Data were obtained prospectively for 7,912 head-injured patients admitted from 1980 to 1981 to 41 hospitals in 3 U.S. metropolitan areas. Highly significant independent predictors of mortality were motor score, number of reactive eyes, systolic blood pressure, abdominal injury, chest injury, age, and hospital unit (hospital or group of hospitals). The difference between the number of deaths observed and the number expected, by hospital unit, ranged from 43% below expectation to 52% above expectation. The 2 hospital units with the lowest mortality had only 1 death, where 9.4 would be expected, in patients whose risk of death was estimated to be less than 10%. In the single hospital with clearly excess mortality, however, the standardized mortality ratios were 208, 135, and 144% for the risk groups whose probabilities of death were 0 to 50, 50 to 80, and 80 to 100%, respectively. The lack of deaths in those hospitals with the best mortality experience and the excess mortality in the hospital with the worst mortality experience appear to be explained largely by differences in patients who, by clinical criteria, seem to be at low risk for death, and not by severely injured patients.