• Can the market risk adjust itself?

    One of the key concerns about private plans participating in Medicare (or private plans in general) is that they may “cream skim,” or establish benefits and policies that select for healthier than average enrollees. More precisely, under Medicare, if a plan can enroll a beneficiary for which it receives a government payment (subsidy) that is above the cost of financing care for that beneficiary, it will earn additional profit at taxpayer’s expense.

    Taxpayers should not be happy about this. What can be done? The standard economist’s answer is risk adjustment, that is, to pay plans more for sicker enrollees, less for healthier ones. The key is to pay just the right amount more and less so that the plan is indifferent to enrolling a sick vs. healthy individual. The extra payment for the sick and the lower payment for the healthy do not create any incentives for the plan to behave any differently than if all beneficiaries were equally healthy.

    But any economist who knows enough to suggest risk adjustment as a solution to cream skimming also knows that it may be insufficient. The government is not likely to be able to know enough to risk adjust around all cream skimming. The profit motive is compelling. Plans will likely find a way to achieve favorable selection, enrollment of “good risks” relative to payment.

    The problem here is that the government, with incomplete information, is telling the plans what they’ll be paid instead of the plans, which have more information, revealing their differential costs for sick vs. healthy beneficiaries. How does one get plans to reveal their costs? The standard answer is bidding.

    I’ve already written about how plans could bid for government subsidies. The way that system is normally described is for plans to bid the cost of a standard, minimum set of benefits for an average risk (average health) beneficiary. The government then sets subsidies at the lowest bid and pays all plans that rate. (Traditional Medicare can be one of those plans.) Costlier plans or plans offering richer benefits must charge more to beneficiaries in the form of a premium. The incentive to keep costs low is clear. Quality monitoring and bonuses/penalties could inject incentives to provide adequate services. Low-income subsidies can protect the poor.

    One normally adds that subsidies will also be risk adjusted, which under Medicare today amounts to a scaling of the average-risk subsidy (the one the plans bid on). In math, for a specific beneficiary, plans are paid an amount equal to

    S = (beneficiary health risk adjustment scale factor) x (subsidy for average risk beneficiary).

    But that risk adjustment scale factor need not be set by the government, as it is today under Medicare Advantage. Plans could bid on that too. The simplest way is for plans to bid a multiplier. If the current way of computing the beneficiary-specific subsidy is given as S in the equation above, plans could bid on a multiplier r that scales it:

    S’ = r x S.

    The actual r chosen by the government could be the one that minimizes taxpayer costs, or cost to the Medicare program. That is, the beneficiary-specific subsidy would be the lowest according to the competitive bids on the average risk subsidy and the competitive bids on the risk-adjustment multiplier. If a beneficiary wished to enroll in plan that cost more or offered richer benefits, (s)he would have to pay the marginal cost. (For this to work, i.e., for there to be some plan for which the beneficiary-specific, risk-adjusted subsidy was sufficient, the bidding would probably have to occur in stages: Stage 1, the subsidy for the average risk beneficiary is bid and set; Stage 2, the risk-adjustment multiplier is bid and set, with plans taking into consideration the average subsidy from stage 1.)

    One can imagine more complex schemes like breaking up the risk adjustment schedule into segments (ranges of health risk) and having plans bid on a set of multipliers that apply to the set of segments. But those are details. The key is to recognize the distinction between harnessing a market signal (a bid) and setting values by fiat. The former puts the market to work, forcing it to reveal relevant cost information. The latter is a best guess with incomplete information. Taxpayers and beneficiaries should want the former. Provided there is sufficient competition, it would produce the lowest-cost system that minimized cream skimming. The incentive for plans to “game” the system, designing benefits to shed or deter high-risk enrollees would be relatively low.

    The other important thing to recognize is the crucial role of government. It establishes the structure in which to capture the market signal conveyed through bids. That structure is critical. Without it (i.e., in what some might call a more “free market” system), the market signals are indecipherable. A market without structure is cacophony. There may be competition, but that’s all there is. There is no bidding that drives the system in the direction we would like it to go, with lower cost, higher quality, and fewer perverse incentives to game the population risk profile.

