McAllen and El Paso Redux: New Evidence from the Insured Under-65 Population

Last year, Atul Gawande wrote in the New Yorker about the remarkable differences in health care spending for two Texas cities: McAllen and El Paso.  In 1992, according to the Dartmouth Atlas, the two cities were essentially identical with respect to per capita Medicare expenditures.  By 2007, McAllen’s spending had surged, with overall expenditures nearly twice as high as in El Paso.  Dr. Gawande visited the two communities, and brilliantly documented a culture of entrepreneurship among McAllen physicians that seemed to explain their elevated rates of hospital admissions, end-of-life care, and home health care.

But what about the under-65 population?  Dr. Gawande spoke with two independent firms about their measures of under-65 utilization, and found generally higher rates in McAllen.  My colleague Thomas Bubolz studies the under-65 Medicare population – primarily people on Social Security Disability Insurance — and his preliminary results also point to much higher utilization in McAllen compared to El Paso.   Another study using national data by Michael Chernew and colleagues (here) found a strong positive correlation between utilization rates for Medicare and the under-65 population insured by large firms.  (That they also found a negative correlation between Medicare spending and the negotiated price per procedure in the under-65 population points to another source of regional variation: market concentration.)

So when Luisa Franzini and Osama Mikhail, professors at the University of Texas School of Public Health, first offered me the opportunity to work with them using Blue Cross-Blue Shield data on under-65 spending in Hildago (McAllen) and El Paso Counties, I had strong expectations that we’d end up with  pretty much the same result.

I was wrong.  In a recent Health Affairs article, we found that, on average, overall spending per patient in McAllen was about 7 percent below that in El Paso.  Granted, we found the familiar Medicare utilization patterns among people over age 50: McAllen admission rates were 89 percent higher than those in El Paso, and overall expenditures 23 percent higher.  But outpatient visits and spending were lower across the board in McAllen, as was total spending for those under age 50.  What was going on?

We ruled out a variety of hypotheses – that the Medicare differences weren’t measured properly or biased by lack of illness or price adjustment (they’re not).  Indeed, the over-65 population in McAllen, comprising many retirees attracted by warm weather and low living costs, are a remarkably healthy group with mortality rates lower than those in Provo, Utah.

Some of the difference is explained by types of disease: there is virtually no home health for privately insured patients under age 50, and end-of-life care is a rarity.  Typical hospital admissions for the younger population — births and automobile accidents – are more urgent and entail far less discretion on the part of the physician. But when the diseases are similar, such as circulatory disorders, the data are at least consistent with another story:  physicians behave differently depending on who is paying the bill.  Blue Cross – Blue Shield includes preauthorization requirements and utilization review, but Medicare does not.   This gives a different interpretation to the regional variations story – it’s not just about enthusiastic physicians in a region doing more for all their patients, it’s about health care providers who are most enthusiastic when Medicare is paying the bill.

While the evidence supporting this hypothesis is circumstantial, the results are consistent with another recent study by Tomas Philipson and colleagues suggesting that private insurance payment mechanisms attenuate variations.  And importantly, we cannot generalize from a sample of two Texas counties to say anything about overall variations in the under-65 population.  Certainly, Richard Cooper’s work suggests as much unexplained variation in health care expenditures for the under-65 population (and for non-Medicare spending) as the Dartmouth Atlas finds in the Medicare population.  We hope that our ongoing research looking at all Texas counties will shed light on this question.

Despite spending more than 17 percent of GDP on health care, we are at a loss to even document variations in prices and quantities of services provided in the under-65 population.  This is because of a paucity of data; private insurance companies are hesitant to release data about their negotiated prices, and population-based measures of utilization can be problematic.  But it is also important not to become too distracted by our ignorance of under-65 regional insurance markets.  The key financial threat facing the U.S. federal government today is out-of-control Medicare spending growth.  Whether Medicare expenditures are positively or negatively associated with under-65 spending is less important for the federal deficit than is developing new policies to contain excess Medicare cost growth.

Related stories:

Categories: Uncategorized

Tagged as: , ,

24 replies »

