Perspective: Geographic Variation In Health Care Spending—What Direction For Policy?

Last month, the Institute of Medicine (IOM—now known as the National Academy of Medicine) published their much-awaited report titled “Variation in Health Care Spending: Target Decision Making, Not Geography,”[i] which recommends that CMS (Centers for Medicare and Medicaid Services) should not adjust Medicare payments geographically; instead, CMS should continue to focus on value-based payment reforms, such as patient-centered medical homes, bundled payments, and accountable care organizations. Their rationale is that geographic adjustments in payments “would unfairly reward low-value providers in high-value regions and punish high-value providers in low-value regions” (IOM Report Brief, p. 2). The IOM Committee found considerable variation in costs and health care utilization, even with very small units of analysis such as individual hospitals and multi-provider practices. Thus, it would not be fair, and possibly not effective, for CMS to develop payment rewards and penalties at an aggregate geographic level like the Hospital Referral Region (HRR). This report concluded that the differences may result from multiple elements, such as: (1) demand-side factors (patient preferences), (2) supply-side factors (provider supply and preferences), and (3) demographic factors (especially health status). 

Close on the heels of this IOM report, a group of health economists from Harvard and Dartmouth have conducted analyses that indicate[ii] that patient demand does not explain much of the regional variation in health care spending [iii]. These authors conclude that, while peer pressures and financial incentives matter, the most important determinant is physician beliefs about treatment. Their analyses show that 36 percent of end-of-life spending, and 17 percent of U.S. health care spending, are associated with physician beliefs that are unsubstantiated by clinical evidence. This observation calls for understanding how physician beliefs are formed and how they could be altered. This paper presents a detailed discussion, considering multiple explanatory models, but the results are ultimately inconclusive. Of course, one might wonder whether physician beliefs that lead to higher utilization can fairly require taxpayers in a low-utilizing area to subsidize higher expenditures in another area. Indeed, the other two explanations from the IOM Committee, patient preference and service supply, might well be true and still be unfair.

Others have proposed alternative explanations for geographic variation. The authors of another recent paper [iv] concluded that geographic variation in costs of care for traditional fee-for-service Medicare beneficiaries are largely explained by population health and disease burden. These authors contend that even people who die soon are not equally sick in different parts of the country. They agree that bias in case-mix adjustment arising out of regional coding or diagnostic practices are important; however, in their analyses, these factors are small contributors to regional cost variations.

If geographic variation is less affected by systemic inefficiencies or physician/patient preference patterns but more by regional difference in health status, then payments might correctly be higher in areas with higher disease burden. This contention leads to consideration of how this situation could arise. To have higher disease burden in the last months of life in one area than in another, the patients would need to have a different  and more expensive mix of non-lethal illnesses or the high-cost areas would have to be sustaining life longer than low-cost areas. These possibilities, it seems, would be difficult to test, but possible.

Although the debate over the causes of geographic variation is not new, the disputed reasons and policy strategies are thought-provoking for policy makers. Clearly, the data does not yet provide a satisfactory causal link for regional variation. Armed with less-than-perfect evidence and resources, we are still faced with policy choices. For example, should we aim for more standardization in care delivery across the country, and, if so, for which reasons? If reducing costs is our major concern, should we aim to reduce the expenditures of high-cost regions to approximate those of the lowest quartile, especially if the extra expenditures come to be perceived as being unfair? And if we aim to achieve this, what policy levers are available, if reductions in broad areas would unfairly penalize the efficient provider in high-cost areas, as the IOM committee has concluded?

Also, given the geographic variation in socio-economic conditions, policy makers probably need better information about how differences in demographic factors like life expectancy, poverty, and employment affect and result from the delivery of care. Would it be prudent to invest in non-medical services like housing, meals, or family caregiver support to reduce our need to resort to acute medical care? I believe the discussion ought to be around the consequences of substantial differences in consumption of public funds in different parts of the country and the best ways to address variations perceived to be unfair, rather than imposing a simple geographic incentive and penalty.