Welcome to the Cutpoint Multiverse

As CMS noted in the 2027 rate announcement, they are looking to simplify the star rating process as much as possible and to that end have entertained the possibility of a new cutpoint algorithm. The only alternative mentioned so far is simply “quantiles” but today I’d like to dive into what quantiles we may be talking about exactly as well as what other realistic alternatives exist.

To start, let’s look to other managed care spheres in Medicaid and Marketplace to see what star models are already in place. NCQA’s HPR (health plan rating) system that scores managed care plans uses 4 quantiles to determine star ratings — the 10th, 33.33rd, 66.67th, and 90th—flipped depending on whether the measure is lower-is-better. The simplicity of this system is very attractive. Outliers are no longer a factor. It’s very easy to compute and to understand—simply stack rank the plans in each measure and find the quantile you want with a quick SQL query or a couple lines of code. Just one small problem for payers, though. It completely decimates the quality bonus payments. My estimates for 2027 ratings are that we’d be looking at something like 59% of members in 4+ star plans under that quantile plan. Now there’s nothing magic about those 4 percentiles so perhaps others could be chosen. I tried another symmetrical distribution of 20/40/60/80 and this produced about 71% of members in 4+. This is a much more palatable number but it’s certainly not the same plans making 4 stars so it does produce some winners and losers. Honestly I tried this split as kind of an afterthought (though it’s probably the most obvious idea), but the more I think about it, the more sense it makes.

Let’s look at some of our other options, though. The QRS (quality rating system) used for marketplace plans already is governed by CMS, so that’s a checkmark in the “Pro” column. The way this system works is by setting one reference benchmark based on the pooled performance of the top 10 percent of performers in a measure (on a denominator-weighted basis). The star benchmarks are then set as percentage targets of that number.

This is a hard one to recreate because we don’t know the denominator sizes for all plans in individual members, so the best I could do was weight by enrollment to try to proxy it. The results are equally astonishing in the opposite direction—97% of members in 4+ star plans. I’m sure plans would like this but at this point why are we even doing a star rating system at all? I personally find this measurement system deeply unsatisfying statistically. What is special about these arbitrary scalars of the benchmark?

Thank you for coming to my TED talk of bad cutpoint ideas. So where do we go from here? The examples here show the stakes are very high—existential in fact for many plans. Yet the conversation now needs specific proposals and transparent impact analyses. Porting another system is not going to satisfy all the stakeholders involved. The current clustering system has certain pathologies that create uncertainty. A system that defines quality performance relative to traditional medicare seems to me to get to the heart of what the medicare advantage project is for in the first place, but building out a rigorous way to do this would likely involve even more complexity than the current system—not less. I don’t think complexity is the enemy necessarily though if it’s in service of objectives all can agree to. One thing the industry seems to favor according to the AHIP comment letter is knowing cutpoints ahead of time. I think the obvious counterargument here is that this could potentially lower the quality ceiling. Plans might be incentivized to take their foot off the gas after reaching their benchmarks. While that could be true within a given year I think there would be more benefit over a longer horizon. It could help address the disconnect between effort and results that sometimes hamper plans struggling to move up their ratings.

Feel free to reach out if you want to discuss this further or run through scenarios for your plan. This is a real opportunity to have a voice in reshaping what could be one of the most impactful rulemaking changes in years. [email protected]