The following originally appeared on The Upshot (copyright 2017, The New York Times Company). It was coauthored by Aaron Carroll and Austin Frakt.
As a program for low-income Americans, Medicaid requires the poor to pay almost nothing for their health care. Republicans in Congress have made clear that they want to change that equation for many, whether through the health bill that is struggling in the Senate or through future legislation.
The current proposal, to scale back the Affordable Care Act’s Medicaid expansion and to cap spending each year, would give incentives to states to drop Medicaid coverage for millions of low-income Americans. It would offer tax credits toward premiums for private coverage, but those policies would come with thousands of dollars in new deductibles and other cost sharing. Despite the much higher out-of-pocket costs, some policy analysts and policy makers argue that low-income Americans would be better off.
To take one highly placed example, Seema Verma, the leader of the agency that administers Medicaid, recently cited studies questioning the program’s effectiveness and wrote that the health bill “will help Medicaid produce better results for recipients.”
What is the basis for the argument that poor Americans will be healthier if they are required to pay substantially more for health care? It appears that proponents like Ms. Verma have looked at research and concluded that having Medicaid is often no better than being uninsured — and thus that any private insurance, even with enormous deductibles, must be better. But our examination of research in this field suggests this kind of thinking is based on a classic misunderstanding: confusing correlation for causation.
It’s relatively easy to conduct and publish research that shows that Medicaid enrollees have worse health care outcomes than those with private coverage or even with no coverage. One such study that received considerable attention was conducted at the University of Virginia Health System.
For patients with different kinds of insurance — Medicaid, Medicare, private insurance and none — researchers examined the outcomes from almost 900,000 major operations, like coronary artery bypass grafts or organ removal. They found that Medicaid patients were more likely than any other type of patient to die in the hospital. They were also more likely to have certain kinds of complications and infections. Medicaid patients stayed in the hospital longer and cost more than any other type of patient. Private insurance outperformed Medicaid by almost every measure.
Other studies have also found that Medicaid patients have worse health outcomes than those with private coverage or even those with no insurance. If we take them to mean that Medicaid causes worse health, we would be justified in canceling the program. Why spend more to get less?
But that is not a proper interpretation of such studies. There are many other, more plausible explanations for the findings. Medicaid enrollees are of lower socioeconomic status — even lower than the uninsured as a group — and so may have fewer community and family resources that promote good health. They also tend to be sicker than other patients. In fact, some health care providers help the sickest and the neediest to enroll in Medicaid when they have no other option for coverage. Because people can sign up for Medicaid retroactively, becoming ill often leads to Medicaid enrollment, not the opposite.
Some of these differences can be measured and are controlled for in statistical analyses, including the Virginia study. But many other unmeasured differences can skew results, even in studies with such statistical controls. The authors of the U.V.A. surgical study and of studies like it know this, and say as much right in their papers. They practically shout that the correlations they find are not evidence of causation.
That hasn’t stopped policy makers and others in the media from asserting otherwise.
Other approaches to studying Medicaid more credibly demonstrate the value of the program. The most straightforward way is a prospective randomized trial, which gets around the subtle biases that plague studies that use only statistical controls. There has been exactly one randomized study of Medicaid, focused on an expansion of the program in Oregon.
Because demand for the program exceeded what Oregon could fund, in 2008 the state introduced a lottery for Medicaid eligibility. A now famous analysis took advantage of this lottery’s randomness, finding that Medicaid increased rates of diabetes detection and management, reduced rates of depression and lowered financial strain. It did not detect improvements in mortality or measures of physical health, but it did not have enough sick patients or enough time to detect differences we might have expected to see. In other words, it was not powered to detect changes in mortality or physical health.
Saying that this study proves Medicaid doesn’t work ignores this limitation. Regardless, there was nothing to indicate that having Medicaid worsened health.
Another way to tease out the causal effect of Medicaid is to look at variations in Medicaid eligibility rules across states. With respect to health outcomes, these state decisions are effectively random, so they act like a natural experiment. Older studies based on this approach, using data from the 1980s and 1990s, have not found that Medicaid causes worse health.
Findings from more recent studies looking at expansions in enrollment, in the 2000s and then under the Affordable Care Act in 2014, are consistent with older ones. One can argue that Medicaid can be improved upon, but the credible evidence to date is that Medicaid improves health. It is better than being uninsured.
Here’s another telling way to test the idea that Medicaid is harmful. Some of the studies that associate Medicaid with worse health, as compared with private insurance, also find the same association with Medicare. No one argues that Medicare is making people sick.
A very recent New England Journal of Medicine review by Ben Sommers, Atul Gawande and Kate Baicker found that Medicaid increases patients’ access to care and leads to earlier detection of disease, better medication adherence and improved management of chronic conditions. It also provides people with peace of mind — knowing that they will be able to afford care when they get sick.
Research is clear on how people react when asked to pay more for their health care, as the Senate would ask many of those now on Medicaid to do. As the Congressional Budget Office reported, many poor people would choose not to be covered, because even if they could afford the premiums with help from tax credits, deductibles and co-payments would still be prohibitively expensive. No studies prove that removing millions from Medicaid in this way would “produce better results for recipients,” at least as far as their health is concerned.