BACKGROUND: Death during acute care hospitalisation is commonly used as a principal outcome indicator in injury research. This endpoint excludes post-hospital trauma-related deaths, which are substantial according to recent US studies. Two additional ways of defining outcome in trauma victims are also used; by end of somatic care, and at 30 days after injury. Our primary aim was to analyse how the different definitions of trauma outcome influence performance analyses. Secondly, we wanted to evaluate whether 30 days mortality after injury, which is widely used in other parts of biomedicine and recommended by the United Nations for use in transport statistics, is a suitable endpoint in trauma research. MATERIALS AND METHODS: We conducted a retrospective analysis of prospectively collected data from the hospital based trauma registry at Ulleval University Hospital (UUH) in Oslo, Norway. Outcome measure was mortality at discharge from UUH, i.e., by "end of acute care", at end of somatic care defined as discharge from final acute care hospital, and at 30 days after injury. Analyses were performed according to conventional TRISS methodology. RESULTS: 3332 of 3446 patients from the years 2000-2004 were included. Of these, 323 (9.7%) died within 30 days of injury or during somatic care more than 30 days after injury. Mortality varied with outcome definition, with 264 deaths (81.7% of total deaths) before discharge from UUH, 318 (98.4%) before end of somatic care, and 308 (95.4%) within 30 days after injury. TRISS-based trauma system performance evaluation for blunt trauma showed significantly better outcome than predicted with discharge from UUH as outcome definition, whereas for category 30 days after injury, performance was clearly less favourable. Performance for penetrating trauma was not affected, since all deaths occurred before 30 days, and nearly all before discharge from UUH. CONCLUSIONS: A substantial number of in-hospital deaths following blunt trauma occurs after discharge from the primary institution, i.e., unnoticed when "end of acute care" is used as outcome definition. Consequently, outcome definition influenced performance when comparing our institution to an acknowledged standard. We recommend mortality occurring within 30 days of injury as endpoint in trauma research.