BACKGROUND: Previous studies have suggested that there may be long-term harm associated with postoperative complications. Uncertainty exists however, because of the need for risk adjustment and inconsistent definitions of postoperative morbidity. METHODS: We did a longitudinal observational cohort study of patients undergoing major surgery. Case-mix adjustment was applied and morbidity was recorded using a validated outcome measure. Cox proportional hazards modelling using time-dependent covariates was used to measure the independent relationship between prolonged postoperative morbidity and longer term survival. RESULTS: Data were analysed for 1362 patients. The median length of stay was 9 days and the median follow-up time was 6.5 yr. Independent of perioperative risk, postoperative neurological morbidity (prevalence 2.9%) was associated with a relative hazard for long-term mortality of 2.00 [P=0.001; 95% confidence interval (CI) 1.32-3.04]. Prolonged postoperative morbidity (prevalence 15.6%) conferred a relative hazard for death in the first 12 months after surgery of 3.51 (P<0.001; 95% CI 2.28-5.42) and for the next 2 yr of 2.44 (P<0.001; 95% CI 1.62-3.65), returning to baseline thereafter. CONCLUSIONS: Prolonged morbidity after surgery is associated with a risk of premature death for a longer duration than perhaps is commonly thought; however, this risk falls with time. We suggest that prolonged postoperative morbidity measured in this way may be a valid indicator of the quality of surgical healthcare. Our findings reinforce the importance of research and quality improvement initiatives aimed at reducing the duration and severity of postoperative complications.