Re-defining a high volume center for pancreaticoduodenectomy

Background: The purpose of this study was to re-evaluate the previously utilized definitions of high volume center for pancreaticoduodenectomy to determine/establish an objective, evidence based threshold of hospital volume associated with improvement in perioperative mortality. Methods: Patients who underwent pancreaticoduodenectomy were identified using the National Cancer Database from 2004 to 2015. The relationship between hospital volume and 90-day mortality was assessed using a logistic regression model. Receiver Operator Characteristic analysis was performed and Youden's statistic was utilized to calculate the optimal cut offs. Results: 42,402 patients underwent elective Pancreaticoduodenectomy at 1238 unique hospitals. A logistic regression was performed which showed a significant inverse linear association between institutional volume and overall 90 day mortality. The maximum improvement in 90 day mortality is seen if the average annual hospital volume was greater than 9 (OR = 0.647 (0.595–0.702), p < 0.0001). When analysis is limited to hospitals that performed >9 cases per year, the maximum improvement in 90 day mortality was noticed at 36 cases per year (OR = 0.458 (0.399–0.525), p < 0.0001). Conclusions: Based on our results, we recommend defining low, medium, and high volume centers for pancreaticoduodenectomy as hospitals with average annual volume less than 9, 9 to 35, and more than 35 cases per year, respectively.