Free for All? Lessons from the RAND Health Insurance Experiment.

• Author: Neuberg, Leland G.

Failure of the 1966 Coleman educational policy report to find causes of student achievement differences soured many econometricians on nonexperimental studies of public policy issues, and turned their attention to field experimentation. RAND's experiment began in response to an early 1970s health care reform debate. By the mid 1970s that debate had receded from the public policy agenda leaving RAND to finish work in political tranquility and publish findings in a series of technical reports, journal articles, and book chapters. This book synthesizes those results (listing the Experiment's publications in a 28 page appendix) and, with the reemergence of a health care reform debate, appears politically timely.

To participate in the Experiment, RAND randomly selected from strata a total of 2000 families in six U. S. locations (chosen primarily to represent variations in health care availability), and sorted them into fee-for-service and HMO insurance plans. Fee-for-service plans differed as to annual cost share percent (i.e., proportion of billed charges paid by participant) - 0 (free); or 25, 50, or 95 up to 5, 10, or 15 percent of family income to a maximum of $1,000. HMO care was free. RAND acted as insurer, charging no premiums and paying medical service providers the proportion of billed charges that participants did not. Rates of refusal to participate and attrition of selected families were low.

Newhouse et al. place the Experiment in political economic context:

Cost sharing appealed to many conservatives, who believed that patients sought (or physicians delivered) much "unnecessary" medical care when care was free. But many liberals saw cost sharing as a barrier, especially for the poor, to receiving "necessary" care. Both, of course, could be right - cost sharing might reduce both necessary and unnecessary care - in which case the overall appraisal might turn on cost sharing's ultimate effects on health and cost. These effects were unknown in the 1970s.

For those on fee-for-service plans, RAND found decreased service use and per-patient cost from 0% to 25% cost sharing (but no further decrease for more cost sharing). All fee-for-service plans enjoyed the same proportion of medically appropriate service use (judged by two physicians blind to patient plan). For the general population, there was no difference between free and other fee-for-service plans on 32 health status variables; for the poor, blood pressure, corrected vision, dental...