Abstract Objectives: To characterize variation in Medicare expenditures on initial breast cancer care across hospital referral regions (HRRs) and to examine the relative contribution of patient characteristics and treatment factors to such variation. Methods: This was a retrospective cohort study using the 2003-2007 Surveillance, Epidemiology and End Results (SEER)-Medicare linked database and the Medicare 5% random sample of non-cancer beneficiaries. Each woman with localized (stage I-III) breast cancer (“case”) was matched to a woman without cancer (“control”) based on HRR, age, comorbidity and Medicare expenditure in the year prior to cancer diagnosis. We defined initial phase of care as the period from two months prior to breast cancer diagnosis (for cases) or index date (for controls) through 12 months after the diagnosis or index date. For each HRR, we calculated the risk-standardized cancer-related Medicare expenditure as the difference in total expenditure between cases and controls, using hierarchical generalized linear models to control for clustering by HRR and adjust for patient characteristics (age, race, comorbidity, and tumor characteristics) and treatment factors. Treatment factors were first assessed by whether different treatments were used (surgery, radiation therapy, chemotherapy, growth factors, and imaging services), and then assessed by specific treatment modalities (breast conserving surgery, radical mastectomy, intensity modulated radiation therapy, external beam radiation therapy, brachytherapy, traditional chemotherapy, biological therapy, growth factors, and imaging). All estimates were reported in 2009 U.S. dollars. Results: There were 35,055 patients with breast cancer and an equal number of controls in our cohort. After excluding HRRs with fewer than 25 cases, there were 78 HRRs in our final analysis. Unadjusted Medicare expenditure on breast cancer-related care averaged $19,207 per patient. HRRs in the highest quintile had an average expenditure of $23,522 per patient, which was $8,032 higher than HRRs in the lowest quintile (mean expenditure = $15,490). Patient characteristics explained only 18.5% of the difference in total cancer-related expenditure between the highest and lowest-expenditure quintiles. Treatment factors explained more variation across the HRRs. Adjustment for whether different treatments were used (e.g., chemotherapy: yes/no; radiation: yes/no) explained an additional 26.0% of the difference between the highest and lowest-expenditure HRR quintiles. In contrast, adjustment for specific treatment modalities (e.g., specific type of radiation) explained 36.4% of the variation. Variation in expenditures on radiation therapy contributed the most to the difference in total cancer-related expenditure between the highest and lowest-expenditure HRR quintiles. Conclusions: There is large regional variation in Medicare expenditures on initial breast cancer care, even after accounting for patient characteristics. Treatment factors (whether cancer therapies were used at all and the specific modality of therapy used) were important contributors to such regional variation. Future work exploring the relation between cancer treatment intensity, costs, and outcomes is warranted. Citation Information: Cancer Res 2013;73(24 Suppl): Abstract nr P6-07-05.