Making Sense of Geographic Variations in Health Care: the New IOM Report

Since 1973, when Jack Wennberg published his first paper describing geographic variations in health care, researchers have argued about both the magnitude and the causes of variation.  The argument gained greater policy relevance as U.S. health care spending reached 18 percent of GDP and as evidence mounted, largely from researchers at Dartmouth, that higher spending regions were failing to achieve better outcomes.   The possibility of substantial savings not only helped to motivate reform but also raised the stakes in what had been largely an academic argument.   Some began to raise questions about the Dartmouth research.

Today, the prestigious Institute of Medicine released a committee report, led by Harvard’s Professor Joseph Newhouse and Provost Alan Garber, that weighs in on these issues.

The report, called for by the Affordable Care Act and entitled “Variation in Health Care Spending: Target Decision Making, Not Geography,” deserves a careful read. The committee of 19 distinguished academics and policy experts spent several years documenting the causes and consequences of regional variations and developing solid policy recommendations on what to do about them.  (Disclosure: We helped write a background study for the committee).

But for those trying to make health care better and more affordable, whether in Washington or in communities around the country, there are a few areas where the headlines are likely to gloss over important details in the report.

And we believe that the Committee risks throwing out the baby with the bathwater by appearing, through its choice of title, to turn its back on regional initiatives to improve both health and health care.

What the committee found

The report confirmed three core findings of Dartmouth’s research.

First, geographic variations in spending are substantial, pervasive and persistent over time — the variations are not just random noise. Second, adjusting for individuals’ age, sex, income, race, and health status attenuates these variations, but there’s still plenty that remain. Third, there is little or no correlation between spending and health care quality. The report also effectively identifies the puzzling empirical patterns that don’t fit conveniently into the Dartmouth framework, such as a lack of association between spending in commercial insurance and Medicare populations.

The committee also confirmed earlier work by Harvard investigators showing that, for the commercially insured population, variations in the prices paid by private health plans explain most of the variations in private insurance spending.  The committee deserves considerable credit for deepening our understanding of this irrational world of pricing commercial health care services.  Yet as the report finds, even in the commercially insured population, there are substantial differences in utilization rates across regions.  We would therefore argue that for commercial populations both price and utilization deserve attention, especially because in many regions, avoidable utilization may be easier to address than price.

It is Medicare spending growth, however, that represents arguably the greatest risk to the financial health of the U.S Treasury, and in Medicare, variations are almost entirely the consequence of utilization of services, not prices.  The report finds that the single largest component of the variation in Medicare spending across regions that remains after risk and price adjustment is due to post-acute care (including skilled nursing facility services, home health care, hospice, inpatient rehabilitation and long term acute care). These services have also been a major source of growth.

But this focus on post-acute rather than acute hospital and physician services misses the key point that dysfunctional regional health systems are characterized both by hospitals providing fragmented and expensive care and by a large and thriving post-acute care sector ready and eager to absorb the discharged patients.  For example, Joan Teno and colleagues at Brown University have established the strong association of inpatient treatments with no medical benefit, such as feeding tubes for people with advanced dementia, with high rates of regional resource use.

Which brings us to…..

The IOM committee’s policy recommendations: Where they hit the mark …

The committee makes five policy recommendations — and we agree with all of them.  First, they call for making more and better data available, on both Medicare and commercial populations.  Second, they recommend that CMS continue to test new payment models that encourage clinical and financial integration.  Third, they call for timely and iterative evaluation of current and new payment reforms so that improvements can be made to the models.  Fourth, they call on Congress to grant CMS the flexibility to accelerate the transition to value-based payment models as successful approaches emerge.

The fifth recommendation focuses on whether Congress should adopt a geographically based payment adjustment. When the committee was first mandated by Congress in the midst of health care reform in 2010, congressional members from regions with lower costs espoused a “Value Index” in which Medicare would reward low-spending regions with higher reimbursements, at the expense of high-spending regions. The committee concluded that payment mechanisms should not be tied to region, but instead targeted to individual providers, rightly criticizing the Value Index approach as not providing institutions and systems with the right incentives to reduce costs and improve quality.

… and where they fall short:  Geography does matter

We believe, however, that the committee, by subtitling the report “Target Decision-makers, Not Geography,” will confuse the media and casual readers (for example, those who don’t make it to page 3-3 in the full report) by appearing to cast doubt on the promise of geographic and regional efforts to improve the quality and efficiency of U.S. health care.

