patient demographics, admission vital signs, laboratory studies, and injury characteristics) from trauma patients enrolled within 6 hours of injury were collected. Mann–Whitney U test and analysis of variance test assessed significance ( P ≤ .05). Logistic regression analyses determined the association of the studied variables with 24-hour mortality. Results Seven hundred ninety-five trauma patients were enrolled, of which 55 died within 24 hours of admission. Admission variables which independently predicted 24-hour mortality were as follows: Glasgow Coma Scale ≤8, hemoglobin 1.5, Ly30 >8%, and penetrating injury ( P < .05). This 5-variable model's area under the receiver operator characteristic curve was .88. The Hosmer–Lemeshow goodness-of-fit test was .90. Conclusions This 5-variable model provides a rapid prediction of 24-hour mortality. The inclusion of rTEG Ly30 demonstrates the association of fibrinolysis with outcome and may support the early use of antifibrinolytic therapies.