A little bit about the RAND health insurance experiment

You may recall (though it hardly matters), that I’m preparing a little talk for some Princeton University Woodrow Wilson School undergrads. My charge from their professor is to convey, in 45 minutes, the value for health care policy of research and research methodology. To convince them that health care policy is pretty important and in need of good, evidence-based ideas, I will first subject them to a parade of horrifying graphs.

That will lead to the problem of how to control health care costs, which offers a natural segue to the issue of moral hazard, which slams right up against the RAND health insurance experiment (HIE). I claim that if you only know one thing about the body of research-based evidence relevant for health policy, it should be the RAND HIE. (I’ll have to explain what moral hazard is and the challenges it poses.)

And, if I’m to say anything about the RAND HIE, I ought to at least provide the basic facts about the experimental set-up. Those facts are these:

  • The experiment was conducted between 1971 and 1982. (This is an old study. None of the students to which I will be speaking were anywhere near alive. I was not even a teenager by the end of the study. And yet, it remains one of the most influential studies relevant to health policy. Hundreds of papers are based on it.)
  • It was a randomized experiment. (That means something important that relates to why this is such an influential study, which I’ll explain.)
  • 2,750 families consisting of about 7,700 non-elderly (under 65 years old) individuals were recruited. (The exclusion of the elderly is important for interpretation of results, which I’ll explain.)
  • Recruits were drawn from six regionally diverse sites that also provided an urban/rural balance. (This is relevant for the generality of findings.)
  • They were randomized to four types of health insurance plans, which varied in level of cost sharing from none to 95 percent before reaching an out-of-pocket spending cap, set no higher than $1000 (over $6000 in today’s dollars; cost sharing and this cap were income-adjusted, but that’s a detail). The free care plan also had HMO-style care management. (Plan types with these features exist today: some care management, some high-deductible plans, and so forth. I’ll have to explain what all this gobbledygook means.)
  • Recruits participated for three to five years.
  • The research questions were: (1) How does cost sharing and care management affect health services use? (2) How does it affect appropriateness and quality of care? (3) What are the health consequences?

In a subsequent post and in the talk, I’ll discuss the answers, though they’ve been provided in posts on this blog before, not to mention most readers no doubt know them from their own reading. But still, I’m making slides so I might as well share.

I’ll be you think I can’t do all this in 45 minuets. I’d take that bet. How much? What do you require for proof? (Remember, I’ve had a lot of practice on this blog simplifying things down to their essentials. Not every nuance related to what I conveyed above is essential.)

Source: RAND Research Highlights, The health Insurance Experiment, 2006.

Hidden information below


Email Address*