Insurance Coverage is Central to U.S. Health Policy. But Should It Be?
Why a singular focus on increasing health coverage won't deliver the health gains we seek, and detracts from more productive efforts to reform our healthcare system
Since the 1960s, expanding coverage has been the primary goal of U.S. health policy.
Although simple comparisons of insured and uninsured Americans show that insurance coverage is strongly correlated with morbidity and mortality, these are not reliable estimates of the causal effect of health insurance on health.
The best empirical evidence (derived from randomized experiments) suggests health insurance coverage, on average, does little to improve health.
Policymakers should devote more resources toward achieving other health policy outcomes, such as lowering the cost and increasing the quality of care.
For decades, the terms “health insurance” and “health” have been virtually synonymous in U.S. policy discussions. Decision-makers have taken for granted that expanding health insurance coverage will generate large, beneficial downstream effects on health. Accordingly, a single measure is often used to convey the flaws in America’s health system: the uninsurance rate. It’s not surprising, then, that the three largest health reforms of the last 60 years – the creation of Medicare and Medicaid in the 1960s and the passage of the Affordable Care Act in 2010 – focused mainly on increasing the number of insured Americans. More recently, the GOP’s American Health Care Act of 2017, which failed in Congress, was mostly debated on the basis of its projected effects on coverage, and proponents of Medicare-for-All tout their plan’s promise of near-universal health insurance.
This obsession with coverage is misguided. The highest-quality empirical evidence indicates that health insurance has (at best) a modest effect on overall well-being. Not only is expanding health insurance unlikely to deliver significant health gains, but the myopic pursuit of that objective diverts attention and resources away from more productive policy reforms. Larger improvements in Americans’ health would likely be achieved if policymakers devoted more energy to lowering the cost and enhancing the quality of the care that’s delivered – not merely increasing the number of people with an insurance card.
To be sure, many studies report that insured individuals tend to be in better health, suffer from fewer diseases, visit the doctor more often, and live longer than the uninsured. But it would be a mistake to interpret these correlations as capturing health insurance’s causal effect on health. Insured and uninsured people differ in all sorts of ways that confound simple comparisons. People with higher levels of income, for example, are more likely to be insured. But higher income affects health in many ways that have nothing to do with insurance (such as reducing stress and allowing people to buy healthier food and safer cars). Some of the differences between the uninsured and insured populations – such as income – are readily observable in standard datasets and can be controlled for statistically.
Subtler, unobservable differences between the two groups pose a more serious challenge. For example, people with a high aversion to risk are more likely to have insurance. Risk-averse people are also less likely to engage in unhealthy behaviors like smoking or driving without a seatbelt. Since risk aversion is unobserved, it’s not clear if the better outcomes among insured people are caused by having insurance or are the result of the fact that insured people are generally more cautious in life. Other unobserved differences between insured and uninsured people may matter too, such as the willingness to comply with medical advice and the ability to interpret health-related information. If insured people, on average, are higher on these dimensions than uninsured people – and it is plausible that they are – then the correlation between health insurance and health could be a mirage.
The best tool to resolve these thorny issues and to make reliable causal inferences is a randomized controlled trial (RCT). Because of their cost and complexity, RCTs are rarely used to study health policies. But happily, two RCTs have been conducted in the last half-century that shed light on the impact of health insurance on health.
The first is the RAND Health Insurance Experiment, conducted from 1971 to 1986. In this study, researchers randomly assigned different groups of participants to health insurance plans with varying degrees of cost-sharing. At one extreme, a subset of participants was offered completely free care. At the other extreme, another subset of participants faced a coinsurance rate of 95 percent (the percentage of medical charges the consumer had to pay). There were also gradations of cost-sharing between these two extremes. The results, predictably, showed that people with higher cost-sharing used fewer services. For most participants, however, this decrease in healthcare utilization led to negligible changes in health. A policy brief from RAND provides the following summary:
“In general, the reduction in services induced by cost sharing had no adverse effect on participants’ health. However, there were exceptions. The poorest and sickest 6 percent of the sample at the start of the experiment had better outcomes under the free plan for 4 of the 30 conditions measured. Specifically,
Free care improved the control of hypertension. The poorest patients in the free care group who entered the experiment with hypertension saw greater reductions in blood pressure than did their counterparts with cost sharing. The projected effect was about a 10 percent reduction in mortality for those with hypertension.
Free care marginally improved vision for the poorest patients.
Free care also increased the likelihood among the poorest patients of receiving needed dental care.
