To validate the TRISS method as an audit system on a group of patients with severe trauma admitted to an Italian general intensive care unit (ICU).Prospective, cohort study of consecutive admissions to the ICU.A 6-bed general ICU in a 500-bed general hospital.190 patients with severe trauma admitted from January 1992 to December 1993 were considered eligible. Patients lacking the data necessary to calculate the TRISS probability of survival, or for whom the ultimate outcome was unknown, were excluded. 162 patients were included in the study.None.Vital status at discharge from the last hospital that admitted the patient for the trauma being considered.The Hosmer-Lemeshow goodness-of-fit tests were: H = 16.9, df = 10, p = 0.076; C = 5.8, df = 10, p = 0.831; H 3.5, df = 3, p = 0.31. The area under the receiver operating characteristic curve was 0.963 (SE +/- 0.019). Classification measures at a decision criterion of 0.5 were: sensitivity 0.857, specificity 0.964, positive predictive value 0.782, negative predictive value 0.978, total correct classification 0.950, and the Youden index 0.821. The positive likelihood ratio (LHR) was 24.17, whereas the negative LHR was 0.14.The results of the validation of the TRISS method showed adequate calibration and high discriminatory power in Italian ICU trauma patients also, allowing confidence in the use of this method as an audit tool in our ICU. Some caution is advisable in extending these results to patients with operable intracranial injuries, due to the relatively low number of such cases included in the study.