The importance of admission physiological and biochemical variables was modeled on data from 185 patients with blunt liver trauma with regard to their significance in prediction of mortality. The variables used were admission Glasgow Coma Score, base excess (or deficit), arterial lactate, Injury Severity Score, and initial 24-hour volume of blood required for replacement. Each variable was modeled as a predictor of survival alone and in combination, using a linear logistic model. In any two-variable combination, Glasgow Coma Score had a high likelihood ratio for prediction representing the influence of brain injury. But as a single variable reflecting the probability of death, both base excess (LD50 = -11.8 mmol/L) and initial 24-hour volume of blood (LD50 = 5.4 L) were highly significant. A combined logistic model of admission Glasgow Coma Score and base excess had the greatest likelihood of accurate prediction of outcome: P death = e lambda/l + e lambda; where lambda = -0.21(Glasgow Coma Score) -0.147(base excess) + 0.285. Testing of this predictive model on data from 323 additional patients with multiple trauma who had pelvic fracture as their index injury also showed it to be a highly significant early predictor of outcome.