Thursday, July 1, 2010

NEJM -- Geographic Variation in Medicare Spending

Editorial by Arnold M. Epstein, MD
 
In my third and fourth years of medical school at Duke University, I was totally engaged in learning clinical strategies and algorithms: when to admit a patient with chest pain from the emergency ward, when and how to work up dyspepsia, when to schedule laboratory work and ambulatory follow-up for stable hypertension. These protocols are the nuts and bolts of day-to-day practice. I learned them as an apprentice, trailing fellows and attending physicians on ward rounds and in the clinic, since the level of detail needed to guide these decisions never seemed to be in the scientific literature. After finishing medical school, I moved to Boston for house-staff training. To my great surprise, I encountered a whole new set of protocols for exactly the same conditions. Their proponents espoused them with equal vigor and certainty as my mentors at Duke, yet the strategies were often substantially different. It was then that I realized the scary truth: the science behind medicine is sorely lacking, and often there is no clearly right answer.


Nearly 40 years ago, Wennberg published his seminal work showing variation, by a factor of two or more, in the rates of common surgical procedures such as tonsillectomy and cholecystectomy across small geographic areas in Vermont, areas that had no apparent differences among their populations of patients.1 In subsequent years, Wennberg, Fisher, and their colleagues at Dartmouth have used administrative data and information from surveys2 to show that regional variation is a national phenomenon that persists despite adjustment for differences in the case mix and that greater use of medical services is not associated with higher quality or better outcomes. In fact, the Dartmouth findings suggest just the opposite and are touted as evidence of great waste. Yet critics persist, worrying about the shortcomings of administrative data.


The article by Zuckerman and colleagues in this issue of the Journal3 both supports and modifies the Dartmouth gospel. The authors used Medicare claims data linked to survey data on health status from the Medicare Current Beneficiary Survey to examine spending in Hospital-Referral Regions (HRRs) nationwide. The claims data alone allowed them to recapitulate the Dartmouth findings, showing large geographic variation in expenditures for medical services: a 52% difference in per-beneficiary expenses between the top-spending and bottom-spending quintiles of HRRs. However, sequential controls for demographic characteristics and survey information about baseline health status, subsequent changes in health status, and newly or previously diagnosed illness reduced the difference in expenditures between the top and bottom quintiles to 33%. Because finer categories of risk-adjustment data regarding the health of beneficiaries were used in this study than in previous studies, they explained a greater proportion of the variation in spending. However, much as was found in Dartmouth work,2 substantial regional variation in Medicare spending remained.

..........
Of course, we should anticipate some variation across geographic regions. With the constant flow of new technology and drugs and the ever-changing scientific literature, it would be unreasonable to expect complete consensus or equal rates of diffusion of new clinical practice across regions. Without adjustment, the difference in spending between the quintiles of highest spending and lowest spending was more than 50%, according to Zuckerman and colleagues. That degree of variation, with no apparent difference in outcomes, raises questions of unacceptable inefficiency in the high-spending areas or even of impropriety. In contrast, if the true variation in spending is only 33%, does it still matter? I think yes. Changing the patterns of use of medical resources in the higher-spending areas to resemble more closely the patterns in the lowest-spending areas would save an enormous amount of money. Moreover, unwarranted variation reminds us that in too many areas of medicine, we lack knowledge of what works best and far too often fail to follow consistently what we know to be superior practice.


If regional variation could elucidate specific medical services that should be targeted as unnecessary or overused, we would merely study patterns of care in high-use areas. Unfortunately, past studies of services such as surgical procedures indicate that high- and low-use areas have similar proportions of inappropriate care.5,6 Instead, most of the discrepancy appears attributable to discretionary decisions by physicians — decisions (such as whether we ask stable patients with hypertension to return for a routine check in 2, or 4, or 6 months) that lack consensus about the correct answer.7

Read More
Enhanced by Zemanta

No comments:

Post a Comment