The 2012 election campaign is in full swing, and, for better or worse, health care is one of the major defining issues of the election. How can it not be, given the passage of the Patient Protection and Affordable Care Act (PPACA), also colloquially known as “Obamacare,” was one of the Obama administration’s major accomplishments and arguably the largest remaking of the American health care system since Medicare in 1965? It’s also been singularly unpopular thus far, contributing to the Republican takeover of the House of Representatives in the 2010 elections, as well as the erosion of Democratic control of the Senate. Given that this is a medical blog dedicated to discussing the scientific basis of medicine and not a political or health policy blog, I am not going to go into the reasons for a lot of this. What I am going discuss is a recent eruption of the central problem that led President Obama to make the PPACA one of the central policy initiatives, if not the central policy initiative, of his first term. That problem is the issue of people without health insurance, who number roughly 50 million, with a further estimate that 86.7 million people were uninsured at some point during the two year period from 2007 to 2008, representing about 29% of the total U.S. population under 65.
The question that bubbled to the surface last week in the form of a statement by Republican challenger Mitt Romney, and a tear-inducing op-ed piece published yesterday in the New York Times by Nicholas Kristof entitled A Possibly Fatal Mistake, is what the health impact of not having insurance is for those millions of people. This is a question that can be addressed scientifically and is, despite its politically charged nature, correctly within the purview of science-based medicine. What to do about it, in contrast, is a matter for politics and public policy. So first let’s examine the question.
A clueless statement and an op-ed that puts a face on the problem
Before we discuss the evidence regarding the health effects of being uninsured, let’s look Romney’s statement and why it resulted in so much blowback. Romney made his assertion during an interview with the editors of The Columbus Dispatch:
“We don’t have a setting across this country where if you don’t have insurance, we just say to you, ‘Tough luck, you’re going to die when you have your heart attack,’” he said as he offered more hints as to what he would put in place of “Obamacare,” which he has pledged to repeal.
“No, you go to the hospital, you get treated, you get care, and it’s paid for, either by charity, the government or by the hospital. We don’t have people that become ill, who die in their apartment because they don’t have insurance.”
He pointed out that federal law requires hospitals to treat those without health insurance — although hospital officials frequently say that drives up health-care costs.
Romney was referring to the Emergency Medical Treatment and Active Labor Act (EMTALA), a federal law passed in 1986 under the Reagan administration that requires hospitals to provide care to anyone needing emergency treatment regardless of citizenship, legal status, or ability to pay. Hospitals may only transfer or discharge patients requiring emergency care after stabilization, when their condition requires transfer to a tertiary care hospital, or against medical advice. It is highly unlikely that any person who has ever worked in an emergency room or cared for the uninsured would make such a statement. Emergency rooms are not equipped to treat complex conditions; all they can do is to treat the acute problem. In addition, tertiary care hospitals receive a lot of patients admitted under EMTALA, who are transferred at the drop of a hat. Well do I remember my days as a surgery resident rotating in the county hospital, when we used to joke about the Friday afternoon phone calls to transfer patients who had failed a wallet biopsy. We even knew what time was the “witching hour,” when we were most likely to get such calls. Of course, the problem with EMTALA was (and is) that there were no provisions for reimbursement for uncompensated care. Basically, hospitals were forced by law to eat the costs of caring for the uninsured.
Since completing fellowship, I have held faculty positions in two of the 41 NCI-designated comprehensive cancer centers in the U.S., both of which take care of a lot of uninsured patients. In New Jersey, I used to take care of quite a few illegal immigrants. Here in Detroit, it’s the uninsured and the poor, so much so that a large fraction of my practice is made up of the uninsured and Medicaid patients. I’ve seen more women than I can remember who waited far longer than they should have to see a doctor for their breast cancer because they couldn’t afford it. Over the years, all too often my patients have been symptomatic for quite some time, and when they finally do present their tumors are larger, more difficult to treat, and more likely to kill them. They represent the female equivalent of Kristof’s uninsured friend Scott, who is the human face of the issue discussed in his NYT op-ed and tells his story:
In 2011 I began having greater difficulty peeing. I didn’t go see the doctor because that would have been several hundred dollars out of pocket — just enough disincentive to get me to make a bad decision.
