Article Title

Determinants of Patient Choice of Health Care Providers for Breast Cancer Treatment

Publication Date



patterns of care, breast cancer


Background/Aims: Unless they are constrained by their insurance coverage, patients routinely seen within an integrated delivery system (IDS) have choices as to where they will be treated. If the IDS provides a reasonably broad range of cancer services, a major alternative will be an academic medical center (AMC). We hypothesize that among breast cancer patients initially diagnosed in a community-based IDS, the likelihood of seeking cancer-related care in a nearby AMC will be related to: a) cancer severity, b) other comorbidities, c) geographic proximity to the AMC vs. the IDS — likely more important for interventions requiring frequent visits (e.g. chemo vs. surgery), d) sociodemographic factors, e) “connectedness” of the patient to the IDS, and f) enrollment in a capitated plan.

Methods: The Oncoshare database includes all patients diagnosed with or treated for breast cancer between 2000 and 2013 at Palo Alto Medical Foundation (PAMF), a community IDS, or Stanford University Medical Center, a neighboring AMC. The database links electronic health records from both institutions with data from the California Cancer Registry. We will examine treatment venues for patients initially diagnosed at PAMF using electronic health records data to determine baseline health, comorbidities and breast cancer treatments and California Cancer Registry data for tumor characteristics, sociodemographic factors (based on census tract) and primary payer at diagnosis. Patient “connectedness” is measured by duration with the primary care provider and number of other specialists seen at PAMF. We will use multiple variable logistic regression models to examine determinants of choice of IDS vs. AMC providers for breast cancer treatment.

Results: During the studied period, 3,784 women were newly diagnosed at PAMF; 80% received all their treatment at PAMF, 5% received all their postdiagnostic care at the AMC, and 4% received treatment at both sites during their initial cancer episode. Treatment sites were unknown for 11% of patients. Analyses to understand these patterns of where patients seek care are underway.

Conclusion: Observational studies often contrast resource use and outcomes across delivery systems. If patients have choices among systems, however, it is critical to understand whether their reasons for seeking care at different sites may partially account for observed differences in performance.




June 24th, 2016


August 12th, 2016