  1. @ Matthew Holt,
    What a coincidence, on December 7, Mr. Skinner admits to major flaws in the Dartmouth Atlas conclusions citing absence of all-payor data in drawing conclusions. This of course is something dozens have commented on for over a year.
    And now on December 15, Dartmouth announces a pioneering collaboration with other provider organizations to seek insights from all-payor data.
    Well Matt, and those who buy Dartmouth’s spin…the only thing pioneering about this initiative is that you would to have been lost on a wagon train crossing the continental divide to not be aware of the enormous collaborative efforts taking place using all-payor data over the past 20 years in this country.
    A remarkable number of commercial firms, data commissions, hospital associations and agencies and a variety of data collaborative have been sharing all-payor data and effectively using it to define best demonstrated practices, best clinical outcomes and measuring patient safety, quality, outcomes and patient experience. Many of these programs preceded the IOM’s findings and some were instrumental in inciting this further research.
    So if you think that this “thesis” is anything new, go back to pre-dot-com and collect $200. This admission of error by Dartmouth combined with a two decade late start on using all-payor data puts these folks at the back of the bus for experience and reputation. Candidly, it is just this fact that had so many of us proving that the Dartmouth Atlas thesis was completely flawed…yet their spin continued and folks like you are still drinking their Kool-Aid.
    I call for a restatement of the litany of claims made by these folks now that they have been newly enlightened. I also call for an apology to the media and the analysts who tried to educate them about their aberrant findings and their defiant and un-science-like demeanor. Absent that, I don’t think they have any credibility nor does their research hold water with these admissions. Good luck to the providers who have hitched their wagon to that horse.

  2. “End of life utilization” isn’t even a type of utilization. It’s only defined by the fact that someone died at some future date. There is no billing code for, or clinical service (except maybe hospice) that is for ‘end of life’.
    But more important, Dr. Skinner now explains that the regional variations in spending are insight into, well, that there are regional variations. In his post now he explains that what high spending regions share is as follows: “But the common feature of these regions is a health care system that appears optimized to soak up vast dollar amounts from the U.S. Treasury.” In other words, what high spending regions share is that they are high spending. Sorta circular I think. Where are all those other conclusions Dartmouth once drew that have now been debunked?
    Spending driven by the number of specialists: oops, McAllen has very few specialists.
    NY is a high spending because of a culture of greed and intensity: oops, New York is only high spending if you fail to account for medical education dollars.
    Spending is driven by the number of hospital beds: oops, that’s a feature of urbanicity.
    High spending is a geographic phenomenon: oops, medicare spending is not correlated with non-medicare spending in the same region.
    So now we know what the Atlas shows according to its lead health economist. High spending is proof that, well, there are places with high spending, nothing more. No overarching pattern. And no obvious policy implication except “McAllen is a good place to start” (except it shares nothing with other high spending regions according to Skinner). Wow, good thing we got that nailed down.

  3. End of life utilization is a good snapshot of end of life utilization, period! And by the way, not only do all chronically ill patients eventually die…all humans eventually die regardless of their condition at death.
    Now that you have “found” all-payer data, maybe you will also look at beginning of life care, but please don’t jump to conclusions too fast as people’s whole lives are in front of them. Rationing takes on a different meaning when the NICU is involved.

  4. Regarding end of life care — it turns out that end of life utilization is a good snapshot of spending on the chronically ill, many of whom die. It is highly correlated with all kinds of utilization measures in the Medicare population. Whether these measures are also correlated with fertility rates (and hence the incidence of maternity stays among women in their 20s) or trauma rates among men in their 20s, is less clear.
    Great question about whether McAllen is like other high-cost communities. LA is high cost largely because of heavy use of hospitalization services, while NYC is high cost largely because they rake in so much through graduate medical education payment supplements, and let’s not even talk about Miami. But the common feature of these regions is a health care system that appears optimized to soak up vast dollar amounts from the U.S. Treasury. So understanding the McAllen puzzle won’t solve our health care problems, but it’s a good place to start.

  5. Other commenters have already noted many of the contradictions in Dr. Skinner’s posting. Medicare different than the under 65 but also the same for 50 and above. If that’s due to Blue Cross prior authorization then do they shut that off at age 50? Utilization in the Chernew paper is inpatient only – but this is inpatient and outpatient. The two don’t contradict each other.
    But most worrisome – Skinner seems here to be saying that ‘end of life care’ is a type of care, distinct from other types of care given to the younger insured. That assertion contradicts 20 years of the Dartmouth Atlas using end of life measures as a proxy for spending rates that they believe are ‘risk adjusted because everyone died’. But here in this post it appears that Skinner is saying that end of life care is actually a type of spending, and that it may differ from other types of spending. I’d love to know which one it is.
    BTW – am I the only one left wondering if McAllen is representative of other high spending places (which is why it matters) or unique (in which case it doesn’t really matter that much from a national health policy perspective). I’m still not sure.

