Introduction
Few topics ignite such political fervor as health insurance. What exact benefits does coverage confer? Should the United States’ healthcare system be universalized, equally available to individuals of all socioeconomic backgrounds? How do expansions of public health insurance impact the American populace at large? These questions spark controversy, death-gripped by the partisanship which pervades American politics. While the landscape of healthcare and health insurance in the United States continues to metamorphose, so to do the arguments on both ends of the aisle crescendo in their ardor. And this debate is nothing new. Even a decade and a half ago, when the Affordable Care Act was still in its nascent stages, the dialogue surrounding health insurance was politically charged and hot-button. The truth is, however, that despite the torrent of argumentation and evidence leveraged by either side, before 2008 there was little certainty about what kind of consequences something as monumental as the Affordable Care Act would effect. This informational gap was bridged, at least in part, by one key experimental effort, which would leave an indelible imprint on health policy discourse within the United States.
Randomized controlled trials, hailed as the ideal experimental technique in medical and scientific studies, are seldom possible in social policy research. In 2008, the Oregon Health Plan (OHP) Medicaid Program – which, preceding the Affordable Care Act, provided Medicaid coverage for non-disabled and non-pregnant Oregonians in poverty – afforded principal investigators Katherine Baiker and Amy Finkelstein this precise opportunity to oversee a randomized controlled trial. After originally closing itself off to new enrollees in 2004, the OHP grew in capacity such that by 2008 it was able to extend coverage to 10,000 more people selected through a lottery process which a total of 70,000 people ultimately signed up for. Baiker and Finkelstein seized this lottery as a means of studying the effect of expanding public health insurance on a number of factors, including health care use, health outcomes, financial strain, and the overall welfare of low-income adults. The quasi-experimental and randomly controlled nature of their study accorded Baiker and Finkelstein the advantage of being able to assume that their comparison group (those who signed up for the lottery but were not selected) only differed from their treatment group (those who were selected for OHP Medicaid coverage via the lottery) in one key respect: “because of the increased Medicaid coverage among those selected by the lottery.” This direct comparison and simple design is what lended their project, coined the Oregon Health Insurance Experiment (OHIE), such acclaim.
Over the course of this article, I will focus attention upon the momentous Oregon Health Insurance Experiment and the implications of this study on the costs and benefits of Medicaid expansion, most particularly in terms of health care usage and health outcomes. Of course, potential limitations on the significance and generalizability of these findings must also be considered, which will become especially pertinent as analysis moves from 2008 Oregon to the questions which still linger about the costs and benefits of further expansion of Medicaid coverage under the Affordable Care Act. It will become clear that regulations and policy shifts concerning Medicaid coverage must incorporate concerns raised by the Oregon Health Insurance Experiment into their framework if they seek to optimize the efficiency and efficacy of public health insurance in the United States.
The Findings of the Oregon Health Insurance Experiment
The impact of health insurance on healthcare outcomes and use remains a point of contention within the present corpus of academic health insurance literature. For example, one analysis of Massachusetts’ 2006 expansion of public health insurance (often regarded as a precursor to the Affordable Care Act, both centering an effort to universalize health insurance coverage) finds that this health policy reform reduced emergency department usage about 8% to 5%. Meanwhile, the infamous RAND Health Insurance Experiment from the 1970s, which randomly assigned people to insurance plans with varying degrees of cost-sharing, supplied evidence to the contrary: more comprehensive insurance coverage was associated with more frequent emergency department use. The Oregon Health Insurance Experiment offers a uniquely compelling perspective within this discourse, circumventing certain confounding variables which could complicate purely observational studies.
In the vein of health care usage, the OHIE presented bifurcated results. Initially, researchers estimated that the treatment group reported no increase in emergency department nor hospital admissions, despite their newly afforded Medicaid coverage. At the same time, however, OHP Medicaid recipients did report a greater number of non-emergent office visits for primary and preventive medical care services, including cholesterol screening (an increase of 14.57 percentage points). In 2013, a year after the bulk of the OHIE’s results had been published, researchers returned to this data which informed their previous conclusion on the connection between emergency department use and Medicaid coverage, this time consulting hospital administrative data beyond mere self-reporting. While their analysis of this new data only implicated a third of the original treatment group, and may have been contaminated by certain Medicaid enrollees who did not originally figure into the OHIE (i.e. disabled and pregnant people), they were nonetheless able to tease out a statistically significant 7.0 percentage point increase in emergency-department visits for those with Medicaid as opposed to the uninsured. With this new finding in tow, access to and use of healthcare across the board – including visits to primary care, preventive care, and the emergency department – can be seen to experience an uptick for those who were randomly selected via lottery to have Medicaid coverage. Even data surrounding people’s level of dental care cohered with this trend of the treatment group making greater use of healthcare services than the comparison group. Indeed, following later analysis of administrative records by the OHIE researchers, the percentage of those who reported having unmet dental care needs was determined to decrease by a staggering 13.5 points once they were conferred Medicaid coverage.
