Background: Operative mortality traditionally has been defined as the rate within 30 days or during the initial hospitalization, and studies that established the volume-outcome relationship for pancreatectomy used similar definitions.
Methods: Pancreatectomies reported to the National Cancer Data Base (NCDB) during 2007-2010 were examined for 30- and 90-day mortality. Unadjusted mortality rates were compared by type of resection, stage, comorbidities, and average annual hospital volume. Hierarchical logistic regression models generated risk-adjusted odds ratios for 30- and 90-day mortality.
Results: After 21,482 pancreatectomies, the unadjusted 30-day mortality rate was 3.7 % (95 % confidence interval [CI] 3.4-3.9 %), which doubled at 90 days to 7.4 % (95 % CI 7.0-7.8). The unadjusted and risk-adjusted mortality rates were higher at 30 days with increasing age, increasing stage, male gender, lower income, low hospital volume, resections other than distal pancreatectomy, Medicare or Medicaid insurance coverage, residence in a Southern census division, history of prior cancer, and multiple comorbidities. The lowest-volume hospitals (<5 per year) performed 19 % of the pancreatectomies, with a risk-adjusted odds ratios for mortality that were 4.2 times higher (95 % CI 3.1-5.8) at 30 days and remained 1.9 times higher (95 % CI 1.5-2.3) at 30-90 days compared with hospitals that had high volumes (≥40 per year).
Conclusion: Mortality rates within 90 days after pancreatic resection are double those at 30 days. The volume-outcome relationship persists in the NCDB. Reporting mortality rates 90 days after pancreatectomy is important. Hospitals should be aware of their annual volume and mortality rates 30 and 90 days after pancreatectomy and should benchmark the use of high-volume hospitals.