Since anaesthesia, unlike medical or surgical specialties, does not constitute treatment, this study sought to determine if methods used to assess medical or surgical outcomes (that is the determination of adverse outcome) are applicable to anaesthesia. Anaesthetists collected information on patient, surgical and anaesthetic factors while data on recovery room and postoperative events were evaluated by research nurses. Data on 27,184 inpatients were collected and the analysis of outcomes determined for the intraoperative, post-anaesthetic care unit and postoperative time periods. Logistic regression was used to control for differences in patient populations across the four hospitals. In addition, a random selection of 115 major events was classified by a panel of anaesthetists into anaesthesia, surgical and patient-disease contributions. Across the three time periods, large variations in minor outcomes were found across the four hospitals; these variations ranged from two- to five-fold after case-mix adjustment (age, physical status, sex, emergency versus elective and length of anaesthesia). The rates of major events and deaths were similar across three hospitals; one hospital had a lower mortality rate (P less than 0.001) but had a higher rate of all major events (P less than 0.0001). Of major events assessed by physician panels, 18.3% had some anaesthetic involvement and no deaths were attributable partially or wholly to anaesthesia. Possible reasons to account for these variations in outcome include compliance in recording events, inadequate case-mix adjustment, differences in interpretation of the variables (despite guidelines) and institutional differences in monitoring, charting and observation protocols. The authors conclude that measuring quality of care in anaesthesia by comparing major outcomes is unsatisfactory since the contribution of anaesthesia to perioperative outcomes is uncertain and that variations may be explained by institutional differences which are beyond the control of the anaesthetist. It is suggested that minor adverse events, particularly those of concern to the patient, should be the next focus for quality improvement in anaesthesia.