Network News

X My Profile
View More Activity

Health Reform for Beginners: The Suprisingly Important, Occasionally Controversial, Dartmouth Atlas Studies


I'm not exactly sure how to convince you that this is an interesting link. It is, after all, a salvo in an ongoing methodological dispute between health-care experts. But it could be pretty important.

If you were to distill the administration's health-care reform theory down to two sentences, they would read something like this: As the graph atop this post shows, the cost of health care varies enormously between states, but there is no obvious connection between expensive care and better outcomes. Thus, the most promising goal for reform is to make the high-cost states look more like the low-cost states.

This theory is the product of the Dartmouth Atlas studies. The researchers behind those reports have spent the past three decades documenting the huge regional variations in Medicare spending and trying to discover if the differences in spending tracked differences in quality or outcomes. As far as they've been able to tell, there's no connection. On the strength of this data, members of the administration will tell you that the important comparison of health care reform is not between America and France, but Minnesota and Florida.

Critics, however, have argued that their data is fundamentally flawed. Spending differences, they say, are the product of sicker patients and the pathologies of urban life. If they're right, then the administration's approach to health reform is substantially based on a flawed study.

Over at the New York Times' Economix blog, Jonathan Skinner, one of the economists associated with the study, responds to those criticisms and explains what we know, and what we don't know, about the spending. It's worth a read. So is this post by health writer Maggie Mahar. And, if you want to go really deep, download anything on this page.

(Graph credit: Congressional Budget Office.)

By Ezra Klein  |  June 15, 2009; 11:11 AM ET
Categories:  Health Economics , Health Reform , Health Reform For Beginners  
Save & Share:  Send E-mail   Facebook   Twitter   Digg   Yahoo Buzz   StumbleUpon   Technorati   Google Buzz   Previous: I Am an Important Fact: Predictably Irrational Edition
Next: Top Line


"Thus, the most promising goal for reform is to make the high-cost states look more like the low-cost states."

This doesn't follow at all because it does not take into consideration the difficulty of this goal. The most promising goal is the elimination of disease.

I do not see a single practical measure suggested for accomplishing Ezra's goal. This is in distinction to the more modest goal of an efficient insurance system like the rest of the world, like France.

Posted by: lensch | June 15, 2009 11:39 AM | Report abuse

If we are going to deal with health care in America in a way that provides quality care for all, we must address costs. Even to preserve existing levels of care in the future, we must address costs. The Dartmouth Data offer insights into a way of significantly reducing costs while maintaining or even improving quality of care.

The facts in the Dartmouth Data make many important players in health care very uncomfortable, since they show that too often the health care emperors have no clothes, pointing out serious flaws in systems that otherwise enjoy excellent reputations. In addition, they imply that a large part of the problems of high costs in American health care compared with foreign systems is due to ineffective overuse of high tech management without any positive impact on results.

Attacks on the Dartmouth Data have been rebutted effectively by people within the Dartmouth project in the articles Ezra cites, and by outsiders as well. Among others, Gawande’s now famous article specifically addresses and debunks many of the criticisms.

There is one more important point to be made, however. There are many health care systems that are able to deliver cost effective care despite being located in areas where costs are generally high. Mayo Clinic’s Scottsdale branch, the Cleveland Clinic, the Geisinger Clinic, Kaiser in the Bay Area, and others deliver very effective care at lower costs than the general performance of providers that are their neighbors. Effective high quality care at significantly lower costs can be delivered right in the heart of the highest cost geographic areas, making it clear that it is patterns of management, not the nature of patients or the costs of operation, that causes the variation Dartmouth has documented.

Posted by: PatS2 | June 15, 2009 11:52 AM | Report abuse

it seems to be the case that obama/orzag claim to be able to get savings from medicare to help pay for health reform has suddenly been accepted as valid and an acceptable part of the pro-reform argument

Posted by: jamesoneill | June 15, 2009 12:23 PM | Report abuse

There is so much proof that more is not better in Health Care as in many other things.

It has been shown that the regional differences in expenditures are not due to demographic differences nor in different levels of illness.

Most of the excess expenditure is from doctors preferring the higher cost option even when the results of the lower cost one is just as good or even superior by many measures. That and adding in unnecessary diagnostics just for increasing the providers billings.

Every excess intervention or treatment carries some risks. These added risks can cause discomfort, harmful side effects themselves needing treatment, and even deaths.

Thus decreasing excess procedures will save lives and reduce patient discomfort as well as reducing cost.

Posted by: sam_dobermann | June 15, 2009 3:52 PM | Report abuse

i am familiar with the Geisinger system as I have a client that uses them. They use an approach that they are working together with the client as opposed to "selling them a product".

Posted by: visionbrkr | June 15, 2009 3:56 PM | Report abuse

The Dartmouth Atlas is not medical costs. It is Medicare costs.

When one looks at total medical expenditures by state and at the average per patient cost of medical services, one reaches an entirely different conclusion than the Dartmouth Health study.

In the two states, Colorado and Florida, cited as an example by the Dartmouth study, according to the Meps (Medical Expenditure Panel Survey), the average medical expenditure (including Medicare, private health insurance, out of pocket, uninsured, etc.) by all patients in 2006 in Colorado was $3,585. It was $3,586 for Florida. The Dartmouth Health study, which only looks at Medicare, finds that Miami is $16,351 per patient and Grand Junction, Colorado is $5,873.

