A thought experiment
This suggests I’ve not made something clear. I’ll try again. Aaron Carroll said it well.
A lot of it shows that people with private insurance do better than those with public insurance or those without insurance; that should not be a surprise. Most people (and most of your docs) would rather have private insurance than Medicaid. But would you really rather have no insurance than Medicaid? If so, that is everyone’s right. Don’t get the Medicaid.
He’s right. Medicaid is not mandatory.
Let’s dig deeper. You’re uninsured. Thus, with your own money you can select and pay for whatever care you can afford. You get the quality of care you choose to buy with the resources you have. That’s your lot.
You wake up one day in possession of a magic (or maybe evil) Medicaid card. Let’s presume it permits you to visit for no charge a small number of lower quality health care providers. You can still choose to leave the card at home and visit any provider you could before and pay the same prices you had been paying.
Will the quality of care you receive go up or down with the possession of the Medicaid card? This is an empirical question, but let’s first explore the theoretical possibilities. If it helps, replace “care” above with some other type of good like “food” or “clothes” and “Medicaid card” with a “discount card.”
Consider the options. If quality goes up (your outcomes improve) then we would believe that reducing your out-of-pocket price of care, even for lower quality providers, improves outcomes. We’d say, “Medicaid works!”
On the other hand, if quality goes down (your outcomes get worse) what can we say? What causes this? My best explanation would be that you are such a poor judge of your health care needs that you are seduced by lower out-of-pocket cost, Medicaid care and are harmed by its lower quality. Having access to cheap care induces you to use more care and more low quality care. Making care cheaper, but only for certain providers, actually makes outcomes worse. You’d be better off with no insurance because it imposes resource constraints causing you to receive less care overall and thereby avoid the low quality care offered by providers accepting Medicaid. (Do you believe this explanation? Can you suggest a better one without appealing to selection bias (I’m getting to that)?)
I bet you’d say, “Oh no, not me. I’m smarter than the typical Medicaid beneficiary. I would know not to get more care and, above all, to avoid low quality providers. I would not be seduced by the discount the Medicaid card provides.”
To say that suggests that there is something about the Medicaid population that is different from you. I bet you think you’re different than the typical uninsured individual too (provided you aren’t one). By the same token, it is reasonable to presume there are differences between Medicaid and uninsured populations as well. Some are observable and can be controlled for in a multivariate analysis. Some are not, requiring an instrumental variable analysis, exploitation of a natural experiment, or a randomized trial to obtain unbiased results. Note that the relevant differences pertain to individual characteristics, not those of the providers they visit. The selection of providers is an effect of the Medicaid discount (again, assuming you don’t know enough about your health care needs to make more informed decisions).
There are undoubtedly studies that consider Medicaid vs. uninsured outcomes using the random variations provided by the natural experiment that is Medicaid. Characteristics of the program vary by state and year, making it a perfect set-up for such an analysis of this issue. This second I can’t point to a study. But I know where to look. One place to start would be to examine the literature cited by Stan Dorn on Ezra Klein’s blog at the Washington Post (tinyurl.com/StanDorn), Harold Pollack on The New Republic’s The Treatment blog (tinyurl.com/HPollack), and by J. Michael McWilliams on this blog (tinyurl.com/JMMcWill).
That’s it. That’s my position, and it always has been. If you read carefully you ought to notice that I didn’t actually condemn or praise Medicaid. I didn’t actually say how it should be reformed. I just listed the possibilities. Which you believe depends on a combination of your personal views and your interpretation of the literature. What can actually happen depends on political forces so strong my opinion hardly matters.
Selection bias and the study of Medicaid
Aaron Carroll has some wise words to keep in mind when interpreting studies of the Medicaid or uninsured populations:
Insurance doesn’t equal care. Insurance can affect how likely you are to get care and how quickly you might get it. But any study that looks at insurance has to adjust for many, many other variables in order to get the true effect of insurance. …
Surgery is different than other types of care (like emergency care) in that it is harder to refuse. So it may be that the uninsured are getting care on a compassionate basis. Few would provide a screening mammogram or yearly colonoscopy to someone uninsured, however, and you would get that with Medicaid.
He also cautions against taking conference abstracts too seriously:
[L]ess than 45%of the research presented [at the 1998 and 1999 Pediatric Academic Society meetings] was published in a peer-reviewed journal in the next four to five years. So over half of what was presented at the meeting never was “really” published.
