The Latest on Geographic Variation in Medicare Spending: A Demographic Divide Persists But Variation Has Narrowed

In 2009, the physician and author Atul Gawande focused the attention of the health policy community on McAllen, Texas, linking exceptionally high Medicare spending in that city to what he characterized as a profit-oriented “culture of money” among providers.¹ Although research on geographic variation in health care utilization and spending predated that article by many years, Gawande’s article struck a nerve and attracted significant attention among policymakers and researchers.

Gawande’s work built on examinations of differences in Medicare spending per capita among hospital referral regions (HRRs)² by researchers at Dartmouth University. Dartmouth researchers concluded that such differences could not be explained by differences in health status, and that “increased Medicare spending in high-cost regions provides no important benefits in terms of survival.”³ The Dartmouth researchers suggested that the differences in spending across HRRs resulted mainly from differences in practice patterns, which could be addressed by policy interventions, such as changes in financial incentives for providers.⁴

In contrast, other researchers have emphasized differences in beneficiaries’ health status as drivers of geographic variation in Medicare spending. Reschovsky et al. found that differences in disease burden were largely responsible for geographic variation in Medicare spending, based on their analysis of data from 60 communities.⁵ Similarly, Zuckerman et al. found that beneficiary demographics and health status help to explain geographic differences in Medicare spending, but also found that even after adjusting for other possibly relevant factors, such as provider supply measures, significant unexplained differences remained.⁶ Sheiner also concluded, based on a state-level analysis, that much of the geographic variation in Medicare spending can be explained by differences in health status and demographics, which are less amenable to policy intervention than practice patterns.⁷

In 2009, Congress directed the Institute of Medicine (IOM) to conduct a series of studies on geographic variation in Medicare spending and in the broader health care system. The IOM documented significant variation in spending even within high-spending and low-spending areas; showed that much of the geographic variation in Medicare spending is attributable to differences in post-acute care spending; and ultimately recommended against changes in payment policy designed primarily to reduce geographic variation in spending.⁸ The IOM expressed some concern that reductions in payments to providers in high-spending areas could inadvertently penalize providers practicing appropriately who happened to work in high-fraud areas.

This paper contributes to the body of research on geographic variation in Medicare spending in three ways. First, we analyze variation in Medicare per capita spending at the county level, rather than at the state or HRR level, using the most current data available (through 2013). Second, we analyze detailed data on utilization and spending for specific types of services in our comparisons of high- versus low-spending counties. Third, we examine changes over time from 2007 to 2013 in county-level Medicare per capita spending to compare counties with high versus low rates of growth, and to assess whether geographic variation in per capita spending is increasing or decreasing.

Our analysis addresses the following questions: