250 SURGERY FEW ELECTIVE SURGICAL PROCEDURES are associated with higher morbidity and mortality rates than pan- creaticoduodenectomy (the Whipple procedure). Despite some improvements during the past 2 decades, mortality rates exceeding 10% remain common. 1 Pancreaticoduodenectomy is most fre- quently performed for pancreatic cancer. Of the 29,000 patients diagnosed with this disease each year in the United States, 2 approximately 10% to 20% have localized resectable disease. Although surgery remains the only potentially curative thera- py, 5-year survival rates after pancreaticoduodenec- tomy for pancreatic cancer are low, ranging from 10% to 25% in various studies. 3-6 Given this narrow therapeutic margin, minimizing surgery-related morbidity and mortality is particularly important. Regionalizing pancreaticoduodenectomies to high-volume hospitals may be an important way to Effect of hospital volume on in-hospital mortality with pancreaticoduodenectomy John D. Birkmeyer, MD, Samuel R. G. Finlayson, MD, MPH, Anna N. A. Tosteson, ScD, Sandra M. Sharp, MS, Andrew L. Warshaw, MD, and Elliott S. Fisher, MD, MPH, Hanover, NH, White River Junction, Vt, and Boston, Mass Background. Reports of better results at national referral centers than at low-volume community hospi- tals have prompted calls for regionalizing pancreaticoduodenectomy (the Whipple procedure). We exam- ined the relationship between hospital volume and mortality with this procedure across all US hospitals. Methods. Using information from the Medicare claims database, we performed a national cohort study of 7229 Medicare patients more than 65 years old undergoing pancreaticoduodenectomy between 1992 and 1995. We divided the study population into approximate quartiles according to the hospital’s aver- age annual volume of pancreaticoduodenectomies in Medicare patients: very low (<1/ y), low (1-2/ y), medium (2-5/ y), and high (5+/ y). Using multivariate logistic regression to account for potentially con- founding patient characteristics, we examined the association between institutional volume and in-hos- pital mortality, our primary outcome measure. Results. More than 50% of Medicare patients undergoing pancreaticoduodenectomy received care at hos- pitals performing fewer than 2 such procedures per year. In-hospital mortality rates at these low- and very-low-volume hospitals were 3- to 4-fold higher than at high-volume hospitals (12% and 16% , respectively, vs 4% , P < .001). Within the high-volume quartile, the 10 hospitals with the nation’s highest volumes had lower mortality rates than the remaining high-volume centers (2.1% vs 6.2% , P < .01). The strong association between institutional volume and mortality could not be attributed to patient case-mix differences or referral bias. Conclusions. Although volume-outcome relationships have been reported for many complex surgical pro- cedures, hospital experience is particularly important with pancreaticoduodenectomy. Patients consider- ing this procedure should be given the option of care at a high-volume referral center. (Surgery 1999;125:250-6.) From the Center for the Evaluative Clinical Sciences and the Departments of Medicine and Surgery, Dartmouth Medical School, Hanover, NH, the Veterans Administration Outcomes Group, Department of Veterans Affairs Medical Center, White River Junction, Vt, and the Department of Surgery, Massachusetts General Hospital, Boston, Mass Supported in part by a grant from the Robert Wood Johnson Foundation. Accepted for publication Oct 15, 1998. The views expressed herein do not necessarily reflect the views of the Health Care Financing Administration, the Department of Veterans Affairs, or the United States government. Reprint requests: John D. Birkmeyer, MD, VA Outcomes Group (111B), Department of Veterans Affairs Medical Center, White River Junction, VT 05009. Copyright © 1999 by Mosby, Inc. 0039-6060/ 99/ $8.00 + 0 11/ 56/ 95211 Surgical outcomes research