Volume-outcome relationship in pancreatic surgery.

BACKGROUND Volume-outcome relationships related to major surgery may be of limited value if observation ends at the point of discharge without taking transfers and later events into consideration. METHODS The volume-outcome relationship in patients who underwent pancreatic surgery between 2008 and 2010 was assessed using claims data for all inpatient episodes from Germany's largest provider of statutory health insurance covering about 30 per cent of the population. Multiple logistic regression models with random effects were used to analyse the effect of hospital volume (using volume quintiles) on 1-year mortality, adjusting for age, sex, primary disease, type of surgery and co-morbidities. Additional outcomes were in-hospital (including transfer to other hospitals until final discharge) and 90-day mortality. RESULTS Of 9566 patients identified, risk-adjusted 1-year mortality was significantly higher in the three lowest-volume quintiles compared with the highest-volume quintile (odds ratio 1·73, 1·53 and 1·37 respectively). A similar, but less pronounced, effect was demonstrated for in-hospital and 90-day mortality. The effect of hospital volume on 1-year mortality was comparable to the effect of co-morbid conditions such as renal failure. CONCLUSION Although mortality related to pancreatic surgery is influenced by many factors, this study demonstrated lower mortality at 1 year in high-volume centres in Germany.