• Further adventures in cost-sharing

    We’ve discussed the RAND health insurance experiment a number of times on the blog. In a nutshell, that was the randomized controlled trial that showed that cost-sharing, or making people pay for some of their care, led to them using less care. the caveat to this study,was that people didn’t discriminate between necessary and unnecessary care, making some of us skeptical about the wisdom of indiscriminate increased cost-sharing.

    This month, in Health Affairs, is a study that builds upon this work. It’s very much worth your time.

    At the Mayo Clinic, before 2004, you had one of three choices for a health benefit. One cost a lot, but had no cost-sharing at all. Another cost a middle amount, had no deductible, but some co-payments at the time of care. A final plan was the cheapest, and had a more significant deductible (but no co-payments).

    In 2004, these plans were scrapped and two new plans were introduced. Both completely eliminated any cost-sharing for primary care and preventive services. Both had co-payments for specialty care and for ED visits. One of them cost more, and had no deductible; the other cost less and had a deductible. Both also introduced co-insurance, meaning that people had to pay a percentage of all other care, including diagnostic testing, imaging, procedures, and hospitalizations.

    This means that a natural experiment could be run. Although this was not a randomized controlled trial, because of the changes, a number of groups could be compared, and a number of hypotheses could be tested. These included the following:

    1. Would utilization go down with an increase in cost-sharing for specialty care?
    2. Would diagnostic testing go down with an increase in cost-sharing?
    3. Would primary care utilization go up with a decrease in cost sharing?

    Let’s look at the first question:

    The introduction of a $25 co-pay and 10% co-insurance seems to have had a significant impact on the utilization of specialty services. Not only that, but the impact seems to have been steady for years. Impressive, but not unexpected, given the results of the RAND HIE.

    What about the second question?

    Well, imaging dropped off with the introduction of 10% co-insurance, but then started to creep back up. The authors think that imaging may increase over time naturally, and that the change dropped the line down on the Y-axis, but perhaps didn’t change the overall slope.

    But what about the third question. Did dropping cost sharing increase the use of primary care?

    Huh. Dropping the co-pay from $10 to $0 for primary care visits appears to have had no effect on getting people to increase their use of primary care.

    Now first, some caveats. My concerns about hypothesis #1 and #2 are that we don’t have a lot of actual outcomes to go along with these. Decreased use of services is fine as long as it doesn’t result in bad health outcomes. We do know that people got less testing and went to fewer specialty visits, which did save money. But we don’t know if people fared worse, better, or the same when they went without that care.

    But it’s hypothesis #3 that really has me thinking. One of the ways the ACA is supposed to make things better is by dropping all the cost-sharing for primary care and preventive services. This is meant to increrase utilization of those things, which would improve outcomes, and (some claim) reduce costs. But this study showed that dropping that cost-sharing did nothing to increase utilization.

    In other words, cost-sharing may be useful, when increased, to reduce utilization. It may not be as useful, however, when decreased, to increase utilization.

    As Sarah Kliff puts it (stealing one of my favorite metaphors), cost-sharing may be a good stick, but it’s not a good carrot. We need to find better ways to induce people to use the services we want them to use.


    • These results should be expected.

      Perceived gains (i.e. reducing the cost of primary care/prevention) have a much less dramatic effect on behavior than perceived losses (i.e. increasing the cost of specialty care/imaging services) – this is known as loss aversion.

    • What would be a better long-term reaction would be to pay for preventive care, 100% out-of-pocket.
      In fact, if Obamacare passes, what I hope people will consider is not using the insurance to pay for the preventive care from dollar one.
      What could be the result?
      While the deductible is theoretically zero, the insured voluntarily chooses to not use the benefits in return for a lower premium.
      A $25,000 “deductible” lowers the premium by 60%.
      A $50,000 “deductible” lowers the premium by 70%.
      Don Levit

    • Forgive me if this issue as addressed by the study–I don’t have access to the full study– but It would be interesting to know if the length of time people spent in the old plan before changing to the new plan had any impact.. I would think that given time people would adjust to the ten dollar copay to the point where it had minimal impact on their behavior.

    • Dr. Carroll, how do you discriminate between necessary and unnecessary care?

    • I’m a bit skeptical of these numbers, but unfortunately do not have a subscription to Health Affairs so I can’t dig into the details.

      Based on my experience (I’m a health actuary), the 25% reduction in specialty office visits is huge. The “cliff” nature of the graph makes me wonder whether the benefit change wasn’t accompanied by some change in the methodology of counting specialty visits.

      Another question I would have is whether there were any changes in the population between 2003 and 2004. For example, if a Mayo Clinic employee also had the option of being covered under their spouse’s plan, would the high utilizers of specialty care opt out of the Mayo Clinic plan in favor of their spouse’s plan. That would give the appearance of cost sharing influencing behavior, when in fact it is just the result of self selection. Although, to be honest, I don’t think that dynamic could move the needle that much.

      Similarly, Medigap plans are often blamed for over-utilization of Medicare services, with the rationale being that utilization is higher for members with Medigap plans than for those without. But that could just be because those that know they are going to need more services will be more likely to buy a Medigap plan. Blaming the Medigap plan for the higher utilization may be reversing cause and effect.

    • “One of the ways the ACA is supposed to make things better is by dropping all the cost-sharing for primary care and preventive services. This is meant to increrase utilization of those things, which would improve outcomes, and (some claim) reduce costs.”

      What’s the evidence that inducing/compelling everyone to visit their GP more regularly would have a significant effect on “improving outcomes?” Which outomes? For whom?

      People who look after their health and visit their doctors already aren’t likely to do so significantly more often and for them the marginal benefit is likely to be exceptionally small relative to the increased costs.

      “Our findings suggest that the broad generalizations made by many presidential candidates can be misleading. These statements convey the message that substantial resources can be saved through prevention. Although some preventive measures do save money, the vast majority reviewed in the health economics literature do not”


      Those who want to improve health care would do well to identify and focus on specific incentives that will lead to specific outcome-improving-and-cost-saving interventions for specific populations.

      Paying morbidly obese depressed diabetic chain-smokers to see their doctors on a regular basis probably passes the cost/benefit test with flying colors. The case for incentivizing vegetarian marathoners with no obvious health problems to do so is much less clear.

    • Back in early 2010, I came up with my own health care reform proposal based on what I learned from other countries’ health care systems. My big incentive for preventative care on chronic health conditions was to allow insurers to charge up to an extra 50% (that was ineligible for subsidy) on otherwise community-rated premiums for individuals who did not undergo a yearly physical exam or did not actively treat chronic conditions. Like others have previously said in this thread, people are more aversive of losses than gains, and I think this would provide a strong enough nudge to get people to actively take care of their health.

    • Somebody has to point out that employees of Mayo Clinic may not be particularly representative of the general population. Also, I would like to point out that primary care is fundamentally different from other kinds of care, in a way that is not easy to quantify, thus making it difficult to study: people often have a personal relationship with their primary care providers; few have a personal relationship with their radiologists. The nature of the relationship can be expected to have an impact on the response to changes in cost. People may tend to not worry about the cost so much, if their perception is that the money is going to someone they like. They will be less willing to part with their own money, if the money is going to someone that they don’t even know.