Serious symptoms were less prevalent for poorer people on the free plan.
Cost sharing also had some beneficial effects. Participants in cost sharing plans worried less about their health and had fewer restricted-activity days (including time spent in seeking medical care).”
At best, the results of the RAND experiment indicate that generous health insurance coverage can cause improvements in some health outcomes among the poorest and sickest segment of the population. For the vast majority of conditions measured (26 out of 30, or 86%) and the vast majority of the population (those above the poverty line and without a severe health condition), the RAND experiment gives no indication that health insurance coverage produces better health.
The second RCT on the health effects of health insurance is the Oregon Health Insurance Experiment. In 2008, Oregon expanded its Medicaid program through a lottery, selecting names from a waiting list to fill a limited number of available slots. A team of health economists capitalized on this opportunity to compare lottery-winners – most of whom gained Medicaid coverage – with lottery-losers, many of whom remained uninsured. The experiment spawned many research articles, but one paper published in the New England Journal of Medicine in 2013 is particularly relevant here. The authors summarized their findings as follows:
“Medicaid coverage had no significant effect on the prevalence or diagnosis of hypertension or high cholesterol levels or on the use of medication for these conditions. It increased the probability of a diagnosis of diabetes and the use of medication for diabetes, but it had no significant effect on the prevalence of measured glycated hemoglobin levels of 6.5% or higher. Medicaid coverage led to a substantial reduction in the risk of a positive screening result for depression. This pattern of findings with respect to clinically measured health — an improvement in mental health but not in physical health (Table 2) — was mirrored in the self-reported health measures, with improvements concentrated in mental rather than physical health (Table 3). The improvements appear to be specific to depression and mental health measures; Medicaid coverage did not appear to lead to an increase in self-reported happiness, which is arguably a more general measure of overall subjective well-being.”
These results from the Oregon Health Insurance Experiment largely corroborate the findings of the RAND study. Though Medicaid coverage modestly improved mental health, the analysis failed to detect any statistically significant changes in physical health outcomes or in general happiness.
There are many possible interpretations of these results, many of which continue to be debated. Here are some thoughts to keep in mind:
Health insurance creates moral hazard (i.e., the willingness to accept additional risk when the personal costs of an adverse outcome are reduced). It’s the old “seat belt laws increase traffic accidents” phenomenon. When people feel protected, they tend to take more chances. In the case of health insurance, an insured person may be less careful about their diet or less cautious about avoiding injuries because they know that whatever medical costs they incur will largely be paid by insurance. For example, the Oregon Health Insurance Experiment suggests (though the finding did not reach statistical significance) that people who gained coverage were subsequently more likely to smoke.
Mental health seems to be more responsive to health insurance than physical health. The Oregon Health Insurance Experiment supports this view. Perhaps peace of mind – the knowledge that an illness or accident won’t wreak one’s finances – is really what most people are buying when they choose a health insurance plan.
Both the RAND study and the Oregon Health Insurance Experiment found that generous health insurance coverage increased healthcare utilization (i.e., insured individuals were visiting the doctor more often, taking more pharmaceuticals, and undergoing more preventive procedures than if they had remained uninsured). But this increase in utilization, for the most part, did not produce tangibly better health. Two possibilities are worth considering:
Perhaps healthcare is simply less important to health than we often assume. The massive gains in life expectancy and well-being Americans have enjoyed over the last century are more often attributable to advances in public health – such as new vaccines, safer automotive designs, the recognition of tobacco as a health risk, the fluoridation of drinking water, and reductions in workplace hazards – than to breakthroughs in healthcare.
Perhaps visiting the doctor is offset by more negative risks than we realize. Many patients are harmed by medication errors, hospital-associated infections, and other dangers tied to the health system. By utilizing fewer services, uninsured people forego the many benefits that healthcare can provide – but they also reduce certain risks.
The RAND study and Oregon Health Insurance Experiment may have had design flaws that understated the positive effects of health insurance on health. Empirical research is always subject to caveat and criticism. Although these studies were carried out with great care – large sample sizes, rigorous randomization protocols, a broad selection of outcomes – there is always room to wonder whether the findings might have been biased. (For a deeper discussion of these experiments, see here, here, and here.)
Viewing health insurance coverage as a panacea is not only mistaken; it is preventing important issues about healthcare access, value, and delivery from receiving the attention they deserve. Expanding health insurance does nothing to address the growing shortages of healthcare providers, needless delays in the FDA approval process, or the artificial restrictions placed on healthcare entrepreneurs and innovators.