by Allan Joseph
Welcome to Week 5 of Project Millennial’s summer Journal Club. Previous posts can be found here. This week’s post focuses on “Health Insurance and the Demand for Medical Care” by Willard Manning et al., more commonly known as the RAND Health Insurance Experiment (HIE), after the think tank that conducted the study.
In an ideal world, every policy decision would be informed by reams of research. Even better, that research would primarily be randomized controlled trials (RCTs), the type of research the physical and biological sciences are known for. You can think of them as the closest thing to “perfect” experiments (in most cases): you keep everything the same between two groups of subjects (plants, rockets, lab rats) and change only one thing (sunlight, weight, food intake). Any differences you see (increased growth, shorter distance traveled, obesity) can be attributed to the change, because that’s the only difference between the two groups, ensured by randomly assigning the individual subjects between the groups.
Unfortunately, life isn’t so easy in the social sciences. Want to test how children from single-parent households are different? Well, you can’t just go around randomly assigning certain kids to grow up with only one parent (with good reason), and you can’t just compare single-parent households to two-parent households, because there are a lot of differences beyond the number of parents. This holds true in health policy, too. Very rarely do we get to see an RCT, because they’re ethically difficult. Moreover, they’re expensive.
In the 1980s, however, the federal government commissioned RAND to conduct what would become the most famous experiment in all of health policy — spending the equivalent of $305 million today to figure out how changes in insurance policies affect how people access healthcare.
If you aren’t familiar with some of the terms about health insurance here, I’ve defined them for you in the footnotes — but let’s dive right in.
What did the RAND HIE test? Basically, it was trying to figure out how increased cost-sharing (higher coinsurance ) affected how much people used healthcare services — that is, went to the doctor or hospital. It did this by literally randomly assigning patients to one of four different standard insurance plans, ranging from no coinsurance to 95% coinsurance as well as a plan providing 95% outpatient coinsurance and free (0% coinsurance) inpatient care. For most of the analysis, the “free plan” is considered the baseline, and the authors measure the effects of increasing cost-sharing.
What did it find? Quite simply, the study showed that increasing cost-sharing reduced how much patients utilized healthcare services. Table 4.2, below, shows how that works. Notice how increasing the coinsurance rate down the first four rows results in decreases in utilization, and the final plan (the one with 95% coinsurance for outpatient expenses but free inpatient expenses) goes back up again as the plan gets more generous. The effect was the same across in different income groups, though interestingly, at any given level of insurance, richer patients spent more than poorer ones — see Table 4.3 for details. (Don’t worry about the word “predicted” in the title below, as that has to do with the statistical technique used to come to these results — this isn’t the hypothesis).
The authors can even put a number on this response, called the “elasticity of demand.” That’s an economist’s term to describe how much demand changes in response to a change in price, but it’s a simple concept, measured on a scale of 0 to negative infinity. Imagine the price of a good goes up by 1% and the demand for it drops by 0.5% — the price elasticity would be -0.5, signifying a relatively inelastic good, which is one that doesn’t respond much to price. What if demand went down 5% instead? The price elasticity would be -5, signifying a good that is very responsive to demand.
There’s some nuance to this, but basically, the HIE found the elasticity of demand for healthcare was between -0.1 and -0.2 — that is, a 1% increase in the price the consumer must pay (as opposed to the insurance company) would result in a 0.2% decrease in the amount demanded. By the same token, doubling the price (100% increase) would mean only a 20% decrease in demand. That means while healthcare use does change in response to changing co-pays, it doesn’t do so by a whole lot.
Why does it matter? Good question. Of course, the biggest finding of the experiment is that the more consumers have to pay for their healthcare, the less they use. That would seem to be a point in favor of high-deductible  health plans, which have become a central part of conservative efforts to reform the healthcare system, as it suggests that could reduce costs. However, the evidence is mixed, because it’s important to know whether the patients reduced care they needed or simply just unnecessary care. As one of the RAND authors wrote in 2004, “Those on the political left generally espoused the view that the services were necessary; those on the right, that they were unnecessary. On this score, the results of the HIE had something for both sides.” In that paper, he wrote that, for Americans covered by something like employer-sponsored insurance, the reduction was in unnecessary care, but for the poor and sick, the reduction came out of necessary care, and was harmful to health. Without further definitive research, the debate is unsettled — but the RAND HIE will continue to be cited by both sides for years to come.
Experiments like this one sound like a good idea. Can we run more? The answer is usually no. They’re too expensive, and they’re ethically tricky. Yet they’re still the gold standard for research, and sometimes, circumstances fall into place to allow one — which is just what happened in Oregon recently, on the question of whether health insurance improves health. Project Millennial covered that extensively, but as that debate showed, even RCTs don’t settle questions for good. In Oregon’s case, the study probably wasn’t big enough to get good results on health outcomes, and it’s only been running for two years. The longer the trial runs, the better answers we can get — again, as long as the study population is big enough. After spending millions of dollars on the RAND HIE, at least, the federal government, and the policy community as a whole, got some answers. It would be nice if we got more.
Next week: What are some of the problems in the American healthcare system?
1. “Coinsurance” refers to the proportion of costs the patient might pay when using healthcare. For example, if a doctor’s visit costs $100, someone with 10% coinsurance would pay $10, and the insurance company would pay $90. “Copays” are very similar, but instead of a percentage, they’re just a flat amount: $25 for every doctor’s visit, for example.
2. The “deductible” varies from plan to plan, but it often functions as a limit on how much a patient can pay out of pocket in a given year.
Allan Joseph is a first year medical student at the Warren Alpert Medical School of Brown University, where he is pursuing an MD/MPP. You can follow him on Twitter @allanmjoseph.