As the late Nobel-Prize winning economist Elinor Ostrom has emphasized, successful management ofcomplex social problems can best be achieved through sustained collaboration among diverse stakeholders, often across traditional political boundaries.  She demonstrated that cooperative agreements are often the most effective approach to solving the kinds of problems we face in health care. Among these are the natural instincts of physicians and hospitals within local health care systems to protect their financial health by expanding capacity and defending market share, whether by opening new cardiac centers when the one at the nearby hospital is perfectly adequate, or by buying proton accelerators that will be used to treat conditions where they offer no demonstrated benefit.

The rationale for a geographic focus on health care reform is strong:  the factors that determine population health are largely local, rooted in the environmental, social, economic, and behavioral determinants of health.  Many of the factors that influence health care quality and costs are also local, including local supply, pricing behavior, and the relative emphasis of providers on profit.  For example, in the widely cited New Yorker article by Atul Gawande, Medicare utilization in McAllen was found to be nearly twice as high as that in another Texas border town, El Paso, despite the existence of multiple hospitals in both McAllen and El Paso, nearly identical Medicare prices, and common Texas malpractice laws.

Many regional multi-stakeholder initiatives have been established.   Although most began with a focus on quality, many are beginning to act more broadly to both improve health and lower costs: Three examples include Pueblo Colorado (Regional Triple Aim),  Akron, OH (Accountable Care Community), and the Atlanta Regional Collaborative for Health Improvement (focused on driving provider transitions to global payment, capturing savings, and reinvesting in strategic population health initiatives).

While the IOM Committee is exactly right to call for improved financial incentives for health care providers, we should also remember that both health and health care are local.  Geography matters.

Elliot Fisher, MD, MPH and Jonathan Skinner, PhD are professors at Darmouth’s Geisel School of Medicine and The Dartmouth Institute for Health Policy and Clinical Practice.  Fisher is a  principal investigators, and Skinner is a senior scholar of The Dartmouth Atlas Project.

Fisher, Elliot S. & Skinner, Jonathan S., Making Sense of Geographic Variations in Health Care: The New IOM Report, Health Affairs Blog, 24 July 2013. Copyright ©2013 Health Affairs by Project HOPE – The People-to-People Health Foundation, Inc.

24 replies »

  1. In being concise my words may not sound polite so I apologize in advance, it’s not intentional:
    1) Geography is, at best, an indicator
    2) A “zip code” is not a causal factor
    3) Chasing geographic variation = running in circles = getting nowhere
    4) Geographic variation statistically comes out as an instrumental variable approach
    5) Instrumental Variables introduce bias (search pubmed for Pearl, Hernan, others), and while you may trim some error out of the model, due to bias you won’t know if you’re even aiming at the correct target
    6) Decision making is more closely aligned with causal factors
    7) Find the causes of the variations, not the indicators (correlations)

  2. i was getting the same symptoms.– David Andreatta (@david_andreatta) “Some scary headlining – but 23 deaths, and Lack poke-checked the puck away from Brandon Saad. Versteeg opened the scoring at 7:49 on a deflection of Kane’s shot from just inside the right point.Remarkably,C.?That??s the way the world has worked since people first discovered the utility of information and the power of secrecy. having agreed verbally with the reporter that the conflict of interest in his department is egregious,What is the city’s most famous moment in football history? so no need to be worried about competition for attracting an NHL team.

  3. Doctors that accept cash payments are likely to be popular in areas where costs of health care are high as well. This puts the money where it needs to go in more realistic quantity. Will that lower the costs in such a region, or will it drive some institutions to hike their prices further? Can we get a study on that?

  4. Hmmm. How do the blind get onto peer review committees? Are the members elected by the local physicians? Appointed? Are they volunteers?

    What would be the best way to ensure a quality peer review committee?

  5. My physician friend goes to 4 hospitals in the same region. At one, friend was criticized for ordering subcu heparin in a patient with a recently drained subdural hematoma and GI bleed, and yet, at another hospital, with a similarly at risk patient (IC Bleed and iron deficiency), there was repeated insistence (augmented by the EHR CDS) that heparin be used to prevent DVT is a 91 year old frail patient.


    Cultures are different in hospitals and when the blind lead the blind, variations occur and become ingrained.

    And when the blind are sitting on peer review committees, you tend to do what they demand or face sanctions and ridicule.