  6. Tim,
    I share your view of how and why ‘the left’ invested in the Dartmouth thesis. I think that the problem with the Dartmouth work is its prescriptions not their descriptions. Buz Cooper, on the other hand, managed to get his own exhibits confused. There’s no ‘MAYBE BC is right’

  7. Tim,
    Since the vast majority of the “maybe” came from me, I would like to clarify that I am as “Left” as you can get without waving red flags.
    I am not vision-invested in anything and I don’t believe in grand designs. I do however believe in rigorous science and fact based conclusions. I don’t think we have enough science to draw definitive conclusions regarding the reasons for geographical variations in spending, or any variations between public and private payers, let alone assigning responsibility to providers, payers and/or patients.
    I do believe that these phenomena should continue to be objectively researched and I like the fact that Jon Skinner is acknowledging the uncertainty of the current state of knowledge.

  8. I wish I had a nickel for every “maybe” on this page. A rational person would conclude we don’t yet know why we see what we see in the data.
    MAYBE what Guwande said about a “money culture” wasn’t true. MAYBE the explanation for “geographic variation” is something else, like…poverty? MAYBE Buzz Cooper is right.
    The Left is vision-invested in the Dartmouth thesis, because they need to get control of the private money to make their grand design come true. The most conflicted voices in this debate are the wonks in the health care think tanks.

  9. Jon–
    A very interesting discussion . . .
    As you say at the end of your post, it is most important that we get a handle on reining in Medicare spending. That is driving the deficit. And the over-65 population is most vulnerable to overtreatment because
    a)they are more worried about their health b) they have more time to visit specialists, and undergo tests and procedures and c) Medicare does not require pre-approval before visiting a specialist.
    If pre-approval is needed and you go to your primary care doctor first to talk about your angina, he may recommend a change of diet and medication. If you go directly to a cardiologist, he is more likely to recommend what he knows how to do-expensive tests which may well lead to angioplasty w/ stenting.
    From the physician’s point of view it’s easier to go ahead and prescribe tests and treatments for Medicare patients because it’s usually quite clear what Medicare does and doesn’t cover. And Medicare covers virtually evenything that the FDA has approved.
    If your patient is covered by a private insurer, things are more ambiguous. Some insurers will cover certain treatments under certain circumstances; others won’t. You may wind up on the phone debating whether or not your patient needs that procedure with someone representing the insurance company. You may submit forms for reimbursement and then disdcover that you’re not going to be paid.
    The bottom line: Medicare may be making over-treatment too easy. Going foward, under the Affordable Care Act, if Medicare inflation continues to outstrip ordinary inflation (as represented by the CPI) by more than 1%, the Independent Payment Advisory Board is charged with finding ways to rein in spending. It cannot reduce benefits or hike co-pays.
    But it is likely to try to hone in on unncessary procedures. For example, it might require that two doctors sign off on the need for angioplasty. Today the doctor who makes the diagnosis in the cath lab is often the one who then goes ahead and does the procedure.
    Some insuers, including at least one Kaiser hospital in S. Cal. require a second opinion. I wrote about this in a post on the stent scandal today.
    Let me add that we don’t want Medicare to make things too complicated. It should continue to be as clear as possible as to what it does and doesn’t cover for patients who fit a particular profile. And that info should be well-organized on a its onlite site.

  10. Interesting Blog, even though this was not what i was looking for (I am in search of clinics like this one> http://www.ccsviclinic.ca/ )… I certainly plan on visiting again! By the way, if anyone knows of a good clinic that does CCSVI Screenings? BTW..thanks a lot and i will bookmark your article: McAllen and El Paso Redux: New Evidence from the Insured Under-65 Population…

    The Kool Aid was not safe to drink.
    Keep up the good work. Better information leads to better policy.
    Thank you for your admission, the spin was sickening.
    Thank you for the platform to inform and debate.
    Let’s get back to work, we have a lot to accomplish.

  12. The spending for the under 65 population is less because of the fact that utilization management mechanisms that exist for private insurers may prompt some physicians, who might otherwise overuse certain services, to exercise more restraint.
    The difference is spending can also be attributed to the type of diseases. the younger people are more affected by acute diseases and road traffic accidents, whereas the older people with chronic diseases involving life style modification

  13. “It may very well be….”
    And that is the problem in a nutshell. It may very well be that McAllen has a “money culture”, and it may very well be that those “greedy” doctors are taking advantage of Medicare, but it may very well be that things are different and we are looking at an altogether different issue.
    The HA article acknowledges that the HCCs in McAllen are higher than the national average. It may be that HCCs are manipulated, as they often are, but it may also be that the docs in McAllen are diagnosing earlier and more often and treating everything they find aggressively, not out of sheer greed, but to benefit patients.
    The article also acknowledges that mortality rates and hip fracture rates are slightly lower in McAllen. The article quoted in favor of the argument that aggressive treatment does not have better outcomes deals with people expected to die – end of life again – not necessarily with treatable conditions.
    So maybe, just maybe, the docs in McAllen are only guilty of spending an inordinate amount of money in trying really hard to help patients, and this obviously starts before Medicare kicks in. And maybe we are of the opinion that such small improvements are not worth so much money.
    So maybe we don’t have enough information to establish the existence of a “money culture”, other than Dr. Gawande’s anecdotal interviews. There may very well be a geographical variation, but perhaps it is rooted in good intentions, heroic measures and not spectacular enough success, but success nevertheless.
    I think Austin Frakt’s comment regarding the unfortunate scarcity of commercial insurance data is probably the most indisputable statement in this entire conversation.

