Estimates of the effect of blood alcohol on injury severity obtained through studies of patients at clinical facilities have been found to be quite different from those estimated through the analysis of highway accident data. Several factors have been suggested to account for these differences, including consideration of the relative frequencies of alcohol involved and alcohol free drivers who were killed at the scene or who died before reaching a treatment facility. More generally, the nature of the way that the effect of alcohol varies as a function of time from injury may play an important role. In this paper information from highway accident data is combined with time of death information from state medical examiner files to yield estimates of the function P(T), the probability that an injured driver dies at time T or later from an injury occurring at T = O, for classes of crash involved drivers with differing blood alcohol concentrations. Other factors also considered in the analyses include measures of crash severity, driver age, and restraint use.