Wednesday’s New England Journal of Medicine article on the Oregon Medicaid lottery has provoked a firestorm of reaction. I’m not sure what I can add after reading Austin and Aaron’s great posts. They’re just killing this one.
I’m now pondering the Supplementary Appendix prepared by the study team. It’s a pleasure to view the craftsmanship and clarity on display there. Many people are reading this study with an eye towards the big-ticket questions of the Affordable Care Act. I believe that’s a mistake. The moral and pragmatic case for expanded health coverage was overwhelming before these Oregon findings were released. That case is no less overwhelming now.
As Aaron and Austin recount, Oregon Medicaid lottery winners were markedly less likely to face catastrophic medical bills or endure other forms of financial distress. Lottery winners were more likely to report that they received high-quality medical care through a usual provider who met their treatment needs. Medicaid recipients in the experiment were nine percentage-points less likely to satisfy screening criteria for depression, a 30% decline in the base rate. Lottery winners over age fifty were markedly more likely to receive mammograms, Pap smears, and other preventive services. People were more likely to receive depression or diabetes medications.
So Medicaid coverage brought substantial benefits. Indeed the program compares surprisingly well with Medicare and with private insurance in providing effective coverage for low-income adults. Were these benefits worth the additional $1171.63 per-person annual Medicaid costs? I believe the question answers itself.
Yet the case for Medicaid expansion is not the only or the most interesting question in-play here. Distinguished Harvard economist Joseph Newhouse emailed that the Oregon study “is a Rorschach test of people’s views of the ACA.”
That’s perhaps unavoidable. It’s unfortunate, too. One sad consequence of polarization is our reflex to interpret every finding in light of its polemical utility in the broader political fight.
I’m not one to throw stones over partisan attachments. Yet the partisan lens often obscures what’s most instructive about excellent, complex, and in-some-ways limited studies. The view becomes especially murky because the Oregon study is so under-powered to examine the physical health outcomes of greatest concern.
The most obvious power problems arise because most Medicaid lottery winners never actually enrolled. As the Oregon researchers described in the Quarterly Journal of Economics, lottery winners were mailed an application and given 45 days to submit the appropriate paperwork establishing Medicaid eligibility:
About 30% of selected individuals successfully enrolled. There were two main sources of slippage: only about 60% of those selected sent back applications, and about half of those who sent back applications were deemed ineligible, primarily due to failure to meet the requirements of income in the last quarter corresponding to annual income below the poverty level, which in 2008 was $10,400 for a single person….
Lottery winners and controls sometimes obtained other coverage. Life intruded in other ways. All-in-all, winning the lottery was associated with a 25 percentage-point increase in actual Medicaid receipt.
Second, the study team lacked baseline clinical measures. So they could not specifically control for baseline health. They could segment the analysis into patient subgroups based on age or other factors. When researchers tried to drill down to these subgroups, they didn’t see much, but the confidence intervals were also very wide.
Consider the issue of high blood pressure. Baicker et al. reported no statistically significant reduction in hypertension. That’s disappointing, especially in light of the 1974-82 RAND Health Insurance Experiment, which found important predicted mortality reductions among low-income participants who began the study in ill-health. But the Oregon experiment was only a two-year study. It lacked the sample size and baseline data to properly replicate this analysis.
To give you a sense of the challenge, Oregon Medicaid receipt was associated with an estimated increase of 1.76 percentage points in new diagnoses for hypertension. Since the control group mean was 5.6 percent, Medicaid receipt was associated with a 30% increase in the detection of previously undiagnosed hypertension within a relatively short period.
That’s not statistically significant, but it’s potentially quite important. Unfortunately, it’s inherently difficult to tease out such a small effect. For all the thousands of study participants, back of the envelope calculation indicates this corresponds to something like 33 more detected cases within the treatment than the control group.
Ironically, the overall population effect on hypertension seemed quite similar in the Oregon experiment and the RAND HIE. Free care within the original RAND HIE induced a (non-significant) reduction of 0.7 mm Hg in diastolic BP. In Oregon, Medicaid receipt was associated with an (also non-significant) average reduction of 0.81.
I believe that a longer intervention with a larger sample will yield real, though modest health impacts on hypertension, diabetes, and other conditions. The abstract to the New England Journal study states: “Medicaid coverage generated no significant improvements in measured physical health outcomes in the first 2 years.” That’s correct, but easily misleading given the power limitations and some promising preliminary results.
I believe both ideological camps in American health policy should probably gain some comfort, and some discomfort, from these results. Medicaid absolutely helps people. There’s an OJ-searching-for-the-real-killer quality to arguments against health reform that claim otherwise.
Yet the critics are right that simple Medicaid enrollment failed to powerfully move the needle on basic cardiovascular measures as much as many of us hoped it would. Public insurance can and must do better in these areas.
Medicaid could certainly do a better job promoting individual health. Some issues are pretty basic. The original RAND HIE found that patients were particularly price-sensitive in their demand for dental and vision services.* Low-income participants in the RAND Health Insurance Experiment assigned to free care could see better, and had better oral health than their peers assigned to cost-sharing plans.
These lessons should be more consistently applied. Like many other states, Oregon doesn’t cover non-emergency dental services or vision care in its Medicaid program. Covering these services would bring immediate health benefits. Expanding such coverage would also engage people who have other difficulties such as undetected cardiovascular difficulties.
As in the original RAND Health Insurance Experiment, there’s no evidence that Oregon’s Medicaid enrollment improved health behaviors. Indeed Medicaid receipt was actually associated with increased smoking, presumably because lottery winners had increased disposable income.
Expanded Medicaid coverage is essential. Yet it’s only the first step to improved quality of care and improved population health. Improved disease management, accountable care organizations, and primary care medical homes are important, too. So are measures such as work-based wellness programs, smoking cessation programs, alcohol, and tobacco taxes to address public health concerns.
The Oregon research team should be proud of what they’ve accomplished. Their findings underscore the immediate benefits of health reform, but also how much work we still must do.
*The Affordable Care Act fills many holes in the areas of mental health and substance abuse services, too. That’s a topic for another occasion.