Specifications for calculation of risk-adjusted odds of death using trauma registry data

Logistic regression models, with coefficients developed from normative populations, can be applied to a trauma registry cohort to predict the risk-adjusted frequency of death. Quality of care is judged based on differences between predicted and observed mortality frequency. The goal of these analyses was to determine if decedents who died in the emergency department had independent variables associated with risk of death identical to those who died after hospital admission. This case-control study is based upon decedents in a trauma registry matched to survivors. Backward stepwise linear logistic regression models contained independent variables selected to reflect patients' status before treatment. Beta coefficients and independent variables selected for models of expired emergency department patients were different from those of hospital death patients. To achieve a more precise determination of risk-adjusted mortality for injured patients at a trauma center, two separate analyses are appropriate: death in emergency department and death after hospital admission.