I’m not saying the results … aren’t valid. I’m saying I can’t tell. And neither can you, without more information. The peer review for a meeting just isn’t the same as for full publication. You have less time, different criteria, and almost nothing by which to judge the work. Ideally, meetings would stop publicizing abstracts as if they were full studies, but neither they, nor the press, seem likely to do so.
So, be wary of conference abstracts. Actually, be wary of peer-reviewed publications too. The strongest conclusions are based on a body of work collected in an unbiased fashion. It’s not uncommon for papers to draw conflicting conclusions. But if an overwhelming majority of papers in an area point in the same direction then it’s reasonable to think they’re on to something. Of course, methodological technique matters. It is possible that scholars are all doing something wrong and only a few recent papers actually get it right. So, it is no easy task to interpret the academic literature.
That’s a good segue to this interesting set of results from a recent paper in the Journal of Hospital Medicine, “Insurance status and hospital care for myocardial infarction, stroke, and pneumonia,” by Omar Hasan, E. John Orav, and LeRoi S. Hicks. Click on the following figure to enlarge and study it carefully.
Ignore model 2. Model 1 is adjusted for age group, sex, race, income, emergency admission, and weekend admission and for hospitals’ bed size, control, region, and teaching status. That sounds like a lot of risk adjustment. But model 3 has a lot more. It also adjusts for comorbidities, severity of principal diagnosis, and the proportion of uninsured and Medicaid patients in each hospital.
Now take a close look at the first set of results for acute myocardial infarction. The relative sizes of the odds ratio of uninsured vs privately insured and Medicaid vs. privately insured switch between models 1 and 3. A lot of the other results change quite a bit between the two models too. There are two lessons here. One is that risk adjustment really matters. The other is that we can’t be certain that even model 3 has enough. There may be unobservable characteristics that bias the results.
The only way to fully address the bias due to selection into the Medicaid, uninsured, and privately insured groups is to find a source of exogenous random variation. That could be supplied by a randomized trial or with sound instrumental variables techniques.
Another thing to note is that in the results presented, Medicaid patients have better mortality outcomes than the uninsured for stroke and AMI (model 3). So, from this study alone, one might be tempted to conclude that Medicaid isn’t as bad as some other studies may suggest. But again, should we make a big deal out of this cherry-picked study? No, we certainly should not.
Deathly Medicaid?
Never draw broad policy conclusions from one study. It could hurt your brain. Well it hurts mine, anyway.
Avik Roy writes,
Medicaid so severely underpays doctors—reimbursing them at 72 percent of already-stingy Medicare rates—that many physicians refuse to see Medicaid patients. Medicaid patients, in turn, fill up emergency rooms, where they delay the care of the seriously injured.
Now comes word, via a large study by the University of Virginia (h/t Joseph Colletti), that surgical patients on Medicaid are 13% more likely to die than those with no insurance at all, and 97% more likely to die than those with private insurance.
The policy implication is that we should make Medicaid as close as possible to private insurance, thereby making Medicaid surgical patients 97% less likely to die. Or, if that is too costly, we should instead make Medicaid more like no insurance at all and boost survival by 13%. Talk about bang for the buck!
One way this could make sense is if very little health care (such as that the uninsured might receive) is bad for you, a little bit of, perhaps low quality, health care (such as that Medicaid patients might receive) is very bad for your health, and a lot of perhaps higher quality health care (that the privately insured enjoy so much) is very very good for you.
Or maybe there’s a problem with the study.
Our Unpublished Response to McArdle
Unfortunately, the letter I drafted with colleagues in response to Megan McArdle’s March 2010 The Atlantic article “Myth Diagnosis” was not published in the magazine. Those that were did not make the same points we did (letters available online). Below is the text of our letter.
To The Atlantic Editor:
Megan McArdle’s March 2010 article, “Myth Diagnosis,” distorts the scientific record in asserting that, “Quite possibly, lack of health insurance has no more impact on your health than lack of flood insurance.” Citing a tiny fraction of the literature on this topic, she concludes that we should know far more about the relationship between health insurance and mortality before considering major reforms to the health care system. But we already know vastly more than McArdle lets on.
