The quality of a trauma system can be assessed by the rate of preventable deaths. A random selected sample of 110 trauma patients was examined using both clinical and autopsy data. The assessors were asked the following question: If this patient had sustained the accident in front of the hospital in a normal working day, might death have been prevented? Death was found to be unavoidable in 61 cases; in 25 cases death was classified potentially preventable; 11 cases were classified as clearly preventable death. The main failures of treatment were identified as errors and delays during the first phases of in-hospital assessment and care. An improvement in the pre-hospital phase will be almost useless if the quality of the definitive in-hospital management is not addressed.