As adherents to evidence-based practice, we are used to paradigms changing. From leeches to surgery for low back pain, the medical literature is filled with things that seemed sensible and theoretically sound, but that on rigorous study turned out not to be correct. This is why providers need to keep up on the literature. But there are caveats. First, we must balance an openness to changing practice when the evidence supports or even demands doing so, with a healthy skepticism and critical evaluation of the evidence to be sure we draw the right conclusions from what are often imperfect studies. We can all think of examples of papers that at first blush appeared to be true landmarks, only to have substantial flaws revealed, or be contradicted by subsequent data. In addition, data are merely facts; to become information, data must be interpreted, and those interpretations can be subjective. Finally, most progress in health care is at best incremental. It is rare that any one study singlehandedly changes what we do.
A recent paper in the New England Journal of Medicine has been hailed by at least some commentators as one of those rare solo game changers. In my mind, though, I believe its data are being widely misinterpreted. I am referring to the study of the Oregon Medicaid Experiment. Briefly, in 2008, Oregon was expanding its Medicaid coverage for childless adults. However, there was less funding available than originally intended, so they allocated the coverage to the applicants using a lottery. This was the holy grail of health services research – a randomized controlled trial (albeit a naturally occurring one) of insurance vs. no insurance. Such rigorous study designs almost never occur in the area of health policy. This was a rare opportunity to answer the question of how insurance coverage affects utilization of services and, most importantly, health, without the confounding and other flaws that occur when, for example, comparing different states with different levels of coverage.
The authors found that when comparing those who were randomly selected to get coverage with those who remained uninsured, those with Medicaid used more health services. This is perhaps not terribly surprising. But after two years of follow up, while the newly insured had lower rates of depression and less financial stress, there were no differences in several measures of health status including prevalence of diabetes and hypertension, cholesterol levels, or hemoglobin A1c levels in diabetics. These results are consistent with one of the only other RCTs of insurance coverage, the RAND study of the 1970s. All subjects in that study had coverage, but with varying levels of cost sharing. Better coverage led to more utilization, but without any clear overall difference in health status.
Some commentators, particularly those opposed to the Medicaid expansion included in the Affordable Care Act, have touted these studies as proving that comprehensive health insurance in general, and Medicaid in particular, do not work. Many others have pointed out specific flaws with the study that might limit this conclusion. But I think there are two additional major errors of interpretation here that we might heed.
What if the proper conclusion is not that health insurance doesn’t improve health, but that heath care does not improve health? After all, in both studies there were more doctors visits, prescriptions, etc., but no better health status. That might be a leap, but we do know that not all medical interventions (tests and treatments) are beneficial. Moreover, medical care is but one determinant, and a minor one at that, of a person’s health. Finding that having insurance by itself does not decrease the rate of diabetes isn’t terribly unexpected. But one potential lesson to draw from the Oregon study – and, I believe, and important one – is that health insurance is being spent on the wrong things. If health coverage, and health care, are to have a positive impact – if they are to have value – what we do may need to be more focused on prevention, on promoting adherence on the part of both patients and providers to proven management strategies, and on care coordination.
The second thing to keep in mind is that health is not merely the absence of disease. The World Health Organization, among others (including the American Academy of Pediatrics) support a more holistic view of health: a positive state of physical, mental, and social well-being. Yes, the lucky people who received Oregon Medicaid had similar rates of several measures of physical health. But they had lower rates of depression and of economic stress. If we had some composite measure of the comprehensive meaning of health, insurance would undoubtedly have been shown to improve it.
Surely at least a few of those who pay for health care will look at this study and draw a similar conclusion. If they pursue evidence-based policy making, they will develop ways to move models of care and payment in that direction. Fee-for-service may become the bloodletting of the health payment world.