Early this year, I began seeing blood in my urine, and then I got scared. I Googled “blood in urine” and turned up several possible explanations. I remember sitting at my computer and thinking, “Well, I can afford the cost of an infection, but cancer would probably bust my bank and take everything in my I.R.A. So I’m just going to bet on this being an infection.”
I was extremely busy at work since it was peak tax season, so I figured I’d go after April 15. Then I developed a 102-degree fever and went to one of those urgent care clinics in a strip mall. (I didn’t have a regular physician and hadn’t been getting annual physicals.)
The doctor there gave me a diagnosis of prostate infection and prescribed antibiotics. That seemed to help, but by April 15 it seemed to be getting worse again. On May 3 I saw a urologist, and he drew blood for tests, but the results weren’t back yet that weekend when my health degenerated rapidly.
A friend took me to the Swedish Medical Center Emergency Room near my home. Doctors ran blood labs immediately. A normal P.S.A. test for prostate cancer is below 4, and mine was 1,100. They also did a CT scan, which turned up possible signs of cancerous bone lesions. Prostate cancer likes to spread to bones.
I also had a blood disorder called disseminated intravascular coagulation, which is sometimes brought on by prostate cancer. It basically causes you to destroy your own blood cells, and it’s abbreviated as D.I.C. Medical students joke that it stands for “death is close.”
I realize that right now I’m referring to my anecdotal experience. However, one anecdote is that of a man who gambled and lost because health insurance was too expensive. The rest is my experience in a highly specialized field in a city with high unemployment and poverty. It is quite possible that such experience can be misleading, and certainly one of the key messages we promote on this blog is that anecdotal experience is inherently potentially misleading. (That’s why it’s the primary evidence used by promoters of unscientific or pseudoscientific medicine.) In a way, Kristof’s friend’s story would seem to confirm Romney’s statement, at least on the surface. Scott did, after all, end up getting excellent medical care for his stage IV prostate cancer, and, although he probably could have afforded health insurance if he had stretched a bit, did make a choice not to purchase insurance. But, then, as I said, anecdotes can be misleading.
Before we get into the data itself, it is not much of a stretch to imagine that not having health insurance would result in worse health outcomes. What I am trying to say using SBM-speak is that the hypothesis that people without health insurance will be more likely to have health problems and die unnecessarily than people who have decent health insurance is a hypothesis with a fairly high degree of prior plausibility. After all, if you’re uninsured, you’re less likely to see a physician except when you get sick, less likely to be able to pay for your medications (particularly if they are expensive), and less likely to undergo routine preventative care. It’s thus plausible that being uninsured would be associated with an increased risk of death or poor health outcomes. None of this means we don’t have to do the research and look at the evidence; all it does is to suggest hypotheses to test and emphasize that these hypotheses have a reasonable chance of being consistent with the data.
Even twenty years ago, this question was of intense interest. One of the seminal studies examining the relationship between health insurance and health outcomes was published in JAMA by Franks et al., who prospectively followed 4,694 adults older than 25 years who reported they were uninsured or privately insured in the first National Health and Nutrition Examination Survey (NHANES I), a representative cohort of the US population from initial interview in 1971 through 1975 until 1987. They found a 25% higher risk of mortality in the uninsured after adjusting for age, smoking, alcohol consumption, obesity, education and income. This effect was evident in all sociodemographic health insurance and mortality groups examined.
In 2002, the Institute of Medicine estimated that over 18,000 Americans between the ages of 25-64 die annually because of lack of health insurance, a number comparable to the number who died of diabetes, stroke, or homicide in 2001. Among the conclusions of this report:
- Uninsured adults are less likely than adults with any kind of health coverage to receive preventive and screening services and to receive them on a timely basis. Health insurance that provides coverage of preventive and screening services is likely to result in greater and more appropriate use of these services.
