OBJECTIVE: To evaluate the efficiency (sensitivity, specificity, positive predictive value, overtriage, and undertriage) of activation of the trauma team in a Norwegian trauma referral centre. DESIGN: A cohort study with univariate and multivariate analysis. SETTING: A primary trauma hospital and trauma referral centre, Norway. SUBJECTS: 3391 injured patients admitted during a 12 months period, starting January 15th, 1996. MAIN OUTCOME MEASURES: Activation of the trauma team for severely injured patients and factors associated with correct activation. RESULTS: Of the 3383 injured patients admitted, 283 (8%) were classified as severely injured. Of 507 activations of the trauma team, 240 (47%) were for severely injured patients (sensitivity 85%, undertriage 15%, specificity 91%, overtriage 9%, positive predictive value 0.47). The system of activation was significantly more efficient for patients admitted by anaesthetist-manned ambulances than by ordinary ground ambulances (sensitivity 94% compared with 83%, corresponding positive predictive value 0.55 and 0.33, p