Ignorant Night Thoughts on Regional Medical Cost Disparities: Wednesday Focus for September 17, 2014

What I think of as the Louise Sheiner fact, as set out by David Cutler (paraphrasing from memory):

Q: How much of regional variation in real health-care (Medicare) costs is due to the fact that some regions have sicker populations than others?

A1 (micro): If we examine how much sicker people in different regions are, and multiply the difference in average sickness by how much extra treatment sicker people get on average, we get an incremental regional R² ~ 0.1: an extra 10%-points of the regional real cost variation can be accounted for because some regions are sicker than others.

A2 (macro): If we just regress regional real costs on some plausible indicator of regional average sickness, we get an incremental regional R² ~ 0.5: an extra 50%-points of the regional real cost variation can be accounted for because some regions are sicker than others.

Clearly the facts are: (a) sick people in high-average-sickness regions are getting treated more than sick people in low-average-sickness regions, and (b) not-very-sick people in high-average-sickness regions are also getting treated more than not-very-sick people in low-average-sickness regions, and (c) there are (by definition) more sick people in high-average-sickness regions than in low-average-sickness regions, and (d) fact (c) is not a large amount–is, in fact, only one-fifth–of the association between average sickness and regional real Medicare costs, and (e) the computer is almost (but not quite) as happy using regional Confederate sympathies in the winter of 1861 to predict regional real cost differentials as it is using current regional average health status:

This seems to me to be an extremely interesting fact about the U.S.’s health care system and health care economy. Complicating things further for me trying to write about this extremely interesting fact about the U.S.’s health care system and health care economy: I am not a health-care economist. I did play one in the Treasury Department from 1993-1995 because the slot of Deputy Assistant Secretary for Health-Care Programs was occupied by Marina Weiss, Lloyd Bentsen’s long-time health-care policy strategist, and the woman whom all of us in the U.S. Treasury in 1993-1994 expected would someday have a lunch with Sheila Burke, Senate Minority Leader Bob Dole’s (R-KS) long-time health-care policy strategist, and they would do a deal, but Dole–apparently, I have never heard an explanation from anyone in the Dole camp, only evasions–decided that he couldn’t win the Republican presidential nomination in 1996 if he did a deal with Clinton and so demonstrated that he was really a RINO, failing to recognize that he couldn’t win the presidency in 1996 unless he did do a deal with Clinton and could sell himself as somebody who got things done and had serious legislative accomplishments, but I digress…

At any event: Marina told me that I was closer to being a health-care economist than she was, and so I needed to man-up and do the job–which I did essentially by asking health-care economist David Cutler (plus Jon Gruber, Sherry Glied, and anyone else I could find) and health-care lawyer Ann Marie Marciarille (my wife) what I should say, and simply parroting them, but I digress again…

Let me start again: This is a very interesting and important fact:

• Louise thinks: regional average health status -> regional patterns of practice -> health care costs -> cheerleading won’t work because exiting regional patterns of practice exist for good reasons and will be very hard to change without first changing regional average health status

• David thinks: other stuff -> regional patterns of practice -> health care costs -> cheerleading has a good chance of working because regional patterns of practice are a complicated emergent phenomenon only tangentially related to regional average health status.

David’s interpretation, however, leaves a puzzle: why are the “other things” that produce high-cost cowboy medical practice patterns so highly correlated with the “other other things” that produce a sick population–as Louise says, a five-year differential in life expectancy at age 65 between Minnesota and the next state on the list, Mississippi? You can say “social capital”, but then why is, say, high Diabetes prevalence a better measure of low social capital than the measures political scientists construct of low social capital? The standard economist’s response is: “well, political scientists aren’t terribly good at their jobs, and ought to construct better indexes of low social capital”. Other economists may nod sagely at this observation and have their prejudices about political scientists confirmed, but I do not find this satisfactory.

The only even half-intelligent thing I have to say is that this smells to me a lot like the cross-country growth regression wars, in which there are massive vicious cycles of relative poverty at work, and as long as you give the computer something that hooks into those vicious cycles–even if it is just “located in sub-Saharan Africa” or “closeness to the equator” the computer is very happy telling you it is the secret to failures of economic development…