Avalere Health updates their Medicaid analysis, now within 6-13% of mine*

Hat Tip To: 
Dan Diamond

*(when each is corrected for certain factors...read for explanation)

A month ago, Avalere Health released an analysis of the Medicaid/CHIP data released by the HHS and CMS for the first 3 months of the ACA enrollment period, October through December. At the time, they estimated that the number of new Medicaid/CHIP enrollments due to ACA expansion as being somewhere between 1.1 million - 1.8 million. I questioned some of their methodology (counting households instead of individuals for some states, not clarifying whether the figures included "woodworkers" or not, not accounting for seasonal variations in enrollment patterns, and not including any January data at all). They acknowledged some of these points, but also pointed out a few valid flaws in my own methodology (primarily forgetting to correct for "baseline churn").

Originally, our respective Medicaid tallies at the time came in at:

  • Avalere Health: A high-end estimate of 1.8 million
  • ACASignups: Around 3.7 million

Correcting Avalere's data to account for the 224,000 extra people that they forgot to acknowledge in "households vs. individuals" states brought their tally up to 2.02 million, while adding in the 984,000+ people automatically transferred from existing state-level programs over to Medicaid itself on January 1st brought their high-end estimate up to right around 3 million people.

Avalere later clarified that their analysis did include "woodworkers" and dismissed the seasonality factor as not being a major factor, but also promised to make both of these factors clearer in a future update. At the same time, they also acknowledged that their first analysis didn't include any January enrollments, which I was already including.

After revising my own methodology to account for the "baseline churn" factor--while still including January's data--I posted the following new numbers last week:

  • Legal ACA Expansion ONLY: 3.27 million
  • Expansion + Woodworkers: 4.29 million

Today, Avalere published an updated version of their analysis, now including the January HHS & CMS reports as well. Their conclusion?

  • Avalere's Low-End Estimate: 2.4 million
  • Avalere's High-End Estimate: 3.5 million

Not only are these numbers much closer to my own (the difference is only around 20-25% now instead of the 2x - 3x difference we saw last month), but they've made sure to include a few other things as well:

...They've made it much clearer which enrollments are included and not included, specifically stating the "woodwork" factor:

Avalere’s estimates include both individuals who are newly eligible for Medicaid as a result of states’ expanding eligibility under the ACA and those who were previously eligible but newly enrolled due to greater awareness of coverage options (also known as the “woodwork” population). Avalere’s estimates remove individuals who would have been enrolled in Medicaid absent the ACA, including eligibility renewals in some states.

...They've acknowledged the seasonality factor and the fact that they haven't accounted for it (not that I blame them):

There is some evidence of seasonality in enrollment tends, and this analysis does not control for such fluctuations given data limitations.   

Finally, they've acknowledged the "households vs. individuals" factor, although they still are not correcting for that (I use 1.8x as a household mulitplier):

in a limited number of cases, states have reported households as opposed to individual applicants, and this is not adjusted for.  

Here's the good news--I do adjust for the household/individuals factor for the 5 states which report Medicaid/CHIP in this method (Alaska, Connecticut, Nevada, North Carolina and Oregon), using a very conservative 1.8x multiplier factor (the U.S. Census Bureau says it's 2.6x). I don't include Nevada at because they unfortunately also mix redeterminations (renewals) with their numbers, but if you add up the other 4 states, the additional individuals per household add up to about 319,000 people!

My new model assumes that only around 133,000 of these are either Expansion Only or Woodworkers, but even at that, Avalere's tally would still go up a notch:

  • Avalere Low End (thru 1/28), corrected for Households factor: 2.53 million
  • Avalere High End (thru 1/28), corrected for Households factor: 3.63 million

There's a bunch of additional caveats/disclaimers for individual states, but those were mostly already included in their prior analysis.

Finally, at my end, there's the February vs. January factor. Avalere's data only runs through the end of January, while mine includes a lot of data from February. Retroactively applying my new projection model to January 28, I get the following:

  • ACASignups.net ACA Expansion ONLY (thru 1/28): 2.9 million
  • ACASignups.net Expansion + Woodworkers (thru 1/28): 3.83 million

Voila.

A true apples-to-apples comparison using the same time periods and correcting for the households/individuals factor brings Avalere's latest total to within 6-13% of mine depending on whether you're looking at the high or low end of the range. I may be overestimating the true numbers by a few hundred thousand, or they may be underestimating them, but it looks like we're both pretty confident that the truth is somewhere in that range.