- Uninsured cancer patients generally are in poorer health and are more likely to die prematurely than persons with insurance, largely because of delayed diagnosis. This finding is supported by population-based studies of persons with breast, cervical, colorectal, and prostate cancer and melanoma.
- Uninsured adults with diabetes are less likely to receive recommended services. Lacking health insurance for longer periods increases the risk of inadequate care for this condition and can lead to uncontrolled blood sugar levels, which, over time, put diabetics at risk for additional chronic disease and disability.
- Uninsured adults with hypertension or high blood cholesterol have diminished access to care, are less likely to be screened, are less likely to take prescription medication if diagnosed, and experience worse health outcomes.
- Uninsured patients with end-stage renal disease begin dialysis with more severe disease than do those who had insurance before beginning dialysis.
- Uninsured adults with HIV infection are less likely to receive highly effective medications that have been shown to improve survival and die sooner than those with coverage.
- Uninsured patients who are hospitalized for a range of conditions are more likely to die in the hospital, to receive fewer services when admitted, and to experience substandard care and resultant injury than are insured patients.
- Uninsured persons with trauma are less likely to be admitted to the hospital, more likely to receive fewer services when admitted, and are more likely to die than are insured trauma victims.
- Uninsured patients with acute cardiovascular disease are less likely to be admitted to a hospital that performs angiography or revascularization procedures, are less likely to receive these diagnostic and treatment procedures, and are more likely to die in the short term.
In 2008, the Urban Institute updated the IOM numbers by applying the methodology used by the IOM to more recent Census Bureau estimates of the uninsured, and estimated that in 2006 there were 22,000 excess deaths that could be attributed to lack of health insurance. The Urban Institute also suggested that the IOM analysis might have underestimated the number of deaths resulting from being uninsured. Its rationale was as follows:
The underlying longitudinal studies on which IOM relied did not specify the impact of insurance coverage on mortality by 10-year age groups. Rather, they documented the relationship between insurance and mortality across the sum total of all surveyed age groups. The IOM’s methodology implicitly assumed that insurance reduces mortality by the identical percentage for each 10-year age band, which the underlying research did not show. More grounded in the research would be an application of differential mortality estimates to all adults age 25–64, as was done for those longitudinal studies, rather than separately to each age group within this range. For 2000–06, this alternative approach raises the estimated number of excess deaths by an average of 20.5 percent a year.
In 2009, in a study from Harvard Medical School and the Cambridge Health Alliance, Wilper et al. published updated estimate of excess mortality associated with lack of insurance in the American Journal of Public Health. This analysis used methodology similar to that of Franks et al. applied to the third National Health and Nutrition Examination Survey (NHANES III), specifically 9,004 patients between ages 17 and 64 with complete baseline data for interview and physical examination. They found that the hazard ratio for death for the uninsured was 1.40 (confidence interval 1.06 to 1.84) compared to those with private health insurance. This particular study is the source of a rather famous number: 45,000 patients die due to lack of insurance each year. This particular study is at the high end of the estimates of excess deaths associated with lack of health insurance, which is why it not surprisingly often comes in for the most criticism, particularly given that it was supported by a partisan group, Physicians for a National Health Program. That’s why I tend to view this study as an outlier, but even outliers can sometimes tell us something. Whether the Harvard study was an outlier or not, that same year, the IOM updated its 2002 report. One of its conclusions was:
In contrast, the body of evidence on the effects of uninsurance on adults’ health has strengthened considerably since 2002. Numerous studies have addressed some of the methodological shortcomings of past research. As discussed further below, 17 observational and 13 quasi-experimental rigorous analyses have reported significant findings related to health insurance and adults’ health (Table 3-3) (McWilliams, 2008). The quality and consistency of the recent research findings is striking. As would be expected, health insurance is clearly most beneficial for adults who need medical attention, particularly for adults with common chronic conditions or acute conditions for which effective treatments are available. Furthermore, national studies assessing the effects of near-universal Medicare coverage after age 65 suggest that uninsured near-elderly adults who are acutely or chronically ill substantially benefit from gaining health insurance coverage.