  6. As Dr Reinhardt implies, part of the problem is that Medicare does not have a hard budget…i.e., when the annual allocation is spent, there is no more money available.

    Instead, Medicare is still an open-ended, claims-driven, self-reporting system with a highly graded and easily-manipulated fee schedule.

    If Medicare did have a hard budget, then payments could be ratcheted down right in the middle of a fiscal year, in a given state.

    If Florida was home to 7% of Medicare recipients, Florida would receive 7% of Medicare dollars. If total claims were running well above 7% in the middle of the fiscal year, then each payment thereafter would be reduced.

    This would cause the medical community to police itself. Honest providers would turn on the McAllens in their midst.

    Of course Congress would hate this,, and the pubiic relations outcry would be deafening.

    But that is what grown-up fiscal budgeting is all about.

    (see also Joseph White on entitlement vs. bureau budgeting.)

  7. It is as simple as “monkey see, monkey do”. The role of behemoth medical centers in the run up of costs and variation ought not be neglected.

  8. Others have noted the significant differences in physician practice patterns from one state or region of the country to another. If it were up to me, I would attack this in two ways.

    First, I would publish lots of data by state and region covering such statistics as hospital inpatient bed days and ICU bed days per thousand Medicare beneficiaries, numbers of common surgical procedures like back surgeries, hip and knee replacements, CABG’s, stents, etc. per thousand Medicare members, skilled nursing facility bed days and home healthcare visits per thousand people and several others. Ensure that the mainstream media has full access to the data and encourage them to write about significant state and regional differences and pointedly ask doctors what accounts for them.

    Second, I would let Medicare Part B premiums vary by state. Currently, the Part B premium is uniform across the country and is set to cover 25% of the program’s costs. I think the Medicare beneficiaries from each state should pay a premium that covers 25% of the program’s cost in that state, perhaps with a cap of $150 per month while residents of lower spending states would pay less than they do now but with a floor of $75 per month.

    The idea here is to create countervailing power and pushback against excessive physician driven utilization in high spending states. To the extent that some of the excess cost might be patient driven, let’s expose that as well.

  9. This is a complex issue that probably calls for different strategies to approach different facets of the problem.

    For example, it’s easy to understand why hospitals want to expand in areas that are profitable and well reimbursed. This includes cardiac care and new treatments like proton beam therapy. As long as hospital Boards of Directors tie CEO and CFO incentive compensation largely to growth in revenue, profit and market share, this trend will persist. For priorities like cost-effectiveness, patient safety, adhering to evidence based guidelines and protocols and patient satisfaction to gain more traction, the metrics that drive senior management bonus compensation will have to change in a way that values they newer priorities much more highly. The proliferation of proton beam centers could also be mitigated by adopting reference pricing so that it wouldn’t get reimbursed at a higher level than IMRT or other equally effective treatments.

    Greater use of hospice care can be a good thing if it comes at the expense of futile ICU care for dying patients. More aggressive efforts to get patients to execute living wills and advance directives would be helpful here and the information should be stored on a registry so it’s available to doctors and hospitals when needed. It would also be helpful if doctors, especially oncologists, were more honest with patients and family members regarding the prognosis and make sure that they fully understand what they are signing up for if they opt for aggressive treatment.

    With respect to skilled nursing, home healthcare, and inpatient rehabilitation, these services are generally not painful or invasive and are sometimes even pleasant. Profit margins for service providers are often quite high and the number of visits often bumps up against the maximum that insurance will pay for. Some of this post-acute care utilization could probably be mitigated by a move toward bundled payments for an entire episode of care so providers will have less incentive to provide more treatment than is medically necessary.

    Medicare can dictate prices but it has a poor record of controlling utilization. It needs to be more than a big dumb payer.