For example, she characterized one study, which did not find a decrease in mortality risk due to insurance, as “what may be the largest and most comprehensive analysis yet done of the effect of insurance on mortality.” That sounds as if this single study is determinative. Yet no study in a social science could be. In truth, that insurance and the access to care it facilitates improves health and reduces mortality risk is as close to an incontrovertible truth as one can find in social science.
Viewed as a whole, the body of evidence shows that this relationship is well established. Last year, comprehensive literature reviews conducted by the Institute of Medicine and published in the Milbank Quarterly concluded that the overwhelming majority of well-conducted studies have found important health benefits of insurance, including lower risk of mortality. In addition to quasi-experimental research, several observational studies by leading researchers that controlled for a robust set of characteristics have demonstrated a 35-43% greater risk of death within 8-10 years for adults who were uninsured at baseline and even higher relative risks for older uninsured adults with treatable chronic conditions, such as diabetes and hypertension. These and other relevant studies are described in three online summaries posted in response to McArdle’s article—by Stan Dorn on Ezra Klein’s blog at the Washington Post (tinyurl.com/StanDorn), Harold Pollack on The New Republic’s The Treatment blog (tinyurl.com/HPollack), and by J. Michael McWilliams on Austin Frakt’s blog The Incidental Economist (tinyurl.com/JMMcWill).
But McArdle did not make her readers aware of this body of evidence. Instead, she cherry-picked work that supported her conclusion, ignoring every study published since 1994 that is inconsistent with her argument. It is one thing to argue that we should reassess proposed approaches to health reform. It is quite another to misrepresent a body of work in support of that conclusion and further mislead readers that such work does not exist.
No one could object to The Atlantic‘s support for a wide range of opinion columns. But The Atlantic is a respected, widely read home to intellectually honest and rigorous journalism. One hopes that, before publishing an article like McArdle’s at a key juncture of the national debate over health reform, the magazine’s editors would have made sure that the article fairly reflected the available evidence. Sadly, McArdle’s article did not come close to meeting that standard.
Austin Frakt, PhD
Assistant Professor of Health Policy and Management
School of Public Health
Boston University
Stan Dorn, JD
Senior Fellow
Urban Institute
Jack Hadley, PhD
Professor and Senior Health Services Researcher
Dept. of Health Policy and Management
George Mason University
Aaron E. Carroll, MD, MS
Associate Professor of Pediatrics
Director, Center for Health Policy and Professionalism Research
Indiana University School of Medicine
Lisa I. Iezzoni, MD, MSc
Professor of Medicine, Harvard Medical School
Director, Mongan Institute for Health Policy
Massachusetts General Hospital
Will Lack of Insurance Kill You?
You sick of this issue yet? It’ll die soon enough. I’m not trying to keep it going. I’m just posting this for the record. (Posting things helps me keep track of them.)
Over on The Treatment blog, Harold Pollack does a first rate job parsing the issues. He reviews some literature and raises a few points not mentioned recently by others in this debate. His conclusion:
Does health insurance save lives? Almost certainly it does, though the size and the causal pathways of the protective effect remain uncertain. Should we continue epidemiological research to pin down these relationships and improve other, non-insurance strategies to save lives? Absolutely.
On this first money question, then, would universal coverage make people tangibly healthier? You betcha.
Then Pollack turns to the best possible, yet still reasonable, view of McArdle’s article. Unfortunately for her and The Atlantic she didn’t make these points in these ways and blew her own credibility by misrepresenting the evidence. So, I’m happy to let Pollack’s take be the stand in. Here it is:
In addressing the second question, McArdle’s skepticism deserves a more sympathetic hearing. Suppose we accept that universal coverage could save 22,000 lives every year. That’s a large number, but there are other ways to save thousands of lives that are much more cost-effective than expanding health insurance coverage. We systematically neglect these other opportunities.
… More than 400,000 Americans die every year from tobacco use, for example. A stiff increase in cigarette taxes (with the proceeds used to finance other needed tobacco control measures) would probably prevent more deaths than universal health coverage would.
… Ethicist Daniel Callahan muses that every American city includes gleaming hospitals and crumbling schools. That’s not sustainable or wise, even if our only goal were to promote population health. For this reason and others, controlling the long-term growth of health care spending is essential. Health care spending is already crowding out other investments required to address critical national needs.