There are 13 recent studies on the health effects of health insurance coverage for children, including 5 studies that used quasi-experimental methods (Aizer, 2007; Bermudez and Baker, 2005; Cousineau et al., 2008; Currie et al., 2008; Howell et al., 2008a). These studies suggest that health insurance is beneficial for children in several ways, resulting in more timely diagnosis of serious health conditions, fewer avoidable hospitalizations, better asthma outcomes, and fewer missed school days (Aizer, 2007; Bermudez and Baker, 2005; Cousineau et al., 2008; Currie et al., 2008; Damiano et al., 2003; Fox et al., 2003; Froehlich et al., 2007; Howell and Trenholm, 2007; Howell et al., 2008a,b; Maniatis et al., 2005; Szilagyi et al., 2004, 2006).
But that’s not all. Since it’s my specialty, I’ll start by looking at some of the evidence regarding the outcomes of breast cancer patients who are uninsured compared to those who have health insurance. For example, a study published this year examining the outcomes of 2,157 hospital admissions for women with spinal metastases from breast cancer. Analyses were adjusted for differences in patient age, gender, primary tumor histology, socioeconomic status, hospital bed size, and hospital teaching status. The investigators found that women operated on for spinal metastases from breast cancer tended to do worse and have a higher risk of death if they were uninsured than if they had private insurance. The authors concluded that the poorer outcomes observed among the uninsured were primarily due to the uninsured patients being significantly more likely to have a nonelective hospital admission and present with myelopathy. Although this study had some limitations, namely that it couldn’t account for lesser quality private insurance (for instance, plans with high copays and/or poorer coverage) and variations in Medicaid eligibility by state. Also, the database used only includes in-hospital data and therefore couldn’t examine long-term outcomes.
Since surgery is also my specialty, I thought I’d also point out that there is considerable evidence that being uninsured or underinsured is associated with worse outcomes after surgery. For example, a recent study published in the Annals of Surgery from LePar et al. at the University of Virginia examined outcomes from 893,658 major surgical operations and found that mortality was considerably worse in Medicare, Medicaid, and the uninsured than they were in patients with private insurance. Adjusting for age, gender, income, geographic region, operation, and 30 comorbid conditions eliminated the outcome disparity for Medicare patients, but Medicaid and uninsured payer status still independently conferred the highest adjusted risks of mortality.
In fact, the list of conditions and procedures for which being uninsured is associated with poorer outcomes and higher mortality goes on and on: cardiac valve surgery, surgery for colorectal cancer, breast cancer treatment and outcomes, trauma mortality (including among children), and abdominal aortic aneurysms, to name a few. Moreover, analysis of survey data from patients who were uninsured but then became old enough to be enrolled in Medicare suggests that “acquisition of Medicare coverage was associated with improved trends in self-reported health for previously uninsured adults, particularly those with cardiovascular disease or diabetes.” In summary, there is a large and robust body of evidence suggesting that people do, in fact, die because of lack of health insurance.
Not so fast…
Of course, for a question as complex and prone to confounders as whether lack of health insurance is associated with poorer outcomes, including mortality, there are always those who are not convinced by existing observational data. Certainly, lack of health insurance can be a marker, not a cause, for poor health and subsequent poor outcomes, and teasing out the various confounding factors is not a trivial task. Perhaps the most widely cited study questioning this relationship was featured prominently in an oft-cited article in the lay press by Megan McArdle in The Atlantic entitled Myth diagnosis. It’s a study by Richard Kronick published in Health Services Research in 2009 that questions the IOM report from 2002:
These results demonstrate that if two people are otherwise similar at baseline on characteristics controlled for in the model presented in Table 3, but one is insured and the other uninsured, their likelihood of survival over a 2–16-year follow-up period is nearly identical. Further, I show that survival probabilities for the insured and uninsured are similar even among disadvantaged subsets of the population; that there are no differences for long-term uninsured compared with short-term uninsured; that the results are no different when the length of the follow-up period is shortened; and that there are no differences when causes of death are restricted to those causes thought to be amenable to the quality of health care.
