Background: Care bundling is an emerging health financing innovation to change the incentives of care, intended to improve quality of care and promote better resource use. In 2016, Medicare outlined a proposal for changing Medicare reimbursement for outpatient drugs through pre-determined care bundles. To gauge the potential for care bundling, we examine one of the first comprehensive efforts, the Oncology Care Model (OCM). This paper shows that the oncology care bundles likely used by OCM have large variation in cost per patient across the United States. Methods: For this analysis, we utilized five years (2010-2014) of the Medicare 5% limited data set (LDS) of fee for service claims. All seven claims segments were used in the analysis including: physician/carrier Part B, durable medical equipment,outpatient hospital, inpatient, skilled nursing facility, home health, and hospice. The 5% LDS sample of Medicare beneficiaries used to identify patients with cancer bundles totaled 17,143 in 2014. An approximate national estimate would be 20 times 17,143, yielding 342,860 beneficiaries. Results: Our analysis of Medicare claims for the three most expensive bundles (lung cancer, prostate cancer and lymphoma) from 2010 to 2014 shows over a 400% difference in per capita bundle reimbursement between US states. Furthermore, we found that the mix of reimbursements within all bundles of fee for service claim types varies meaningfully. Finally, we show that the rank order of most expensive cancers to treat at a patient level is not correlated with the most expensive cancers at a societal level. Conclusions: There is substantial geographic variation in per capita cancer costs that is not consistent for the top 3 cancer bundles. Therefore, policy-making based on system-wide geography will likely not produce a consistent solution. As a result, policy formulation will be challenging when patient cost management is a goal, especially in a healthcare sector where innovation is likely to move faster than robust and thoughtful cost containment strategies.
Stephen T Parente and Lisa Tomai