Actuarial uncertainty as seen by CMS could be reflected in ACA rate requests

Late last week, the Center for Medicare and Medicaid Services (CMS) released an interim final rule on how a variety of federally funded programs will be responding to COVID-19 from a financial plumbing perspective.  The biggest chunk of the rule is for the Medicare Accountable Care Organizations (ACOs).  These ACOs are provider led groups that take on population expenditure risk.  An ACO that meets quality scores makes money if they can hold spending below a benchmark and can lose money if the spending is over a benchmark.  Benchmarks are determined by past spending and has an assumption that the performance year will be fairly similar to the recent past.  That is not the case this year!

CMS actuaries are trying to figure out what COVID does to spending and therefore to benchmarks.  That is a very tough thing to do as most of the data that will inform that decision has not been reported or even created yet.  But the following paragraph is fascinating:

Based on a typical year, we assumed up to a 20 percent reduction in expenditures for 2020 because of a decline in elective services and the deferral of other services, and we assumed
increases in expenditures due to COVID-19 inpatient treatment and related spending. We estimate that this variation in COVID-19 related spending would roughly double the standard deviation in gross measured savings and losses (expressed as a percentage of benchmark) that would have been determined across all ACOs participating in PY 2020. (MY EMPHASIS)

This leapt out at me for two reasons.  First, a good chunk of small ACO performance bonuses  could be from statistical noise.  A 2018 Health Affairs blog lays out the history:

CMS requires savings to exceed anywhere from 2.0 percent to 3.9 percent before they are shared. However, even those corridors allow for a large degree of random earnings due to chance each year. In the initial rulemaking for the program, CMS accepted that an ACO with 5,000 lives would have a 10.0 percent chance of random winnings in any given year in establishing the 3.9 percent threshold, while a 50,000-life ACO would only have a 1.0 percent chance.

Those estimates were made in non-pandemic times. Variance is likely to increase so quite a few more small ACOs could be randomly declared winners and losers due to noise.

More importantly, if we can assume that the actuaries at CMS are thinking along the same lines as actuaries at most Affordable Care Act insurers in expecting high variance, we should be ready to expect wildly divergent initial pricing bids this month as the first states accept rate filings for 2021. Two actuaries looking at the same data and coming up with the same average cost of an event can still create wildly different fair prices depending on how much variance in the outcome is assumed. The actuary that thinks the variance is tighter will price a product lower than the actuary who is modeling more variance. We are not even getting into a discussion as to whether or not the risk is Gaussian or power or any other distribution. But assuming the same distribution shape but with different variances, pricing will change dramatically.

I do not know what the initial rate requests will be for ACA insurers in 2021. I strongly suspect that there will be a lot more variance in each state. Some insurers will be more optimistic than others in terms of both the assumed average and the assumed variances of a number of notable parameters. The most optimistic insurer will, all else being equal, price lower relative to their current rates than the least optimistic insurer.