Basically, Kronick found no relationships between insurance status and mortality. While this study was large (600,000 subjects) and controlled, it is also an outlier, just as much as the Harvard study is an outlier. Again, that doesn’t mean it was a bad study; outliers can often tell us something, and what Kronick seems to be telling us is that the magnitude of the effect on mortality associated with lack of insurance might not be as large as previously thought. Might. It is, remember, just one study, as large as it might be. McArdle might refer to Kronick’s study as “what may be the largest and most comprehensive analysis yet done of the effect of insurance on mortality,” which sounds incredibly impressive, but the simple fact is that no single study can provide the answer, particularly to question as complex as whether having no health insurance status is associated with increased mortality and poor outcomes. Kronick’s study also has a significant problem that was pointed out in this post by Ezra Klein, namely that people in poor health are more likely to seek health insurance, which would tend to obscure any positive relationship between health insurance and health status.
McArdle also makes another argument against such a relationship:
This result is not, perhaps, as shocking as it seems. Health care heals, but it also kills. Someone who lacked insurance over the past few decades might have missed taking their Lipitor, but also their Vioxx or Fen-Phen. According to one estimate, 80,000 people a year are killed just by “nosocomial infections”—infections that arise as a result of medical treatment. The only truly experimental study on health insurance, a randomized study of almost 4,000 subjects done by Rand and concluded in 1982, found that increasing the generosity of people’s health insurance caused them to use more health care, but made almost no difference in their health status.
I hate to say it, but McArdle is drifting rather close to Gary Null territory here, in which she seems to be arguing that whatever benefit having decent health insurance might convey, it’s about the same as the harm that “conventional” medicine does. In other words, her argument seems to be that providing people more access to health care will cause as much harm as benefit, making it a wash whether one is insured or not. Of course, that argument cuts both ways, if you accept estimates of over 100,000 “deaths by medicine” per year in that it would imply that having health insurance confers a benefit in terms of mortality reduction that is much larger than the numbers we have would suggest, making the imperative to improve health care coverage and decrease medical errors a much more reasonable conclusion from such an argument than concluding that striving for universal coverage would not reduce mortality. Be that as it may, more problematic is that like many proponents of dubious medicine and science, McArdle cherry picked the literature, choosing one study that is an outlier and a thirty year old study from the RAND Corporation that showed what she wanted and in essence dismissed the rest. In refuting McArdle, by J. Michael McWilliams, MD, PhD, Assistant Professor of Health Care Policy and Medicine at Harvard Medical School and an associate physician in the Division of General Medicine at Brigham and Women’s Hospital points this out and speculates:
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.
The very term “science-based medicine” was chosen intentionally. Medicine itself is not a science. It can’t be. There are too many other factors that influence treatments, including patient preference, resource allocation, and level of skill of individual practitioners, to name just a few. Our central thesis is that medicine should be based on science and that the best health care is based on science. My purpose in writing this post was not to advocate for any specific solution to the problem of the uninsured, although people who know me know my politics and my position on the matter. Rather, it is to lay out the science studying the question of the relationship between health insurance status and health outcomes. While we do frequently say that correlation does not necessarily equal causation, in some cases the correlation is so tight that it strongly suggests causation. This is one such case. Given that there is no ethical way ever to do a randomized clinical trial in which people are randomly assigned to be insured or uninsured, much as is the case for examining health outcomes between vaccinated and unvaccinated children, we are forced to rely on observational and quasi-experimental data. Those data support the hypothesis that providing health insurance to as many people as possible is associated with better health outcomes and that lack of insurance is associated with poorer health outcomes. That is the science. When someone like Mitt Romney claims that no one ever dies from lack of insurance in the U.S., he is demonstrably wrong.
What we as a society decide do with the results of the science examining this question is less a matter of science than it is of politics and policy.