I don’t think I did a good job in my post this morning of explaining my idea of competitive risk adjustment. So, I’ll try again. Before I do, I want to point out that I can think of many, many variations on the idea. My objective is just to present the simplest possible version I can think of to convey the spirit of it. The rest are details, important for implementation, but this will not be implemented. It might not even be a good idea. I trust people who are qualified to make that judgement will tell me so. That’s why I’m posting about it. Also, should you know of any papers that present this idea or something like it, let me know.
Today, payments to Medicare Advantage plans, as well as the implicit subsidy for traditional Medicare, are set by the Medicare program. Competitive bidding would reverse the flow of information. Rather than the program telling plans what it will pay, plans would provide information, bids, to Medicare on the cost of providing the standard benefit to an average beneficiary. Medicare would then select the lowest bid and pay all plans, including traditional Medicare, that amount as a subsidy for an average beneficiary.
That’s the key. It’s the subsidy for an average risk beneficiary. Medicare doesn’t actually pay the same amount for all beneficiaries. If a beneficiary is sicker, the risk of larger expenses is higher and a plan is paid more. For healthier beneficiaries, a plan is paid less.
Beneficiary risk can be quantified, and is by the Medicare program. For each beneficiary it’s expected (predicted) spending as a fraction of average spending. So, it’s a ratio I’ll call s that is positive and equal to 1.0 for an average risk beneficiary. The expected spending is calculated based on prior diagnoses and demographics.
Now, here’s the competitive bidding part, but with a twist on the style I’ve been writing about for months and described above. Rather than bid one price for an average beneficiary, why not have plans bid on a set of prices for a set of risk ranges? That is, break the range of risk, s, into intervals. As just one example, consider four intervals (the actual values here may not make sense but that is not important conceptually):
- 0.0 < s < 0.5
- 0.5 < s < 1.0
- 1.0 < s < 1.5
- 1.5 < s
Now ask plans to submit four bids, one for each interval of risk. A bid for an interval would be the cost of providing the standard Medicare benefit for an average beneficiary of risk in that interval. The lowest bid in each interval sets the subsidy for that range of s. Actual payment within the interval would have to be scaled since the subsidy amount is for an average beneficiary and, clearly, the interval reflects a range of beneficiaries. But, that’s a detail.
Why do this? One reason is to capture more information about cost from plans than a single bid, across the full range of risk, could provide. By discovering plan costs for high and low risks separately, the program can fine tune payments within risk bands and perhaps do so more efficiently than it can with an administrative pricing system, even one with administrative risk adjustment. Fundamentally, if one believes that the market can reveal costs, but one is worried about favorable selection (cherry picking) with current risk adjustment approaches, then using the market to reveal costs by risk type is one way to address that concern.
Is the basic idea clear? Can anyone see other problems? Can anyone express this more succinctly? Is this nuts?
UPDATE: My brain wandered off while originally writing this and I included a nonsensical paragraph and footnote about low bids. Since what I wrote made no sense, I’ve taken it out. I also added a paragraph that explains why one would risk adjust in this way.