Nothing I just said provides a good argument against health reform. The economic and health benefits of near-universal coverage are quite large. No wealthy society should allow people to lose their homes because they get sick. Health reform would be a historic achievement. Still, it is only one step we must take to create a healthier and more decent society.
As long a quote as that was it still doesn’t do justice to Pollack’s argument. His post is worth a full read.
A Second Lit Reveiw on the Effect of Health Insurance on Mortality
Another literature review on the relationship between health insurance and mortality and health outcomes has been posted on Ezra Klien’s blog. This one is by Stan Dorn, the author of the Urban Institute study that estimated 18,000 deaths could be blamed on lack of insurance. It is a nice complement to the review provided by Michael McWilliams.
In particular, Dorn goes further than McWilliams in his critique of Richard Kronick’s study upon which McArdle’s conclusions are largely based. This may interest readers of this blog, some of whom have asked for more critique of Kronick.
[T]he main point of Kronick’s study is that some of the earlier research may have overstated the effect of insurance on mortality by omitting important variables. Kronick’s study had its own problems because, as his paper alludes, he was not able to address a critically important methodological issue—namely, that people in poor health are more likely to seek health insurance, which obscures any positive relationship between health insurance and health status. Studies that adjust for this factor have found a statistically and quantitatively significant relationship between lack of insurance and increased mortality risk.
Note what has transpired here. McArdle has characterized Kronick’s study as “what may be the largest and most comprehensive analysis yet done of the effect of insurance on mortality.” That sounds very convincing, as if the Kronick study is the definitive word on this matter. In fact, no single study can be. There is no such thing in social science. Every study, including Kronick’s has some limitations. Even a large and comprehensive analysis can suffer from an important methodological limitation, as Dorn believes Kronick’s does.
Therefore, one needs to base conclusions on a body of work. And as Dorn and McWilliams have both found, among recent studies in this area the evidence is greater than three-to-one in favor of an insurance-health outcome link, including mortality. To reach her conclusions, McArdle ignored the entirety of the research in favor of a small number of studies unrepresentative of the whole.
Programming Note
With another post by McArdle the insurance-mortality debate continues and may do so for some time. If you are interested in it and this blog has been your sole source for references to the contributions of others (Klein, McArdle, Drum, etc.) then you should take some action. I’m not going to continue to go post-for-post with the other bloggers on this. I’ve already spoken my mind on the matter, and I’ve facilitated the blog version of a first-class literature review (link to relevant posts). From here on out it would just be correcting folks’ (sometimes willful) misunderstandings and repeating myself. I’ve got better things to do.
Therefore, if you want to keep up with this one then either (a) subscribe to the blogs of others involved or (b) subscribe to my News & Links feed (described on my Subscribe page and visible in the far right sidebar). As I see things of relevance to this topic I will enter them into that feed.
This is not to say with certainty I will not post on this again. I’m just not committing to participate in every round. After all, there is vastly more to health reform, health economics, and health policy than this issue. It’s gotten far more attention than it deserves already.
Clean Up
Apparently there were some technical difficulties with Michael McWilliams’ post that reviews the literature on the health consequences of lack of insurance. They may have only been experienced by Internet Explorer users (seriously, give Firefox a try; it really is better). In any case, the problems should be fixed now. So try again if you couldn’t read the full post earlier.
See also Ezra Klein’s latest on the topic. He’s providing a level of synthesis you won’t find here because it’s not my thing.
I don’t want to be too harsh, and I don’t want to imply that anyone is sitting around twirling their mustache thinking up ways to hurt poor people. But opposition to health-care reform (which is different than opposition to the people who would be helped by health-care reform) is leading to some very strange arguments about the worth of health-care insurance — arguments that don’t fit with previous opinions, revealed preferences, or even the evidence the skeptics are citing.
… Saying that the protective effect of health-care insurance is hard to measure is very different than saying it is “too small to measure,” particularly when the comment is coming from someone who pays for health-care insurance, and is being made in context of whether the uninsured should get insurance rather than whether the insured should let go of theirs. There’s a methodological question here, and then there are political agendas here, and the two are getting mixed up
The hypothesis that some individuals in this debate are motivated by political agendas strikes me as far more plausible than the hypothesis that there is little proven connection between insurance and health and mortality. Though, that bar is pretty low.