  14. Insightful comments — I agree that we lack a full understanding of the interplay among private insurance prices and Medicare (and Medicaid!) reimbursements, for example those in the San Francisco market. Two particularly good studies of these interactions are the Chernew et al paper cited in the posting above, and J. Stensland et al (Health Affairs, 2010, 29(5): 1045-51); this latter paper looks at the association between private-payer profit margins and Medicare margins.
    I’m happy to acknowledge that the under-65 markets can affect how hospitals and physicians treat their Medicare patients — yet another source of regional variation in utilization not explained by patient health or preferences.

  15. “You missed the part that overall spending is lower. Outpatient costs for those under 50 was much lower.”
    I did not miss that part. One way to look at this is that McAllen is frugal with health care for the young and healthy. Health care costs go up as people age, and perhaps cost goes up higher than ElPaso either because McAllen was too frugal with its young, or because McAllen is more aggressive with its old, and perhaps this is why mortality is low.
    Perhaps too many instances of using “perhaps”, but the evidence that costs are correlated to type of insurer is a bit weaker than the evidence that costs are correlated with age.

  16. I think the lower cost for the younger population is probably due to efforts by Blue Cross to monitor and influence physician behavior. It should not surprise us that such efforts can be effective; Medicare does not do a very good job of this monitoring and influencing physician practice

  17. “I would have understood the argument if spending shot up significantly at the exact age of 65, but it does not. It seems docs in McAllen are just treating older people more aggressively as they age, regardless of the insurer.
    Perhaps that has something to do with having “mortality rates lower than those in Provo, Utah”? Or am I missing something here?”
    You missed the part that overall spending is lower. Outpatient costs for those under 50 was much lower. Also, it is very easy for docs to know what kind of insurance patients have. I know of groups that make decisions about the procedure performed based upon insurance.
    Jon- Reinhardt believes, and I am leaning towards the belief (if I am understanding him correctly), that in many ways private insurance costs drive the system. While I agree that we need to concentrate on Medicare, how much divergence between Medicare and private insurance reimbursement is tolerable?
    Steve Sisson

  18. “Particularly every health plan in California thinks that Sutter Health is the most expensive hospital system on earth, yet the Dartmouth data shows Sacramento (which Sutter dominates) to be a model of low volume & therefore low cost Medicare.”
    Matthew – My understanding is that insurers consider Sutter to be expensive based on their prices per procedures and their per diem rates as compared to competitors. Medicare dictates what it will pay whereas insurers have to negotiate their rates. Separately, I have no idea whether Sutter’s Medicare business as a percentage of its total revenue or inpatient bed days is significantly different from other competing hospitals in the region. Does anyone know?

  19. “Granted, we found the familiar Medicare utilization patterns among people over age 50: McAllen admission rates were 89 percent higher than those in El Paso, and overall expenditures 23 percent higher.”
    I assume these over 50 folks don’t have Medicare. So how do we reach the following conclusion?
    “it’s about health care providers who are most enthusiastic when Medicare is paying the bill”
    I would have understood the argument if spending shot up significantly at the exact age of 65, but it does not. It seems docs in McAllen are just treating older people more aggressively as they age, regardless of the insurer.
    Perhaps that has something to do with having “mortality rates lower than those in Provo, Utah”? Or am I missing something here?

  20. Skinner and colleagues have surprised themselves and me here, and he’s even left in a very polite reference to Buzz Cooper (the premier “Dartmouth denier”)
    However, slightly more anecdotal data suggests that similar large variation depending on who is paying exists elsewhere. Particularly every health plan in California thinks that Sutter Health is the most expensive hospital system on earth, yet the Dartmouth data shows Sacramento (which Sutter dominates) to be a model of low volume & therefore low cost Medicare.
    On the other hand, it’s really hard to imagine that providers are so cynical in their ability to treat patients differently based on payer status–they tend not to know between plans, although I guess it’s easy enough to tell by age whether someone is on Medicare or not.
    But I echo Skinner’s end thesis–lets get ALL payer data out into the open, and get the whole picture everywhere.
    Of course now that we have “government run health care” that should be easy….but I fear that the chance to insist on that level of disclosure was lost when the ACA was passed.