Identifying the Optimal case-volume threshold for pancreatectomy in contemporary practice.

Abstract Background The volume-mortality association led to regionalization recommendations for pancreatic surgery. Mortality following pancreatectomy has declined, but case-volume thresholds remain unchanged. Methods Patients undergoing pancreatectomy from 2004 to 2013 were identified in the National Cancer Database (NCDB). Hospitals were divided into low (LV), medium (MV), and high-volume (HV) strata using 30-day mortality quartiles and logistic regression with cubic splines. Adjusted absolute difference and odds of 30-day mortality between strata were calculated. Results Annual volumes for LV, MV, and HV were  18 cases using quartiles and  18 using cubic splines. Absolute 30-day mortality trended downwards, with differential improvements for MV and LV. Benchmark 30-day mortality for hospitals with >18 cases was 2.8%. For this benchmark, the case-volume threshold decreased from 31 in 2004 to 6 in 2013. Conclusion Differential improvement in 30-day mortality at LV and MV hospitals led to similar 30-day mortality odds at MV and HV hospitals by 2013.