Nothing like a good fight among academicians and researchers and that is certainly what the issue of geographic variation in health spending has provoked. Researchers from the Brookings Institution publish their findings in a new report. (Brookings Report) Much of the seminal work in this area was done by researchers associated with the Dartmouth Atlas project, who have largely taken the position that spending variation is due to physician practice pattern differences. Other researchers have found factors such as geographic health status differences to be of more importance. Most of the Dartmouth work was done only with Medicare data; the Brookings researchers and others have used commercial data as well. These authors look at socioeconomic and other population characteristics by state and compare that to the Dartmouth approach, which attempted to assess individual beneficiary status, albeit in a limited way. Using this approach, the researchers are able to explain a much higher percent of the variation is spending. For example, just looking at the rate of diabetes in a state’s population dramatically raises the correlation with spending. Using these variables reduces the variation between the top quintile of states in spending compared to the lowest quintile from 40% to 5%. The difference from some of the Dartmouth research is that the state-level characteristics include all the adult population, not just the elderly in Medicare, and therefore give a truer picture of underlying health needs in an area. The authors work is unlikely to be the last word on the issue, but it increasingly appears that trying to find a policy solution based on geographic factors would be a mistake.
Focusing the debate on larger scale geographic variations misses the most significant opportunity. For now, lets ignore any variation that may be due to different unit prices whether between or within geographic regions. If we truly believe that evidence-based medicine will allow us for many conditions to define a range of acceptable treatment paradigms; then we should be gathering all-payer data on physicians and other clinicians and analyzing their practice patterns for those conditions against our evidence-based norms. Incentives or penalties can be put in place to encourage physicians to stick to those norms, although I always get a little queasy about this approach given that researchers frequently seem to be wrong about the evidence on what is the best approach even for common conditions like hypertension or diabetes. Medicine is practiced at the individual physician level and that is where improvements on the utilization side of the spending equation need to be prioritized. Even if there is some geographic component to variation, it becomes irrelevant if we are measuring individual physicians. But don’t be surprised if that individual doctor approach doesn’t result in a large decrease in utilization and spending; there is a great deal of underutilization of appropriate services that will largely cancel out any reduction in inappropriate ones.