For 25 Years, the RAND Health Insurance Experiment Has Stoked Competing Claims
By Emmett B. Keeler, Joseph P. Newhouse, and Robert H. Brook
Emmett Keeler, who has been an economist and mathematician at RAND since 1968 and a professor at the Pardee RAND Graduate School since 1975, directed segments of the RAND Health Insurance Experiment. He was recently voted membership in the Institute of Medicine. Joseph Newhouse was the principal investigator of the RAND Health Insurance Experiment when he was a RAND researcher in the 1980s. He is now the John D. MacArthur Professor of Health Policy and Management at Harvard University. Robert Brook is vice president of RAND and director of RAND Health.
Completed in 1982, the RAND Health Insurance Experiment stands out as the only long-term, experimental study of cost sharing and its effects on medical service use, quality of care, and health. It is possible to take two contrasting perspectives on the lessons of the experiment with respect to today’s health care debate, in which the notion of cost sharing, or shifting a greater share of health care expense onto consumers, has returned to prominence.
On the one hand, the study suggests that cost sharing can help achieve the fundamental goals of cutting costs and reducing waste without damaging health or quality of care for most people. Participants who paid for a share of their health care used fewer health services than did a comparison group given free care. The cost sharing did not significantly affect the quality of care received by participants.
On the other hand, the experiment shows that cost sharing can be a blunt tool, reducing both needed and unneeded health services in roughly equal proportions. Although cost sharing had no adverse health effects in general, there were exceptions: Care for hypertension, dental health, vision, and selected serious symptoms worsened for the sickest and poorest patients under cost sharing. Therefore, the study concluded that cost sharing should be minimal or nonexistent for the poor, especially those with chronic disease.
Our goal . . .
Subsequent RAND studies have underscored the mixed effects of cost sharing. Our goal in this essay is not to conclude that cost sharing is good or bad but merely to explain its mixed effects so that policymakers can make sound decisions based on a full set of facts.
In the early 1970s, the effects of cost sharing took center stage in the national health care debate. Back then, the debate focused on free, universal health care and on whether the benefits would justify the costs. To inform the debate, a RAND team designed and carried out the Health Insurance Experiment, whose fieldwork began in 1974 and ended in 1982.
The team recruited 2,750 families from six sites across the country, encompassing more than 7,700 people under the age of 65. They were randomly assigned to one of five types of health insurance plans created specifically for the experiment. One type offered free care under a fee-for-service plan (the patient fee was zero). Three types of plans involved varying levels of cost sharing under a fee-for-service plan: 25 percent, 50 percent, or 95 percent coinsurance (the portion of medical charges that the families had to pay). The 95-percent plan closely resembled the high-deductible catastrophic plans being discussed today. The fifth type of plan offered free care from a nonprofit health maintenance organization (HMO).
For poorer families in the cost-sharing plans, the fees were capped at one of three levels (5, 10, or 15 percent of income) or at $1,000 annually (about $3,000 annually if adjusted from 1977 to 2005 levels), whichever was lower. Families participated for three or five years. The upper age limit for adults at the time of enrollment was 61, so that none of them would become eligible for Medicare before the experiment ended.
To assess the use, costs, and quality of care, RAND served as the insurer and oversaw the processing of claims. To assess participant health, RAND oversaw the administration of comprehensive physical exams and also of surveys at the beginning and end of the experiment.
Reduced Use of Health Services
Cost sharing reduced the use of nearly all health services. Both adults and children with cost sharing made one to two fewer physician visits annually and had 20 percent fewer hospitalizations than did those with free care under the fee-for-service plan (see Figures 1 and 2). Declines in use of care were similar for dental visits, prescriptions, and other services.
|Figures 1 and 2 —|
|Both Adults and Children with Cost Sharing Visited Doctors Less Often . . .||. . . and Were Admitted to Hospitals Less Frequently Than Were Patients with Free Care|
SOURCE: Free for All? Lessons from the RAND Health Insurance Experiment, 1993.
Consumers in the HMO had 39 percent fewer hospitalizations than did the consumers with free care who used the fee-for-service system, but the HMO consumers made similar use of outpatient services. Overall, spending reductions under the HMO were comparable to those seen under a high rate of coinsurance.
People in cost-sharing plans spent less than others did on health care, but the savings came from using fewer services rather than finding lower prices. Those with 25-percent coinsurance spent 20 percent less than did participants with free care, and those with 95-percent coinsurance spent about 30 percent less than did participants with free care (see Figure 3).
Figure 3 —
By Using Fewer Services, Not by Finding Lower Prices, People with Cost Sharing Spent Less on Health Care
SOURCE: Free for All? Lessons from the RAND Health Insurance Experiment, 1993.
NOTES: Spending amounts account for adults and children. Annual average amounts have been adjusted to 2005 dollars using the all-items Consumer Price Index.
The reduced use of services resulted primarily from participants deciding not to initiate care. Once patients entered the health care system, cost sharing only modestly affected the intensity or cost of an episode of care.
Similar Appropriateness and Quality
Cost sharing reduced the use of health services at all levels of effectiveness, from highly effective care to less effective care, and in roughly equal amounts among most groups of participants. Both the proportion of inappropriate hospitalizations and the proportion of inappropriate use of antibiotics were the same for cost-sharing and free-plan participants.