Later: See also Kevin Drum who puts up a nice chart from McWilliams’ 2009 Milbank Quarterly paper that summarizes the research. Worth a look.
Letting Perfect be the Enemy of Good?
This is a guest post by J. Michael McWilliams, MD, PhD, assistant professor of health care policy and of medicine at Harvard Medical School and an associate physician in the Division of General Medicine at Brigham and Women’s Hospital. He is also author of the 2009 Milbank Quarterly paper “Health Consequences of Uninsurance among Adults in the United States: Recent Evidence and Implications.” (This post has been cited in the 18 February 2010 edition of Health Wonk Review.)
An Atlantic Monthly article by Megan McArdle questions whether health insurance coverage saves lives, drawing from a narrow slice of the literature to suggest the beneficial effects of insurance coverage on mortality might be negligible. While it is true these effects have been challenging for researchers to assess accurately, this question deserves more than a selective reading of the literature to inform the public and policymakers properly. Indeed, when reviewed comprehensively and with an understanding of key clinical and methodological nuances, the research to date provides consistent and compelling evidence that health insurance coverage significantly improves health outcomes, particularly for adults with treatable conditions (McWilliams 2009).
Studies on the health consequences of uninsurance can be broadly categorized as observational or quasi-experimental. Observational studies compare health outcomes between insured and uninsured adults and use statistical techniques to control for differences in other predictors of health between the two groups. These studies are fundamentally limited because it is usually impossible to control for all possible differences and some differences may be both causes and consequences of insurance coverage. Consequently, observational results may underestimate or overestimate the true effects of coverage. From the sizable observational literature, McArdle selects just one negative study to suggest insurance coverage may not affect mortality (Kronick 2009). Yet several other observational studies that controlled for an equally robust set of characteristics have consistently demonstrated a 35-43% greater risk of death within 8-10 years for adults who were uninsured at baseline and even higher relative risks for older uninsured adults with treatable chronic conditions such as diabetes and hypertension (Baker et al. 2006; McWilliams et al. 2004; Wilper et al. 2009).
Because these observational studies are not sufficiently rigorous to support causal conclusions, we should look to studies that are more experimental in design for more definitive evidence. McArdle cites a principal finding of the RAND Health Insurance Experiment (HIE) that more generous coverage led to more health-care utilization but not better health outcomes on average. However, the set of findings from the RAND HIE that is arguably more salient to this discussion is that more generous coverage did lead to better blood pressure control and lower predicted mortality for low-income adults with hypertension — adults that resemble the uninsured population more closely than the average adult. Moreover, the RAND study was conducted in the 1970s, prior to numerous advances that have improved the effectiveness of medical care for many acute and chronic conditions.
From the quasi-experimental literature, McArdle cites evidence of a lack of immediate survival gains with near-universal Medicare coverage after age 65 in the general population (Card et al. 2004; Levy, and Meltzer 2008). From a clinical perspective, however, we should not expect immediate survival gains for most previously uninsured adults because mortality is such a distal outcome. Survival gains may not manifest for years after improved chronic disease control and cancer screening are established, suggesting much more complex improvements in mortality trends are likely to evolve after age 65 in response to universal coverage. Quasi-experiments that rely on abrupt discontinuities occurring with age are not well suited to capturing these complex but potentially large effects. Consequently, the absence of evidence suggested by these studies is not evidence of absence. In contrast to the general population, immediate mortality effects might be expected for acutely ill patients for whom coverage may improve access to life-saving procedures and therapies. Indeed, a more recent study found age-eligibility for Medicare was associated with a substantial and lasting reduction in mortality for patients who were hospitalized for a range of acute illnesses that were amenable to treatment (Card et al. 2009).
Because many quasi-experimental strategies are geared to capture effects of insurance coverage only if they occur in the short term, they are better suited to examining proximal or intermediate health outcomes. Therefore, perhaps more can be learned about the effects of insurance coverage on mortality from studies that rigorously examine effects on health outcomes that are highly predictive of mortality. To date, numerous studies have found consistently beneficial and often significant effects of insurance coverage on health across a comprehensive set of outcomes and a broad range of treatable chronic and acute conditions that affect many adults in the U.S., including hypertension, coronary artery disease, congestive heart failure, stroke, diabetes, HIV infection, depressive symptoms, acute myocardial infarction, acute respiratory illnesses, and traumatic injuries (McWilliams 2009). In particular, several studies have robustly demonstrated positive effects of near-universal Medicare coverage after age 65 on self-reported health outcomes and clinical measures of disease control, particular for adults with cardiovascular disease or diabetes who make up two-thirds of the near-elderly (Decker and Remler 2004; McWilliams et al. 2007, 2009). Thus, when rigorous study designs have been coupled with appropriate outcomes and applied to clinical populations for whom medical care is effective, the evidence that insurance coverage improves health and survival is consistent and convincing.