    An apt analogy might be to a human and a computer. The former possesses the creativity and perspective to understand what ought to be done. The latter possesses the raw power to calculate rapidly. We would not put the computer in charge of what problems to solve, of what questions to ask. We ought not put the human in charge of rapid arithmetic. Similarly, the market can be exploited as a computing machine. It can tell us what the optimal price or risk adjustment multiplier is, but only if we set up the problem so it produces that answer. Garbage in, garbage out.

    • Austin
      I am not clear on something. Bidding on avg beneficiary cost obvious. However, above a bit fuzzy re: bidding on risk adjuster.

      The point of risk adjustment, is in most cases to retrospectively look back and level the playing field.

      By carriers prospectively submitting bids on their desired tolerance for risk (what they want for the sickest enrollees), are they not subterfuging the very point of why RA exists? RA is performed in the context of what all parties pay and are compensated at the end of the day, no?


      • RA has a lot of approaches. One is prospective, as I wrote. The Medicare Advantage subsidies are risk adjusted in advance. Risk corridors, outlier payments, etc. are also RA, but retrospectively, as you point out. That’s not my focus in this post.

        Point is, patients are heterogeneous. Think of them as different “goods” with the plans “buying” them (don’t think too hard, money flows the other way). Plans bid to buy the patients, to buy their risk, to be responsible for it. Why should they bid one price for all patients? We can break patients up by type (that’s what risk adjustment does). Plans can bid by type. This would be more informative and efficient than the government making an estimate with less information about what the plan’s “bid” should be.

        Get it?

    • No,

      Patients are heterogeneous, too much so. I can follow up if clarity is needed, but proposed system to function, you need to specify what is meant by “break up patients by type.”

      This might be an instance of economist vs clinician seeing the world through a different lens. Are you referring to crude rating bands (basics such as age, gender, geography, ? others)?

      Adding clinical conditions will make the bidding unmanageable I think.


    • Brad, read up on MA risk adjustment. I don’t have time to explain it now. If/when I come across a good resource (there are some, I know) I’ll blog it.

      • This is a good starting-off-point resource… http://www.hccblog.com/

        • Thanks David,
          I am very familar with HCC methodology, its weaknesses, etc.

          It is how these scores will be used in the context of bidding that I am unclear on, “Stage 2.” If you have a reference to make that leap for me, I would be most grateful.


          • Brad, I’m not sure how else to say it, but I will try again. Risk adjustment boils down to a scaling on the base payment rate, that for an average risk beneficiary. Call the scaling factor s. It is in some range of positive values where 1.0 is the value for an average risk beneficiary. s > 1.0 indicates higher than average risk; s < 1.0 is lower than average risk. s is set by Medicare. It's the Medicare Advantage risk adjustment factor. It determines, with the base payment rate, the payment to a plan. It's all set by Medicare.

            Why should Medicare set the payment? Why not have plans tell Medicare how much they require to cover a beneficiary of a given risk s? I tried to sketch out how that might be done in the post, but I don't think I did a good job. I already see problems with it. So, here's another, simpler way (there are many variations, my objective is just to explain the simplest one I can think of):

            Break the distribution of risk into groups like group 1: 0 < s < 0.5; group 2: 0.5 < s < 1.0; group 3: 1.0 < s < 1.5; group 4 s > 1.5. [ACK: FOR SOME REASON WORDPRESS IS SCREWING UP MY ABILITY TO DEFINE THESE GROUPS. YOU DO GET THE IDEA EVEN THOUGH THE TYPING HERE IS WRONG, YES? BREAK S INTO NON-OVERLAPPING RANGES.] now ask plans to bid on payment for each group separately. That’s just competitive bidding stratified by risk. If you understand competitive bidding and understand MA risk adjustment there’s nothing left to understand. Or am I not understanding what you’re not understanding?