Definition of the objective threshold of pancreatoduodenectomy with nationwide data systems

Background: This study aimed to define an objective evidence-based threshold of high-volume hospitals (HVHs) for pancreatoduodenectomy (PD) using nationwide data systems. Methods: A total of 36,453 patients underwent PD in 1,499 hospitals from 2012 to 2015 were collected from the National Clinical Database in Japan. Restricted cubic spline model with risk adjustment was used for definition of an objective evidence-based threshold of HVHs. Results: The restricted cubic spline curve of 30-day and in-hospital mortality showed a continuous decrease with an increase in hospital volume and plateau phase of mortality was detected between approximately 30 and 50 PDs/year. On the basis of this curve, we defined hospitals ≥30 PDs/year as HVHs and ≤29 PDs/year as non-HVHs. We also sub-classified hospitals <5, 5–29, 30–49, and ≥50 PDs/year as low-volume, intermediate-volume, high-volume, and very high-volume hospitals using the spline curve. The odds ratio (OR) of risk-adjusted mortality decreased as hospital volume increased, with an OR of 0.34 for HVHs and 0.26 for very HVHs compared with low-volume hospitals. Conclusions: We consider that this concept is applicable to other high-risk procedures for reducing mortality after these procedures, which could improve medical care and health services.