Association between Oncologist Ownership of Imaging Facilities and Utilization: An Analysis of Positron-Emission Tomography (PET) Use in Lymphoma Settings with Variable PET Utility

Background: Advanced imaging during cancer care is a target for cost reduction, with 6 of 20 Choosing Wisely recommendations published by leading hematology and medical oncology societies focused on reducing low value imaging. Prior research has found self-referral practices related to clinician ownership of imaging equipment to be associated with greater use of imaging with low clinical utility. However, it is unclear whether ownership also influences utilization of appropriate imaging. We examined and compared the association between oncologist PET ownership and the use of pre-treatment staging PET in three non-Hodgkin lymphoma (NHL) settings: diffuse large B cell lymphoma (DLBCL) where consensus guidelines recommend PET; follicular lymphoma (FL) where recommendations are equivocal; and chronic lymphocytic leukemia (CLL) where guidelines recommend against routine use. Methods: We conducted a population-based cohort study using Surveillance, Epidemiology, and End Results (SEER) -Medicare data. Individuals 66+ years old newly diagnosed with DLBCL, FL, or CLL who initiated lymphoma-directed therapy in the community setting between 2004 and 2013 were included. Medicare claims for immunochemotherapy were used to assign each Medicare beneficiary to their treating oncologist. We used a 12-month lookback from each treatment initiation to categorize PET ownership status for each oncologist. Consistent with prior research, oncologists were classified as PET owners if they were listed as billing providers under the technical component or global claim for PET scans. PET claims for all beneficiaries diagnosed with lymphoma in SEER-Medicare were used to identify ownership status. Our primary outcome was receipt of a staging PET scan, defined as a claim for PET within 60 days prior to initiating first-line therapy. In addition to descriptive statistics, we used three separate linear probability models to estimate differences in receipt of a staging PET by ownership status across our NHL subtypes. Our hierarchical multivariable models adjusted for correlation at the provider level and included year of treatment, patient sociodemographic characteristics, lymphoma stage, Elixhauser comorbidity index, and disability status. Results: We identified 9,968 Medicare beneficiaries with DLBCL (n = 5006), FL (n = 2433), or CLL (n = 2529). A total of 2,034 community physicians were identified as their treating oncologists, with 237 (11.7%) physicians having Medicare claims indicating PET ownership. In bivariate analysis, NHL patients (n = 737, 7.4%) who initiated therapy with an oncologist with PET ownership were more likely to receive a staging PET regardless of lymphoma subtype (DLBCL 73.3% vs 65.9%, p = 0.005; FL 71.3% vs 53.2%, p < 0.001; CLL 29.4% vs 14.8%, p < 0.001). In adjusted analysis, PET ownership continued to be associated with higher probability of receiving a staging PET in all three settings. The probability of receiving PET was 17% higher (95% confidence interval [CI]: 8% - 22%, p < 0.001) for PET owners compared to non-owners in the setting of FL, where consensus guidelines were equivocal; and 15% higher (95% CI: 9% - 25%, p <0.001) in the setting of CLL, where consensus guidelines recommend against PET. However, the difference between owners and non-owners was smaller (7%, 95% CI: 3% - 12%, p = 0.003) in the high clinical utility setting of DLBCL, where consensus guidelines recommend use of PET. Conclusion: PET ownership was associated with higher utilization of PET across a spectrum of clinical utility, but the effects were largest in low utility or equivocal settings. Removing financial incentives related to imaging self-referral may reduce utilization of PET scans, but attention to high cost - high utility settings (i.e. PET staging for DLBCL) may be necessary to mitigate unintended consequences.