Of course, the question of healthcare access and utilization must be divorced from the question of health outcomes. Just because people made more substantive and frequent contact with medical care does not mean that their physical wellbeing was automatically bolstered. In fact, after evaluating the effect of Medicaid coverage on a number of physical health measures, such as blood pressure, cholesterol, and glycated hemoglobin levels (a measure of diabetes), the OHIE researchers published their findings that the treatment group only minimally differed with respect to their health status from the comparison group. For example, in one article, they explain that the prevalence of hypertension and higher cholesterol levels remained unchanged irrespective of OHP Medicaid coverage status, with the same holding true for participants’ average glycated hemoglobin levels. Interestingly, Medicaid was found to affect a 3.83 percentage point increase in the probability of participants receiving diabetes diagnoses – a datum which seems inconsistent with the aforementioned unaltered glycated hemoglobin levels. A later evaluation of the OHIE evidence interrogates this inconsistency in chronic disease diagnosis and outcomes to conclude that despite the higher probability of diagnosis which Medicaid confers upon the participant, “it does not significantly increase the likelihood of diabetic patients receiving recommended care such as eye exams and regular blood sugar monitoring.” In other words, the tether between treatment and diagnosis cannot be taken as a given, indicating that Medicaid coverage alone cannot be said to engender significantly better health outcomes, even if it does affect diagnostic rates.
These physical health metrics, however, only comprise a singular narrative in a much broader dialogue on overall patient wellbeing, with self-reported health and mental health data presented by the OHIE as other crucial considerations. Despite not having a statistically significant effect on a number of measures of physical health, those within the treatment group were 13 percentage points more likely to self-report their health as good, very good, or excellent. In 2018, the principal investigator of the OHIE returned to the self-reported evidence collected in 2008 to identify an area within which there was the largest difference in self-reports between the treatment and comparison groups: the prevalence of undiagnosed and untreated depression dropped by over 50% for those with Medicaid. As such, there is something to be said for Medicaid’s positive impact on the participants’ perception of change in health outcomes, particularly within the realm of mental health.
The Limited Generalizability of the Oregon Health Insurance Experiment
Despite the study’s landmark status, researchers have vocalized worries regarding its generalizability, which is a concern requiring address before the OHIE’s findings can be applied more broadly. The research team themselves were quick to note how the results they published on the treatment groups’ physical health metrics (i.e. those which saw limited change) might be limited in their “statistical power.” Indeed, as the OHIE became a widely referenced body of evidence upon the promulgation of the Affordable Care Act, this critique was frequently expressed. Jonathan Cohn, public policy commentator and then senior editor at The New Republic, released a piece in 2013 within which he similarly asserts that “the Oregon study can't disprove that Medicaid produced physical health benefits, because it can't pinpoint the results with enough precision.” If this line of argument has merit, and the OHIE truly was too narrow in its sampling and too hazy in its findings, then researchers’ ability to leverage this evidence in present public insurance and health policy debates is at risk.
However, the OHIE research team, at least in part, anticipates this concern. In order to enhance their statistical power, they “prespecified analyses of subgroups in which effects might be stronger,” which includes near-elderly populations along with those who report diagnosis of certain chronic conditions (i.e. diabetes, hypertension, and elevated cholesterol levels). Put differently, these groups of individuals were identified as those which would reap the greatest potential benefit in terms of health outcomes from Medicaid (a relatively healthy person might not see a boost in health outcomes purely because their health status had less room to improve). Although, one could still plausibly worry that the OHIE has yet failed to categorically deny the existence of any physical health benefits under Medicaid.
