Using Medicare Data to Rate Physician Quality

Last week, the federal government announced that it would allow Medicare claims data to be used for the purpose of disseminating physician quality information to the public. What’s news is not that there will be attempts at creating so-called quality “report cards”–attempts at those have been around for some time–rather it’s that the Centers for Medicare & Medicaid Services (CMS) has finally agreed to let a wide range of folks access Medicare claims data for the purpose, which hasn’t happened before on this level.

Last week, the federal government announced that it would allow Medicare claims data to be used for the purpose of disseminating physician quality information to the public. What’s news is not that there will be attempts at creating so-called quality “report cards”–attempts at those have been around for some time–rather it’s that the Centers for Medicare & Medicaid Services (CMS) has finally agreed to let a wide range of folks access Medicare claims data for the purpose, which hasn’t happened before on this level.

But what are we to make of this new development? Is it a good thing or not? Giving the “consumer” more information on which to base their selection of a physician and their use of health care services seems like a good thing. After all, it’s essentially central to the idea of a well-functioning free market. As any health economist will tell you, the information asymmetry between consumers and providers leads to all sorts of peculiarities that cause the health care market not to behave like the market for other goods and services. This could then conceivably be a step in the direction of correcting some of those peculiarities.

The real question, though, is how good will this information be? Or, said another way, is poor information preferable to no information? Now, that doesn’t mean that there’s not a lot of excellent potential in these Medicare claims data. On the contrary, there’s much to be learned here. Of course, the realization of that potential is a function of the empirical rigor of the analyses researchers like myself undertake. No, the real worry I have is how this translates to the lay public without grossly oversimplifying things.

Let’s say a system is devised that, in true “report card” fashion, assigns physicians a grade ranging from “A” for outstanding to “F” for visit at your own risk. The public would certainly understand such a grading system, and people would be expected to show a clear preference for “A”-rated physicians over “F”-rated ones, but what about the bulk of physicians in the “B” and “C” range? It’s entirely possible, depending on the rating algorithm used, that a physician who excels in one particular area nevertheless gets a “C” rating. Would the public do its homework, or would it avoid doctor “C”? I worry that the latter may be the most likely outcome.

Again, I’m not saying that efforts to monitor quality and report that information publicly are a bad idea. Far from it. I’m merely suggesting that we must be extremely thoughtful in how we engage in such efforts, because the potential for significant unintended consequences is quite real. We must figure out how to approach these data using the most sophisticated of techniques, all the while with an eye on translating what we find in a manner that is accessible to the public without being “watered down” or less than accurate. The risks and the rewards are great.