Back to 2008 Program and Abstracts

Improved outcomes after pancreatic resection (PR) by high volume (HV) surgeons have been reported in single center studies which may be confounded with potential selection and referral bias. We attempted to determine if improved outcomes by HV surgeons are reproducible when patient demographic factors are controlled at the population level.
Methods: Using the Nationwide Inpatient Sample (NIS), discharge records for all non-trauma PR (n=3,705) were examined from 1998-2005. Surgeons were divided into two groups: high volume (HV; >=5 operations / year) or low volume (LV; < 5 /year). The Elixhauser index adjusted for patient comorbidity. We created a logistic regression model to examine the relationship between surgeon type and operative mortality while accounting for patient/hospital factors. To eliminate differences in cohorts and determine the true effect of surgeon volume on mortality, case-control groups based on patient demographics were created using propensity scores.
Results: 128 HV and 1,329 LV surgeons performed 3,705 PR in 449 hospitals across 11 states that report surgeon identifier information over the 8-year period. Patients who underwent PR by HV surgeons were more likely to be male, white race, and a resident of a high-income zip code (p < 0.05). HV surgeons had a lower unadjusted mortality compared to LV surgeons (2.5% vs. 6.8% p<0.0001). Significant independent factors for in-hospital mortality after PR included increasing age, male gender, Medicaid insurance and surgery by HV surgeon (Table). Propensity scoring was used to create matched HV and LV groups; when HV surgeons performed PR an in-hospital mortality benefit was found across all groups.
Conclusions: PR by a HV surgeon in this cohort was independently associated with a 60% reduction in in-hospital mortality. Removal of potential selection bias still resulted in improved outcomes after PR by HV surgeons. To our knowledge, this is the first population-based case-controlled evidence that demonstrates improved in-hospital mortality after PR is directly related to surgeon volume.

Logistic Regression of Operative Mortality

Factor	OR (95% CI)
Age	1.1 (1.0-1.1)
Female gender	0.6 (0.4-0.9)
Black race	1.6 (0.9-2.7)
Hispanic race	1.5 (0.9-2.7)
Medicaid insurance	2.4 (1.1-5.5)
Highest income bracket	1.5 (1.0-2.1)
Urgent admission	0.9 (0.6-1.3)
Elixhauser comorbidity	1.1 (0.9-1.2)
Malignant diagnosis	1.4 (0.9-2.2)
High volume surgeon	0.4 (0.3-0.6)

Back to 2008 Program and Abstracts