Two striking findings emerged with respect to quality of care. First, cost sharing did not significantly alter the quality of care delivered. Second, the overall level of quality was surprisingly low for all participants: The RAND criteria for quality were met only 62 percent of the time. By contrast, patient satisfaction was generally high, and it did not vary at different levels of cost sharing among the fee-for-service plans.
Cost sharing reduced the use of health services at all levels of effectiveness.
It is important for all parties in the debate to acknowledge that the quality of U.S. health care has not improved in the past 20 years. In 2003, a nationwide RAND study showed that quality criteria were met only 55 percent of the time. Thus, despite tremendous technical progress that has raised the potential value of care, there has been no clinical progress in the quality of care received as a proportion of the best care available.
Limited Health Effects
For most participants, the reduction in services induced by cost sharing had no adverse effects on health. However, the poorest and sickest 6 percent of participants fared better under the free plan for 4 of the 30 health conditions measured.
In particular, 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. Although the number of actual deaths in the experiment was too few to be statistically meaningful, the changes in blood pressure that free care triggered would have reduced annual expected mortality rates by about 10 percent among a large population of poor patients with hypertension.
Free care marginally improved vision for the poorest patients. Free care also increased their likelihood of receiving needed dental care. And poorer people on the free plan suffered fewer serious symptoms, such as chest pain when exercising or shortness of breath with light exercise or work.
Cost sharing had some apparently beneficial health effects. Participants in cost-sharing plans worried less about their health and had fewer restricted-activity days (including time spent seeking medical care) than those with free care in the fee-for-service plan.
Health outcomes at the HMO were no worse than outcomes for those with free care in the fee-for-service plan, but patient satisfaction was lower among people initially assigned to the HMO. These people were less satisfied with care overall than either those who had previously chosen to be in the HMO or those who remained in the fee-for-service system.
The experiment also examined whether shouldering more of their own health care costs led people to take better care of themselves. It did not. Risky behaviors were unaffected. The rates of smoking and obesity, for instance, did not change.
Mixed Blessings Continue
Today’s health care environment differs in fundamental ways from the one in which the Health Insurance Experiment took place. The science of medicine has changed across all dimensions. Managed care has become more prominent, as has prescription drug use. Doctors emphasize preventive care to a greater extent and know more about providing it. Given these and many other systemic changes, we cannot be sure that a similar experiment undertaken today would produce similar results.
Nonetheless, subsequent RAND work has reaffirmed that cost sharing alone, while reducing costs and waste, neither improves the overall appropriateness of care sought by patients nor raises the quality of care delivered by doctors. But perhaps appropriateness, quality, and even savings could all be increased if cost sharing were reduced for people with conditions for which treatment is cheap and effective.
Perhaps appropriateness, quality, and even savings could all be increased if cost sharing were reduced for people with conditions for which treatment is cheap and effective.
For example, a 2006 RAND study has shown that eliminating prescription co-payments for the people most in need of cholesterol-lowering drugs could improve their health and help save more than $1 billion annually by raising compliance with the drug regimen and reducing the chance of hospitalizations. The analysis revealed a clear pattern: For each $10 rise in co-payment, the average compliance fell 5 percentage points.
In this case, combining zero co-payments for sicker patients with increased co-payments (by $10 or $20) for healthier patients would cut the number of hospitalizations by 80,000 to 90,000 a year and the number of emergency department visits by 30,000 to 35,000 a year, yielding the enormous net savings. The study, by RAND health economist Dana Goldman and colleagues, confirms the insight from the Health Insurance Experiment that cost sharing can endanger the sickest patients, setting them up for more costly care down the road.
Other recent RAND work on cost sharing has also reached conclusions reminiscent of the Health Insurance Experiment. RAND health economist Melinda Beeuwkes Buntin reviewed studies of the rapidly growing phenomenon of “consumer-directed health care,” which shifts more of the cost of routine care onto consumers by means of high-deductible plans (similar to the Health Insurance Experiment’s highest coinsurance plan) that are often paired with tax-free health savings accounts. The rationale for these plans is typically based on the Health Insurance Experiment’s finding that, on average, greater cost sharing will reduce the use of services and will not affect health.
Buntin and her colleagues found that people who switched to consumer-directed health plans did spend less on health care and did use fewer medical services. However, they also experienced mixed effects on quality of care, reported lower levels of satisfaction, and said they lacked adequate information to make informed choices about medical care. Experts interviewed for the study recommended changes in these health plans to protect vulnerable populations and to provide incentives for appropriate use of health services. In ongoing work, Buntin is evaluating the effects of the next generation of these plans on cost, access, and quality.
Despite calls to modify cost-sharing plans over the years, no one has yet implemented a policy that verifiably exploits the benefits of cost sharing (lower costs, reduced levels of unnecessary care, no overall harmful health effects for most people) while avoiding its negatives (reduced levels of necessary care, some deleterious health consequences for poorer and sicker patients, potentially higher costs down the road). Although poorer patients can be exempted from cost sharing, the question remains open whether a superior policy alternative can be put into place to avert the other negative consequences of cost sharing. That question has gone unanswered since the Health Insurance Experiment first posed it 25 years ago. Perhaps for that reason, proponents and opponents of cost sharing have pointed to different results of the experiment ever since.