We studied the accident department management and survival of all seriously injured patients brought to the Kent and Canterbury hospital over one year. The physiological state on arrival and the injuries of each patient were coded and probabilities of survival calculated by reference to the Trauma Score and Injury Severity Score (the TRISS method). This enabled a broad comparison with North American results but also drew attention to cases in which survival and prediction were inconsistent. We discuss the worth of the method as a tool for local audit of emergency care.