Medical Spending Variation: 1/2 Patients, 1/2 Places

In Miami, health care providers spent about $14,423 per Medicare patient in 2010. But in Minneapolis, average spending on Medicare enrollees that year was $7,819, just over half as much. In fact, the U.S. is filled with regional disparities in medical spending. Why is this?

One explanation focuses on providers: In some regions, they may be more likely to use expensive tests or procedures. Another account focuses on patients: If the underlying health or the care preferences of regional populations varies enough, that may cause differences in spending. In recent years, public discussion of this issue has largely highlighted providers, with the implication that reducing apparently excessive treatments could trim overall health care costs.

But now a unique study co-authored by MIT economists provides a new answer to the medical cost mystery: By scrutinizing millions of Medicare patients who have moved from one place to another, the researchers have found that patients and providers account for virtually equal shares of the differences in regional spending.

“We find it is about 50/50, half due to patients and half due to places,” says Heidi Williams, the Class of 1957 Career Development Associate Professor in MIT’s Department of Economics, and a co-author of a new paper detailing the study’s findings.

That’s MIT News ably summarizing the new Finkelstein, Gentzkow, and Williams paper, Sources of Geographic Variation in Health Care: Evidence From Patient Migration (ungated).

If the half of the variation that is due to place is inefficient (which could mean too low or too high but probably means too high given that the medical care curve is flat) then this puts an upper limit on the gains from standardization but still a quite high limit.

By the way, Finkelstein and Gentzkow are both recent John Bates Clark Medal awardees and Williams is a MacArthur “genius award” winner. Perhaps I should have titled this post, assortative co-authoring.


A third factor may simply be that doctors are cheaper in Mississippi vs NYC.

Doctors are cheaper in NYC

Medicare uses standardized geographic adjustments for payments, so it should be pretty easy to cancel that effect out.

Funny how I immediately associate each view with a political party in the USA.

As a healthcare actuary, I'd love to be able to read that article, but sadly don't have access, unless there is something I'm missing? (or I'll just pay for it later). I'm curious how they measured the half that's due to the patient. How are they factoring in the increasing morbidity of the seniors? I find it hard to believe that a patient simply chooses to have more medical care after they've moved (unless I'm misunderstanding something), unless there's some confounding factor that contributes negatively to health like the stress of a move, etc. From my experience, inefficiencies on the part of the providers is a big chunk, but next are huge regional differences in the leverage held by insurers to negotiate rates with large provider groups/hospitals.

Should say, that I assume this study was probably done with FFS members and not Med/Supp-Medicare Advantage members, so any contracting/managed care aspects probably not a big factor, but still curious how they measured the patient aspect.

I believe the link you want is

Why didn't the authors use parallel construction? Does "health care providers spent . . . . per Medicare patient" mean the same thing as "average spending on Medicare enrollees"?

Also, have they factored in age? Florida may not be a good comparator as many seniors move down there for the climate. I would suspect that those electing to stay behind in Minneapolis would be healthier than those moving south. For example increased cardiovascular problems will lead to increased testing and procedures and someone who gets about in the Minnesota winter might be healthier in this regard. As JShots says, unless you can strip out the morbidity the comparison is really imperfect.

Healthier, or more reconciled to circumstances.

I could see it going the other way, the very healthy may be unable to move to Florida.

Personal experience says that patients who travel south for the winter tend to stay in the south when their health deteriorates to the point where travel is difficult. Winter is physically demanding in the north.

I find it hard to believe that a patient simply chooses to have more medical care after they’ve moved (unless I’m misunderstanding something), unless there’s some confounding factor that contributes negatively to health like the stress of a move, etc.,

It's Florida, a collecting pool of the complainers among the elderly. Rod Dreher said after several years of living there he was in a continual burn at the old.

I lived in Florida 2003-2008, and I never had any major problem with the elderly down there. The hassles in Florida are more generically "Florida man" problems-- a huge pool of idiots of all ages, sexes and races.

And are the health outcomes roughly the same despite the differences in spending? Paging Robin Hanson

Some years ago, I saw a lecture by a pediatrician. The chap said he'd discovered that 80% of his time was spent with 20% of his patients. He was curious about that 20% and reviewed their records. He surmised he'd be looking at kids with juvenile diabetes and other chronic problems. He said that in reviewing the records, he found this wasn't the case. The heavy consumers of his time had chickens*** problems, nothing chronic or serious. He came to the conclusion that the culture of the family was the differentiating factor and that some people have a standard reaction to illness and some have a variant reaction. The variant reaction ("have cake and ice cream and take the week off") is a time and resources sink. As you grow into middle age and into old age, you discover some of your peers are into their medical problems. The one proximate to me is both the wife and the mother of a physician. Her husband made a rule 40 years ago that he wouldn't doctor his family. He had an uncle on his patient roll, but that's as close as he'd take. Her son trained for a subspecialty which ensured his mother wouldn't be picking his brain.

It's Miami, Florida. It has a huge pool of retirees and thus the health care system caters to the clients with the latest and greatest equipment and procedures. And since the government is paying for it, cost is not an objection.

Furthermore, Miami is a significantly larger market than Minneapolis, even without taking the disparity of average age into account:

Metro Miami: 5.0 million

Metro Minneapolis: 3.5 million

So, the paper is written by premier award-winning economists and is to be published in QJE. It is 50 pages of intense analysis and apparently has multiple online appendices with untold number of additional complex analyses. Who am I to criticize? I'd better just accept their findings since I won't live long enough to decipher the paper.

But, after an initial read, I do wonder about a few things. There appears to be no mention of issues regarding finding providers who will accept Medicare. There are regional variations in this - since Medicare pays so much less than private insurers, the ability to find providers who accept Medicare depends greatly on the Medicare population in an area (for example, in Florida most providers accept Medicare since most of their potential customer base is old while in Alaska it can be difficult to find a provider willing to accept Medicare - unless the patient needs a heart specialist or if the provider is a hospital, since their customer base consists of many old people). It would seem that this institutional fact might be relevant to a study purporting to analyze health care utilization with moves between different regions.

The use of "matched" populations I find somewhat unnerving. I've never liked matched samples since the ability to match all relevant characteristics is both unknown and likely to be difficult, if not impossible. But, frankly I can't understand the paper enough to know if the matching is of any relevance to the empirical results.

I guess I'll just have to accept the paper as valid since I am not willing (and possibly not able) to determine for myself whether their analysis is valid. I do which the papers could be released in a version that can be read, rather than the versions required by publishers. Referring to tables and figures that are at the end of the paper, rather than in the body, makes for nearly impossible reading - especially in a paper as complex as this one.

If you get used to getting expensive procedures and move to Mississippi, you still want them? So the place is 100%, not 50%? Or if Mississippi has better preventing care. Or...

Pro Tip:

Nearly all econ papers are available as ungated working paper versions if you simply google them.

You're welcome.

As Medicare provider for over 40 years this all seems same ole same ole. As in the Dartmouth studies. The predominant factors in Medicare spending differentials has been local demographics and regional practice patterns. Or am I missing something?

Alex, Atul Gawande wrote about this exact topic over seven years ago:

It's fraud. I can haz my genius medal now?

Comments for this post are closed