  10. Uwe, a great question. I’d suggest two basic approaches. The first is the collaborative efforts noted in our blog on “multi-stakeholder initiatives.” For example, the goal of the ReThink Health project, funded by the Fannie E. Rippel Foundation, is to bring into negotiation with health care providers those groups who would benefit from a truce in the medical arms race, such as consumers and employers, each of whom has to pay for expensive and fragmented care. Elliott Fisher has been working with the ReThink group and I will defer to his greater expertise on this approach

    Another is to move away from price mechanisms to quantity mechanisms keyed to the local population. I’ve grown cynical watching clever new pricing and reimbursement policies end up being “gamed” by providers who use them to make even more money. Why not consider approaches to reducing capacity at the local level directly, for example (as Elliott has suggested) by extending the “cash for clunkers” program from old cars to old MRIs and CT scanners? Granted, certificate-of-need programs have not worked well in the past. But it may be time to rethink and even expand their use. Some years ago Elliott and I suggested (half-seriously) that physicians require a specialty-specific “medallion,” like those used for taxis in New York City, to set up practice in a certain region. Thus new cardiologists would be more likely to set up practice in under-served areas, rather than paying top dollar for the right to practice in Beverly Hills. This approach may not receive the endorsement of the AMA (!), but environmental economists have understood this general point for years: regulating quantities at the local level can sometimes be superior to fiddling with prices.

  11. A value index would most assuredly make the problem worse, and would in no way represent an “approach to fixing the problem.” Pay the region less and they will do more of what they already do to make up the difference. You can’t change human nature, and to suggest that your nature is different only makes you to appear non-human. Policy that ignores how humans respond to incentives and disincentives is not policy, it is lunacy.

  12. When the Titanic is sinking do you argue about the way the cabins are divided among the rich and poor? The system is dying because of a very basic problem: we pay for far too much care that does not make people healthier and we reward doctors/hospitals/companies who are most imaginative in how they can game the system. If we brought everyone down to the lower cost of care in the US, would it still cost too much? I think that unless we find ways to stop rewarding over-use and over-intervention, we won’t make much of a impact on getting us out of the economic death spiral we are in. In the end, it seems like the whole ship is sinking, not just the McAllen Texases out there.

    The bottom line is that this seems to me to be an argument among academics over the rhythm strip of a dying patient. Regardless of which one is right, the patient will end up having the same fate. I don’t think it’s wrong to have the discussion, but it does seem to miss the more basic problem of gross overspending that is encouraged by our payment system as a whole.

  13. Jon,

    Could be more explicit on concretely what policy that does not ignore geography would look like, given you do not like fee-updates linked to geography?

    You leave me a little confused here with your nuanced comment.


  14. I believe the Dartmouth studies — and especially Jack Wennberg — have done the nation a great service by smacking it in the face with the 2 by 4 called the Dartmouth Atlas. I am both surprised and dismayed how long the Congress and — let’s face it, private grant makers — had to be smacked by that 2 by 4 to finally ask someone to probe beneath the Dartmouth data (although there was a fascinating paper by Hadley, Zuckerman et al doing just that.)

    I recall supporting something like geographic indexing of the payment updates when serving on the Physician Payment Review Committee. The idea was to impose on, say, Florida a Medicare global budget for the coming year, forcing providers there into a zero sum game. The theory was that this would put pressure on the cost-inefficient providers and the locals would clean up the market. The idea was DOA in Congress.

    Germany did something like that with physicians during the 1990s.

    But my hunch is that, had I been on the IOM study panel and seen all of their data and listened to their discussions, I would support the panel’s conclusions and recommendations. We now have much better micro data than we did during the 1990s, and cheaper computational capacity.

  15. This is a great discussion, thanks so much for your contributions. Regarding Vik Khanna’s hypothesis regarding the importance of local culture and physician practice patterns, a paper on this topic is expected to come out on Monday as an National Bureau of Economic Research (NBER) working paper: “Physician Beliefs and Patient Preferences: A New Look at Regional Variation in Health Care Spending,” written with David Cutler, Ariel Dora Stern, and David Wennberg. We use physician surveys with detailed clinical vignettes, and find that, as Vik Khanna hypothesizes, regional physician beliefs in the use of treatments outside of clinical guidelines appear to strongly predict overall Medicare expenditures. (If you can’t find it next week, email me at jon.skinner@dartmouth.edu and I’ll be happy to send a copy.) This study doesn’t have much to say about how these beliefs came about, but I also agree that residency or medical school training is likely key. There is some ongoing research on this question by a former colleague, but it hasn’t yet been published.

    Regarding John Irvine’s questions regarding the IOM recommendations on payments linked to region, and whether that casts doubt on the Dartmouth work: The IOM study clearly established that regional variations are large and important. Thus we view their conclusions as supportive of the Dartmouth work.

    The debate is over how to fix the problem. Some have suggested the “Value Index” whereby low cost regions would get higher reimbursement rates (and conversely for high cost regions getting lower reimbursement rates). Elliott Fisher and I have never liked this approach to fixing the problem — nor does the IOM. So we entirely agree with their recommendations. But this doesn’t reject the existence of geographic variations, only one specific policy approach.

