Background Caseload thresholds for pancreatic surgery are commonly derived by relating outcomes to hospital volume within the same year, a descriptive approach that may overestimate the true volume–outcome relationship because of year-to-year variation in performance. This study evaluated the predictive importance of annual hospital caseload for in-hospital mortality in the subsequent year. Methods In this retrospective population-based study, nationwide diagnosis-related group data from all German hospitals between 2010 and 2019 were analyzed. Adult patients undergoing pancreatic resection for any indication were included. The association between annual hospital caseload and in-hospital mortality was examined using multivariable logistic regression with caseload modeled as a natural cubic spline. Thresholds derived from same-year and previous-year caseloads were compared regarding their independent association with mortality. Results A total of 96,270 patients from 761 hospitals (54.7% males; mean age, 65.4 years). Malignant neoplasm was the most common diagnosis (56.2%), and pancreaticoduodenectomy was the most frequent procedure (58.6%). The overall in-hospital mortality rate was 9.0%. Threshold analysis identified 54 annual resections for same-year caseload and 33 for previous-year caseload. Hospitals performing >54 same-year resections had lower adjusted odds of mortality (OR = 0.73, 95%-CI: 0.63–0.84). Similarly, hospitals performing >33 procedures in the previous year demonstrated reduced mortality (OR = 0.70, 95%-CI: 0.63–0.78). Conclusions Same-year caseload thresholds may overestimate the volume-outcome association. Previous-year caseload provides a more robust and realistic predictor for future surgical outcomes and may represent a more appropriate basis for defining evidence-based caseload thresholds for complex pancreatic surgery.