The link to the Meps data is:

The link to the main Meps data page on state medical expenditures is:

The Dartmouth study seems to be in part an artifact of Medicare reimbursement policies. There are other federal funding programs to medical providers, e.g., Rural Health Flexibility Grants and others, etc., and these vary in per capita amount for each state according to federal law and formulas. Dartmouth only looks at Medicare expenditures and does not include other federal medical reimbursement and funding programs. It does not look at total state medical expenditures.

Complex and strict rules limit what can be included in Medicare reimbursement, but as a general rule, double billing is prohibited. The differences that Dartmouth sees could be that some states get less non-Medicare funding and bill a higher share of costs through to Medicare, while the other states bill the federal government for part of the same costs under other federal programs. Medicare expenditures are higher per procedure in some states than other states, but the total costs of the two procedures could be identical. It could all be due to federal reimbursement laws and rules.

Posted by: MiltonRecht | June 15, 2009 4:05 PM | Report abuse

The problem with the MEPS data is that old bugaboo, mean vs. median.

As a result, MEPS shows a median expense that is just about 50% of the mean expense for the US.

That is not surprising. What this means is something we know already – total health care expenses are based on a very large contribution from a minority of people who consume most of the care: the old and the sick, and a small percentage of people who have bad luck.

That is why insurance works, for one thing.

The Dartmouth Data, on the other hand, look at mean costs for Medicare. These are much higher, since they are mean rather than median, and are also higher because the Medicare population has more old people and more sick people.

The Dartmouth people use Medicare because it is easily accessible and accurate, while data from private insurers are less so. They make the assumption that health care management, while more expensive for the Medicare population, generally follows the same pattern. If management patterns lead to a lot more tests and procedures in the Medicare population, it is likely that the same management patterns lead to more tests and procedures for all patients, since the same patterns probably are followed.

And that is why the Dartmouth Data are very useful, and the MEPS data less so.

Even if this were not the case, since Medicare accounts for about 25% of health care spending and about half of government health care spending, information leading to better management patterns for Medicare would be useful all by itself.

Posted by: PatS2 | June 15, 2009 4:40 PM | Report abuse

My son had nasty, lingering, flu-like illness a couple weeks ago. We finally took him in to see the ped, just in case it was bacterial and an antibiotic could be prescribed.

The ped believed it was viral, but she (unbeknownst to us) called in a prescription for an antibiotic (which we did not use or pay for.)

She also decided to test him for pertussis (whooping cough), which he had been vaccinated against. After she did the test, she then told me that it was a long shot, but she wanted to rule pertussis out.

The bill just came - our statement said the test for pertussis costs $350 - for the test alone. Thanks to the insurance company write-off, we're being billed just $50.

Ped visit cost more than $100.

So more than $150 out of my pocket to find out it was a viral infection that needed to run the course.

I look at the costs of this visit as a showcase for what is wrong with our health care system.

Costs are inexplicable. A lab that charges $350 for a test is willing to settle for just $50. That's ridiculous. The test can't possibly cost $350 at all - so why set that value for it?

Tests are ordered that make no sense whatsoever. Antibiotics are prescribed for viral infections that cannot be cured by antibiotics.

The idea that our problems result from the fact that we pay so much to keep old people and the unlucky few younger people who get sick is absurd.

Our problems result from a system so focused on profit that the actual point of health care has disappeared. We don't really care about getting people healthy any more - we just want the money to keep rolling in.

Posted by: anne3 | June 15, 2009 5:11 PM | Report abuse

Looking at the MEPS data again, I am pretty startled that a reputable organization didn't bother to see if their numbers made sense.

I could also speculate that instead of mixing up mean and median, their sampling could have been extremely messed up in some way.

Any way you cut it, the numbers are way off. In 2007 the mean cost of health care for every man, woman, and child in the US was $7700. Their numbers are from 2006, but still are not sensible. The numbers are even higher today.

You can do a rough double check on the math yourself: total cost of $2.5 trillion, total population a bit over 300 million. That makes the cost per person over $8000 now.

I still suspect that mean vs. median confusion is the most likely explanation, since that would fit.

Posted by: PatS2 | June 15, 2009 8:45 PM | Report abuse

"As the graph atop this post shows, the cost of health care varies enormously between states, but there is no obvious connection between expensive care and better outcomes."

The graph says 'composite measure of quality of care' not 'outcomes'. Do you know for a fact that actual outcomes were measured for that graph, instead of just the usual easily-measured actions such as 'gave aspirin for heart attack' and 'gave flu shot [or pneumonia shot]' that are often used as 'quality measures'?

Digging into the links at Skinner's post, overall mortality for El Paso [low-spending area, and one we presumably ought to emulate per Atul Gawande] is worse than for McAllen [high-spending area]. Same for Florida: mortality is higher [worse] in the low-spending north and lower [better] in high-spending Miami.

Granted, the spending disparities are still huge compared to the small differences in mortality, and thus the topic of just how much do we want to spend to get a little more life? is still up for debate, but the Dartmouth team's conclusion that more care is bad for you is at best simplistic, if not downright wrong.

As for the much-touted Mayo Clinic model for reducing costs, I find it amusing that Mayo is apparently whining about how much they're losing on their Medicare patients:

Posted by: hipparchia | June 16, 2009 12:51 AM | Report abuse

The comments to this entry are closed.

RSS Feed
Subscribe to The Post

© 2010 The Washington Post Company