RAND Then and Now

To commemorate the 60th anniversary of the RAND Corporation, which was established as a nonprofit institution in 1948, stories about “RAND Then and Now” have appeared in RAND Review throughout 2008.

From Microdata to Megatrends

The Continuing Legacy of the RAND Health Insurance Experiment

Although many experts agree that there has been inadequate progress in U.S. health policy over the past three decades — a period in which health care reform has been on and off the policy agenda — there has been steady progress in the analytical tools used to inform health policy. And now those tools may be positioned to play a stronger role than ever before.

The story begins with the RAND Health Insurance Experiment. Begun in 1974 and completed nearly a decade later in 1982, the experiment remains the largest and most sustained evaluation of health insurance design effects ever conducted and stands out today as the only randomized trial of the effects of cost sharing on medical service use, quality of care, and health.

While the experiment’s influence on U.S. health policy has been widely acknowledged, a less visible legacy has been the evolution of health policy “microsimulation models”: computerized statistical tools that model the individual behaviors of people, families, or firms to estimate the likely aggregate outcomes of a change in health policy. The model that RAND researchers created in 1988, based on the results of the Health Insurance Experiment, has been used in many subsequent analyses to predict spending and insurance choices and has inspired the development of many health policy microsimulation models being used today.

AP IMAGES/JOHN MILLER Retirees Harry and Joyce Smith hold on their laps evidence of an emerging trend: their long-term health insurance policy. The Smiths, of Green Valley, Arizona, are among eight million Americans with long-term care insurance — an area in which insurers expect big growth as baby boomers move into their senior years.

Building the Foundation

Because policymakers cannot know in advance how people will respond to a change in policy, being able to simulate likely consequences — both intended and unintended — becomes crucially important, especially in the complex area of health policy. Emmett Keeler, Joan Buchanan, and other RAND colleagues designed the original microsimulation model based on the actual behaviors that were observed in the Health Insurance Experiment.

“Our original model predicted the impact of different kinds of insurance on total medical spending versus out-of-pocket spending and estimated the value of that spending to the patient — precisely, how the mythical ‘economically rational person’ would pick insurance,” said Keeler.

The model used statistical estimates to “simulate spending on episodes of treatment by a representative group of families, an effective and fairly new approach to the study of demand,” Buchanan noted. The simulation showed that cost sharing affects the number of episodes of illness for which people seek treatment and has much smaller effects on the amount of treatment in each episode of treatment. The simulation also showed that small deductibles can restrain excess demand and that large individual caps on out-of-pocket spending can limit risk without greatly increasing spending.

In the late 1980s, the RAND model was used to simulate the effects of a mandate that all employers provide health insurance to their employees. The model estimated that both health service use and employers’ liability for health care costs would increase.

The model was used again in the 1990s to simulate the effect of adding medical savings accounts (MSAs) to the other features of the Health Insurance Reform Act of 1995 (also known as Kassebaum-Kennedy). Republicans had pushed for MSAs, but Democrats feared they would be utilized by only the healthy and wealthy and would damage the traditional insurance market.

“The resulting controversy threatened to kill the legislation,” said Keeler, “but simulations showed that the overall insurance market would not be harmed. As a result, the two sides were able to reach a compromise, and the main part of the bill, including a demonstration of MSAs, passed.”

Modeling the Future

The original simulation model was a “static” one that predicted what would happen in a year. As the field of health policy simulation has progressed, a new generation of “dynamic” models has been developed that can project changes in the health and spending of people over time.

One example is RAND’s Future Elderly Model (FEM), which helps forecast trends in health, health spending, medical technology, longevity, labor supply, and earnings. FEM is being used to examine such trends among people over 50 in the United States, Denmark, France, Germany, Italy, the Netherlands, Spain, and Sweden.

“Because of FEM, policymakers will be better equipped to design social programs that improve health with the lowest effective public and private expenditures,” said Dana Goldman, who developed the model with his RAND colleagues. FEM is at the heart of research being done in the RAND Roybal Center for Health Policy Simulation.

RAND Health’s Comprehensive Assessment of Reform Efforts (COMPARE) initiative has developed a new microsimulation model that builds on what has been learned from prior RAND modeling. “Starting in early 2009, COMPARE will provide online access to the results of simulations on the effects of coverage-related policy changes on the number of people with insurance, spending, consumer financial risk, and health relative to what would occur in the absence of any change in policy,” according to Elizabeth McGlynn, who co-leads the initiative with Jeffrey Wasserman.

Using COMPARE, which will be available to the general public, RAND and other users will be able to see the extent to which various health care reform proposals achieve their objectives. RAND will examine the effects of both incremental and multifaceted policy changes. The results will provide a common base of knowledge for policy debate and development, carrying on the legacy of health policy simulation begun more than 30 years ago.

Related Reading

“Can Medical Savings Accounts for the Nonelderly Reduce Health Care Costs?” Journal of the American Medical Association, Vol. 275, No. 21, June 1996, pp. 1666–1671, Emmett B. Keeler, Jesse D. Malkin, Dana P. Goldman, Joan L. Buchanan. Also available as RAND/RP-540, 1996.

Full Document

The Demand for Episodes of Medical Treatment in the Health Insurance Experiment, Emmett B. Keeler, Joan L. Buchanan, John E. Rolf, Janet M. Hanley, David Reboussin, RAND/R-3454-HHS, 1988, 135 pp., ISBN 978-0-8330-0845-9.

Full Document

Free for All? Lessons from the RAND Health Insurance Experiment, Joseph P. Newhouse, Insurance Experiment Group, Cambridge, Mass.: Harvard University Press, 1993.

Mandating Health Insurance Benefits for Employees: Effects on Health Care Use and Employer’s Costs, M. Susan Marquis, Joan L. Buchanan, Emmett B. Keeler, John E. Rolph, Man-bing Sze, RAND/N-2911-DOL, 1989, 43 pp.

Full Document

Modeling the Health and Medical Care Spending of the Future Elderly, RAND/RB-9324, 2008, 8 pp.

Full Document

“Simulating Health Expenditures Under Alternative Insurance Plans,” Management Science, Vol. 37, No. 9, September 1991, pp. 1067–1090, Joan L. Buchanan, Emmett B. Keeler, John E. Rolph, Martin R. Holmer. Also available as RAND/RP-205, 1993.

Full Document

RAND COMPARE Web site

www.randcompare.org (public access scheduled for mid-January 2009)