To evaluate the incidence and the causes of early intra- and postoperative deaths in a multidisciplinary hospital.Retrospective survey.All patients receiving an anaesthetic between 1992 and 1995.Analysis of all deaths occurring during anaesthesia and in the subsequent 24 hours. Demographic data (age, gender) and medical data (ASA physical class, type of surgery and degree of emergency) were recorded. The contribution of anaesthesia, surgery or patient disease to fatal outcome was analysed.The analysis included 52,654 patients who underwent either general anaesthesia or epidural analgesia. Perioperative mortality (n = 170) was 1/310 patients (0.32%). The risk factors for mortality (multivariate analysis) were: age > 64 years (odds-ratio [OR] 4.8), ASA class > or = 3 (OR 16.6), emergency surgery (OR 3.6), duration of surgery > 115 minutes (OR 37.4). Fifty percent of deaths (95% confidence interval [CI] = 42-58) were related to patient's underlying diseases and 29% to surgery (CI 95% = 22-36). The percentage of deaths linked to anaesthesia was 17.6% (CI 95% = 11.9-23.3, 1/1,755), consisting of 8.2% (CI 95% = 4.1-9.3, 1/3,761) totally due to anaesthesia and 9.4% (CI 95% = 5-13.8, 1/3,291) only partially. The main aetiologies of the deaths linked to anaesthesia were a mismanagement of severe haemorrhages (30%), respiratory complications (23%) or cardiac complications (23%). The mismanagement of an intraoperative critical situation (46%) and a mistake in the post-operative care (33%) were the main causes.In this survey, mortality due to anaesthesia was higher than the rates reported in other studies. Human error remained the main cause.