How many lives would universal coverage save each year? A rigorous body of research tells us the answer is many, probably thousands if not tens of thousands. Short of the perfect study, however, we will never know the exact number. In the meantime, we can let perfect be the enemy of good. Or we can recognize the evidence to date is sufficiently robust for policymakers to proceed confidently with health care reforms that promise substantial health and financial benefits for millions of uninsured Americans.
References
Baker, D. W., J. J. Sudano, R. Durazo-Arvizu, J. Feinglass, W. P. Witt, and J. Thompson. 2006. “Health insurance coverage and the risk of decline in overall health and death among the near elderly, 1992-2002.” Med Care 44:277-82.
Card, D., C. Dobkin, and N. Maestas. 2004. “The impact of nearly universal insurance coverage on health care utilization and health: evidence from Medicare”. NBER Working Paper Series. Cambridge, MA: National Bureau of Economic Research.
Card, D., C. Dobkin, and N. Maestas. 2009. “Does Medicare save lives?” Quart J Econ 124(2):531-96.
Decker, S. L. and D. K. Remler. 2004. “How much might universal health insurance reduce socioeconomic disparities in health? : A comparison of the US and Canada.” Appl Health Econ Health Policy 3:205-16.
Kronick, R. 2009. “Health insurance coverage and mortality revisited.” Health Serv Res 44:1211-31.
Levy, H. and D. Meltzer. 2008. “The impact of health insurance on health.” Annu Rev Public Health 29:399-409.
McWilliams, J. M. 2009. “Health consequences of uninsurance among adults in the United States: recent evidence and implications.” Milbank Q 87:443-94.
McWilliams, J. M., E. Meara, A. M. Zaslavsky, and J. Z. Ayanian. 2007. “Health of previously uninsured adults after acquiring Medicare coverage.” JAMA 298:2886-94.
McWilliams, J. M., E. Meara, A. M. Zaslavsky, and J. Z. Ayanian. 2009. “Differences in control of cardiovascular disease and diabetes by race, ethnicity, and education: U.S. trends from 1999 to 2006 and effects of Medicare coverage.” Ann Intern Med 150:505-15.
McWilliams, J. M., A. M. Zaslavsky, E. Meara, and J. Z. Ayanian. 2004. “Health insurance coverage and mortality among the near-elderly.” Health Aff (Millwood) 23:223-33.
Wilper, A. P., S. Woolhandler, K. E. Lasser, D. McCormick, D. H. Bor, and D. U. Himmelstein. 2009. “Health insurance and mortality in US adults.” Am J Public Health 99(12):2289-95.
Catch Up
Some notes to bring you up to date on the latest blogosphere dust up I’m involved in:
Megan McArdle has a new post relating to her Atlantic Monthly piece on health insurance and mortality and reaction to it. I’m not going to pick it apart. Also, she and I have been conversing in the comments to my prior post. It is clear that our differences won’t be resolved through dialog, nor did I expect they would be. But the debate is thought provoking and potentially educational, if one keeps clear what is sound evidence and what is opinion, which is no easy task.
At the heart of it all are these questions:
- Do we need to know to some level of certainty to what degree health insurance prevents death before deciding how to extend coverage to more Americans or whether we should do so at all?
- What is that level of certainty?
- How sure are we that level of certainty is achievable?
- Or, in order to act is it sufficient to know that lack of insurance increases mortality even if we can’t be as precise as we might like about numbers of lives?
- Is mortality the only relevant measure?
- If we judge the cost per saved life too high does that mean we should reduce overall taxpayer liability by covering fewer people (saving fewer lives) or by redoubling efforts to increase cost-efficiency of care delivered?
Reasonable people can disagree about the answers to these questions.
Subscribe
