ImportanceTreatment at high-volume centers (HVCs) has been associated with improved overall survival (OS) in patients with pancreatic ductal adenocarcinoma (PDAC); however, it is unclear how patterns of referral affect these findings.ObjectiveTo understand the relative contributions of treatment site and selection bias in driving differences in outcomes in patients with PDAC and to characterize socioeconomic factors associated with referral to HVCs.Design, Setting, ParticipantsA population-based retrospective review of the Oregon State Cancer Registry was performed from 1997 to 2019 with a median 4.3 months of follow-up. Study participants were all patients diagnosed with PDAC in Oregon from 1997 to 2018 (n = 8026).ExposuresThe primary exposures studied were diagnosis and treatment at HVCs (20 or more pancreatectomies for PDAC per year), low-volume centers ([LVCs] less than 20 per year), or both.Main Outcomes and MeasuresOS and treatment patterns (eg, receipt of chemotherapy and primary site surgery) were evaluated with Kaplan-Meier analysis and logistic regression, respectively.ResultsEight thousand twenty-six patients (male, 4142 [52%]; mean age, 71 years) were identified (n = 3419 locoregional, n = 4607 metastatic). Patients receiving first-course treatment at a combination of HVCs and LVCs demonstrated improved median OS for locoregional and metastatic disease (16.6 [95% CI, 15.3-17.9] and 6.1 [95% CI, 4.9-7.3] months, respectively) vs patients receiving HVC only (11.5 [95% CI, 10.7-12.3] and 3.9 [95% CI, 3.5-4.3] months, respectively) or LVC-only treatment (8.2 [95% CI, 7.7-8.7] and 2.1 [95% CI, 1.9-2.3] months, respectively; all P < .001). No differences existed in disease burden by volume status of diagnosing institution. When stratifying by site of diagnosis, HVC-associated improvements in median OS were smaller (locoregional: 10.4 [95% CI, 9.5-11.2] vs 9.9 [95% CI, 9.4-10.4] months; P = .03; metastatic: 3.6 vs 2.7 months, P < .001) than when stratifying by the volume status of treating centers, indicating selection bias during referral. A total of 94% (n = 1103) of patients diagnosed at an HVC received HVC treatment vs 18% (n = 985) of LVC diagnoses. Among patients diagnosed at LVCs, later year of diagnosis and higher estimated income were independently associated with higher odds of subsequent HVC treatment, while older age, metastatic disease, and farther distance from HVC were independently associated with lower odds.Conclusions and RelevanceLVC-to-HVC referrals for PDAC experienced improved OS vs HVC- or LVC-only care. While disease-related features prompting referral may partially account for this finding, socioeconomic and geographic disparities in referral worsen OS for disadvantaged patients. Measures to improve access to HVCs are encouraged.