    Where we differ from the IOM report is that we believe policy reforms shouldn’t ignore geography. If health care (like politics) is local, then solutions that arise locally are likely to be an important component of efforts to tame the enormous degree of waste in US health care.

  16. My sense is that the IOM Report is a pretty strong rejection of the idea that payment should be linked to geographic variation in health care.

    I’m not sure what this means, however.

    Does it mean that the work done by Dartmouth is falling out of favor?

    Are we saying that more study is needed to do something like this effectively?

  17. “ that both health and health care are local.”

    Amen to that statement. I would like the authors to comment on the role that medical education plays, and I offer a vignette to set the stage. A friend is in charge of capital projects for a major utility company. He is a former naval officer whose job was design and maintenance of nuclear reactors on surface ships, educated at a Big 10 university. Despite this exemplary background, he is often subject to (mostly good-natured) derision and some serious criticism from his colleagues when he suggests ideas, perspectives, and actions that deviate from the way his co-workers were taught at the large SEC university many of them attended. The “we don’t do things that way here” mentality speaks to the parochialism that still permeates a lot of American culture.

    Is the crux of the problem here medical education? Having been through clinical education myself (physician assistant, internal medicine), I remember many of my preceptors and residents holding very strong opinions about the standards of care in their community and a desire to do things as higher ups (i.e., their administrators, teachers, and evaluators) expected. Does manipulating payment mechanisms (clearly useful on some level) aim to fix one end of the problem without remedying the other? Does the way that clinicians do things in a community reflect not just desire to maximize reimbursement but the inculcation and transmittal of a specific set of values through medical school, to residency, and then out into the commercial marketplace?

    I would be interested to know, for example, about the dispersal of suboptimal clinical approaches to certain problems. If, in region X, outcomes are lower and costs higher for a particular problem and the issue appears to coalesce an identifiable set of procedures or treatment approaches, what are the demographics of that issue? Where are those ideas grounded in that region? Do clinicians who train in that area but leave still adhere to what they were taught or do they adapt to the expectations of their new locale? On the flip side, what of clinicians who move into the region? Do they fold and do as their new peers or risk sticking out a like sore thumb and import beliefs and training from whence they came, which might risk both professional and personal alienation? What about clinicians who stay where they trained (the most common circumstance, I would guess)? How often do they change their clinical practices, and is reimbursement really the only — or even the principal — driver?

    We like to think that medicine is so sophisticated that everyone learns the same thing the same way and operationalizes that information in, if not identical, certainly similar manners. It might not be that straightforward because, not matter what our station in life, we all try to fit in wherever we are.

  18. After a quick read of the IOM Report, I’d say this is a lot more than the “an asterisk.” This is a rejection of the idea that this research should drive policy. The subtitle was not an asterisk. It was an exclamation point.

  19. Those suggestions seem amorphous and unlikely to generate improvement. Isn’t this study the kind of research you’re saying should take place to generate better suggestions? When does someone stand up and actually say that certain sectors are non-value-add and may just need to go away.

    These recommendations might as well be saying “Let’s commission another study!” Academics keep writing these studies but nothing changes.

  20. Geography matters not because there is something magical about “place” that affects utilization patterns, but precisely because (local) politics and policy matter. (All the more reason to remember that the late Elinor Ostrom was in fact a political scientist!) Geographic targeting based on incomplete risk adjustment can easily play into the hands of local policy actors who are, naturally, I suppose, eager to squeeze as much out of the payment system as they can. This, I suspect, is the reason for the committee’s wanting to place an asterisk by geography.

  21. Thanks!


    “The rationale for a geographic focus on health care reform is strong: the factors that determine population health are largely local, rooted in the environmental, social, economic, and behavioral determinants of health”

    I could not agree more. “Geography matters”- big time and more than most of us are willing to admit!

    Dr. Rick Lippin

  22. “The Committee risks throwing out the baby with the bathwater by appearing, through its choice of title, to turn its back on regional initiatives to improve both health and health care.”

    Well, a little Kremlinology is in order then. What’s going on here? My best guess is that the 19 distinguished scholars – or several of them – feel that an asterisk is necessary here to make the point that geography should not be overstated.

    Why do you suppose they’d say something like that?