News

Earlier in June, the New York Times ran an article by Adleson and Reed questioning the findings in the Dartmouth Atlas. Jack Wennberg and colleagues have been working in this field and documenting small area variation in health care since the 1970s. However, the work was not much recognized outside of academic and health care analytic circles until the start of the discussion on national health reform. In a very short period of time, the analysis went from being in the sole domain of providers and policy wonks (hmm, could that be me?) to being on the tip of the tongue of policy makers in Congress and the White House. Tracing the trajectory of this research from relative obscurity to the New York Times article provides an interesting insight into both the policy making process and the risks and opportunities inherent in trying to translate research into public policy.

The basic concept behind small area variation analysis is that health care utilization differs by community in ways that cannot be fully explained by the characteristics and medical need of the population being served in that community. Stated in this way, I think there are few who would actually disagree with that observation. On this point, the data are strong and have been consistent for the more than 40 years history of this kind of analysis. While the methodology has changed over time to look at these trends, the simple fact of unexplained variation is a robust concept. However, taking that observation and deciding what to do about it is an entirely different issue. To craft an intervention that tries to reduce unexplained variation, there must be a theory behind what causes the variation – and therein lies the rub.

There are many different theories to explain why there is so much regional variation in health care. Some believe that the variation is principally driven by the supply of providers (for those of us who went to public health school some time ago, the old Milton Roemer law: “a built bed is a filled bed is a billed bed”). Some believe that the variation is a result of practice patterns that have grown up regionally over time combined with a lack of clarity in the evidence base for treatments. Some argue that the variation has to do with true differences in patient characteristics that aren’t accounted for in the methodology. And, some contend that the way care is organized and delivered accounts for these differences.

While these explanations are not mutually exclusive (and many think there are elements of all of them at work), the explanation one believes is most important will lead to different ways to address the issue. And, beyond that, there is an underlying difference of viewpoint as to whether such variation is good or bad, i.e. whether areas with higher use rates are providing better or worse care and producing better or worse outcomes. For some discussion of this issue, it is useful to look at Dartmouth’s response to the New York Times article.

What happened with these data, however, is instructive and illustrative of the challenges inherent in translating research into policy. When Congress started paying attention to the data and seeing it as an opportunity to help with the cost savings needed to make health reform work – surprise, surprise –the “what to do about it” question became over simplified and the answers started a debate between high spending and low spending states about who should get more of the Medicare pie. That high profile debate resulted in the research itself becoming open to more and more scrutiny and critique and to the ultimate challenge posed in the New York Times article.

In the end, there probably isn’t one explanation for the variation or one set of solutions. The data included in the Dartmouth Atlas and in other analyses like this are a starting point for understanding where the opportunities are for quality and cost improvement – more analysis is really needed to get behind the numbers to understand the dynamics that lead to them. What would be most unfortunate in all of this debate, however, would be to lose sight of the fact that the degree of regional variation in how medical care services are provided in this country is enormous and much of it cannot be easily explained by differences in patient characteristics. The data in the Dartmouth Atlas are important indicators of opportunities to reduce health care spending in this country, and while there can be debate about how much and in what ways, the data must be taken seriously.