What seems to inform an even more potent critique of the OHIE’s findings are the massive shifts which have defined America’s health insurance landscape over the last decade. To what extent is one able to claim that the expansion Oregon’s 2008 expansion of their OHP Medicaid program resembles the broad-scale inflation of Medicaid’s scope beneath the purview of the Affordable Care Act? To what extent can the conclusions based upon a 2008 Oregon-specific Medicaid expansion be applied to present-day questions of Medicaid expansion in other states with markedly different low-income populations? To grasp the reach of this limitation, it is necessary to understand how significantly the Affordable Care Act transformed the face of public health insurance policy in the United States. Before the Affordable Care Act was enacted approximately 18% of nonelderly Americans were afforded insurance coverage through either Medicaid or the Children’s Health Insurance Program (CHIP) – the vast majority of which were children, parents, pregnant, or disabled. For low-income adults that were childless their public insurance options were more bare, with only 9 states providing state-funded benefits to childless adults in 2009 (before ACA). The Affordable Care Act marked a paradigm shift in this standard, requiring states to provide Medicaid to all low-income adults which were below 138% of the federal poverty level – even those without children. Whether or not one stands in support of this piece of legislation the indelible imprint which it has made on healthcare in America is undeniable. Now, compare this behemoth bulk of health policy to Oregon’s 2008 OHP Medicaid growth: the two clearly exist in two entirely different policy landscapes.
Although this does not completely preclude the ability to apply the OHIE’s findings to current issues. While the Oregon Health Plan was not nearly as expansive in its Medicaid coverage as the still-dominant Affordable Care Act (i.e. it was limited in the amount of low-income enrollees it could admit), it similarly keeps the population of childless low-income non-disabled adults in its crosshairs. Moreover, the OHIE never purported to evaluate the comparative effects of different manifestations of Medicaid. Rather, the OHIE researchers had a simpler mission in mind: compare those without Medicaid (the comparison group) to those with Medicaid (the treatment group selected via lottery). Thus, while the Affordable Care Act has radically refigured the character of Medicaid, we can still use evidence from the OHIE to inform policy decisions which would involve granting previously uninsured individuals at least some level of public insurance (most applicably, Medicaid). The OHIE offers answers about a black-and-white distinction – from no insurance to some insurance – it is not primarily speaking in gradations (unlike some of its counterparts, like the RAND Health Insurance Experiment and its language of degrees of cost-sharing).
Implications for Current Medicaid Expansion
The OHIE, despite its incessant place in health insurance discussions, provides no singular prescriptive answer to how future expansion of Medicaid should be handled, but it does highlight salient tradeoffs and benefits which Medicaid coverage might still impart upon its recipients. Upon achieving legislative success, the Affordable Care Act was met with resistance given the degree to which it sought to expand the entitlement for Medicaid. Certain states, backed by a Supreme Court decision, have been able to avoid expanding their Medicaid eligibility requirements in precisely the way Obama had envisioned. However, as political pressure grows, more and more states – such as North Carolina in January of 2023 – begin to adopt these new Medicaid requirements which were stipulated by the Obama Administration. With these policy moves, comes a number of low-income individuals who transition from the status of uninsured to attaining some degree of Medicaid coverage (especially since this expansion tends to involve a raising of the maximum entitlement income). Clearly, questions of Medicaid and its impact on healthcare outcomes and usage remain pertinent.
There are three main OHIE findings which these states, many of which are on the verge of growing their Medicaid programs, should bear in mind. Firstly, the addition of insurance into an individual’s life is likely to increase their access and use of healthcare with respect to emergency medicine, primary care, and preventive medical services. This could likewise increase the incidence of moral hazard, wherein an uptick in healthcare spending is heralded by more comprehensive insurance coverage, which in turn could affect the efficacy of a state’s healthcare system. Second, physical health outcomes are unlikely to experience significant changes as a result of newly provided Medicaid coverage. This is a critical datum for state officials who see a more expansive public insurance program as the mechanism for enhancing health outcomes amongst their constituents. Note, however, that this does not mean that conversations of health insurance should be divorced from conversations of health outcomes. Rather, this finding could equally indicate that health insurance is just one piece of the mosaic of health in America. The fact that other factors should be taken into account when addressing poor health outcomes in underserved communities does not mean that health insurance considerations should be categorically neglected. Lastly, although certain physical health outcomes may not see significant advancements directly following these policy shifts, beneficiaries of newly expanded Medicaid programs are prone to perceive enhanced results regardless of the actuality of these results. This could prove useful in interpreting public opinion on potential Medicaid expansion.