A major new study of the effects of Medicaid published in the New England Journal of Medicine yesterday found that the provision of "Medicaid coverage generated no significant improvements in measured physical health outcomes in the first 2 years." That is not exactly great news for Obamacare, which relies on Medicaid for roughly half of its health coverage expansion. In response, some of the health law's backers are arguing that, well, we can't be sure the study proves that Medicaid has no health benefits in part because the sample size is small enough to mean that the results are statistically underpowered. But that's not how the study's initial results, which appeared far more friendly to Medicaid, were reported and interpreted. Many of the individuals who wrote about the study's initial round of results, released in July of 2011, were quick to tout the study's robust design, and the certainty of its conclusions.
Writing at Kaiser Health News, for example, The New Republic's Jonathan Cohn declared that the study's design "makes it unusually significant." A blog post published by the left-leaning Century Foundation announced that the study's "findings were irrefutable." Aaron Carroll, an influential health policy blogger at The Incidental Economist, emphasized the rigor of the study. "I'd like to reiterate that this was a randomized controlled trial," he wrote. "An RCT is pretty much the best way to prove causality, especially if it's well done." And because it's an RCT, he concluded, "we can even start talking causality." Ezra Klein published a column touting the study with the headline, "Amazing Fact! Science Proves Health Insurance Works." He explained why the randomized study was so valuable: "The gold standard in research is a study that randomly chooses who gets a new treatment and who doesn't. That way, you know your results are unaffected by differences in the two populations you are studying."
Now, well, it's all a little less clear. "The problem with the Oregon study," Klein wrote this morning," …is we don't really know what we're learning." Carroll, who was ready to start talking causality when the first study was published, is now counseling caution. "So chill, people. This is another piece of evidence. It shows that some things improved for people who got Medicaid. For others, changes weren't statistically significant, which isn't the same thing as certainty of no effect. For still others, the jury is still out."
It's notable that the findings from the first round of study results were actually less robust than this week's results. Not only did the first round only measure a single year, there were no objective physical measures of health at all. Instead, the researchers did find big improvements in self-reported health. People who got Medicaid merely said they felt a lot better. And about two-thirds of that self-reported improvement appeared before any medical treatment had been obtained. Yet that was enough for many Obamacare backers to declare that certain victory was at hand. Indeed, despite the lack of objective measures, it was even enough for many reports to declare that we now had irrefutable evidence that Medicaid definitely does improve health.
The White House blog, for example, headlined an item on the first study "Health Insurance Leads to Healthier Americans." An ABC News report on the study opened by saying that the study "proves that being insured through Medicaid benefits people physically. Health policy analyst Harold Pollack used the initial results to ask, "Can conservatives please stop claiming that health insurance doesn't improve health?" Incidental Economist health policy blogger Austin Frakt expressed his confidence that "the research team will find that Medicaid does lead to better health" while singling out the sturdiness of the study's methodology and selection design for praise. A New York Times analysis concluded that "expanding insurance does not save society money — as some advocates of preventive medicine have claimed — but it does appear to make people mentally and physically healthier." Harvard health policy professor John McDonough hailed the study, and dismissed those who counseled caution about the study's results. "Naysayers are already out in force charging that the study results fail to identify actual improvements in enrollees' health status," he wrote. "Those kinds of results are down the road."
We've now gone down that road. But we didn't find those kinds of results. Not with the rigor that the study's authors deemed necessary, anyway. Instead, on the objective health measures, we—or rather the researchers behind the study—found some improvement in objective health measures. But not enough to rise to the level of statistical significance. Not enough to know with high confidence that Medicaid was the cause. This is not nothing. It's even potentially interesting. But it is far from definitive proof, or even just a strong reason to suspect, that Medicaid actually makes a measurable difference in objective health outcomes.
And while some on the left are still claiming victory—in part because of the objective health measures and in part because the study showed decreased risk of health-related financial catastrophe, and decreased probability of screening positive for depression, the particulars of the study's results should at the very least complicate their arguments.
If the primary goal of a program like Medicaid is to protect individuals from financial shocks associated with medical expenses, then why not support a far, far cheaper subsidized catastrophic insurance program instead of low-deductible insurance through Medicaid? If what the poor really need is financial protection, rather than health services, then why not just give them cash?
The depression results are unusual as well, because the study found no concurrent rise in the use of medication for depression. Might some of the difference be attributable to the fact that Medicaid beneficiaries had won the health insurance lottery and, as we know, felt better because of it?
As for the too-small-to-be-statistically-significant improvements on objective measures—even if we could be confident that the improvements were attributable to the provision of Medicaid, would those improvements be worth the high price of the program, both in its current form and its planned expansion under Obamacare? Medicaid currently costs the federal government about $250 billion a year, a figure that's projected to rise past $570 billion over the next decade. That doesn't count the hundreds of billions that state governments also spend on the program. (Health costs, many of which are related to Medicaid, are the biggest cause of budget trouble for states, according to the Government Accountability Office). And for that, beneficiaries are getting health benefits that are, at best, highly uncertain.
For what it's worth, I am glad to see that liberal health wonks are now preaching caution. Given the study's results, they are right to do so. I wish, however, that they had done so from the outset, and I hope that they will adopt a less confident approach in the future. The second-round results of Oregon's experiment with Medicaid suggest that it's possible that Medicaid may have some improvement on a few health measures. But further study could just as easily show that they don't, and that the improvements found here are little more than statistical noise. Science, in other words, has not really proven that Medicaid works, or that it doesn't. But it has strongly suggested, in a gold-standard study, that on objective measures of physical health, those with coverage through the program may not be much better off than those without.