Avalere Health Responds to my Response

Last week, Avalere Health issued a report in which they claimed that the total number of new Medicaid enrollees due to the ACA only come to between 1.1 million and 1.8 million people. I took a look at their study and found what I felt were some significant gaps in their methodology.

I contacted Avalere and asked them to take a look at my own analysis. Caroline Pearson from Avalere got back to me today via Twitter with the following responses (so far):

Regarding my "apples-to-oranges" comparison statement (ie, that my 7.5M number includes both "newly eligible due to the ACA" as well as "woodworkers", while their 1.8M high-end only includes "newly eligible due to the ACA"), she states:

Contrary to your recent post, our analysis does include woodwork AND new eligibles; excludes baseline churn & redetermin.

From their explanation of their methodology, their "excluding baseline churn and redterminations" (ie, the normal new Medicaid enrollees and renewals you'd expect prior to the ACA) was evident. For the "includes woodwork" part, I think what she's referring to here is the part which states (for their "upper-bound" number):

Lastly, we added all assessments/determinations for Medicaid and CHIP by exchanges, effectively assuming all determinations made by healthcare.gov result in a determination for a newly-enrolled individual. 

It sounds like she's saying that 1.8 Million includes "woodwork" enrollees while 1.1 Million would be without "woodwork" enrollees. If I misunderstood this earlier, I apologize for the misread, and instead of comparing my 3.7 million "expansion only" number against 1.8 million (their "upper bound" number), I should be comparing my 3.7 million against their 1.1 million (their "lower bound" figure)...which I did mention as a possible alternative range.

Regarding the "seasonal variation" factor, Ms. Pearson states:

We know there is seasonal variation in application rates, but we believe it varies by state based on a variety of factors.

Good to know...but they didn't mention anything about seasonal variations in their report, even in the footnotes. They simply stated that they compared Oct - Dec 2013 against July - Sept. of 2013 without mentioning anything about seasonal variations. I guess her point here is that Avalere believes that these variations mostly cancel each other out, making the sequential-quarter issue a non-factor. If so, fine...but they didn't mention this, and the woman from CMS did say it's a factor (though not how much of one), which is why I brought it up. As I stated, in my analysis: "I have no idea whether either of these are a factor, but they're both worth looking into."

Finally, Ms. Pearson pointed out an issue with my own analysis:

Major prob with your method is not separating baseline churn from woodwork. New ppl enroll in Mcaid every month absent ACA.

Ouch. Point taken. Here, she isn't referring to Medicaid renewals (which neither of us includes), nor is she referring to Medicaid expansion (ie, those who were not legally eligible prior to the ACA but are now), nor is she even referring to the so-called woodworkers (ie, people who were previously eligible before the exchanges launched but who never enrolled in Medicaid for a variety of reasons). She's talking about people who simply lose their jobs (or otherwise find themselves in a situation where they qualify for Medicaid)...and enroll, as you would normally expect prior to the ACA. This is known as the "baseline churn" she refers to and is a difficult number to pin down, since it's constantly ebbing and flowing as people move onto and off of Medicaid. Supposedly they've accounted for this, while I clearly haven't.

So, that leaves me with one strike against me, but still leaves several questions for Avalere from my post last week:

• The seasonal factor discussed above is still an unknown
• The 984,000 "mass-transfers" from existing (non-Medicaid) programs (the biggest difference)
• The "households vs. individuals" factor (around 224,000 people)
• The new data added since the end of December

So, in all, there's still a good 1.2 million+ difference between our figures...mostly because Avalere isn't including anyone who was transferred into Medicaid from an existing state-level program such as LIHP in California, Commonwealth Care in Massachusetts and so on. Now, you can certainly argue that these people shouldn't be counted as "not being previously insured", but you have to count them as "enrolled in Medicaid due to the ACA" since the ACA is the very law which allowed them to be transferred into Medicaid in the first place.

As for the new data since December, Avalere has already stated that their data only ran through the end of 2013, so there's some amount of additional new Medicaid enrollees who have come in in the 5 weeks since that time not included in their analysis.

I'll post another update if Ms. Pearson gets back to me about these factors or any other items that I may have missed.