This is going to be a bit geeky (even for me).
We have a generalizability problem for ACA research that is likely growing.
Researchers whose questions require individual level data have to go to State Based Marketplaces (SBM). Healthcare.gov is more than happy to release public use files and some limited data sets that are useful for some specific questions. However, any question that requires substantial micro-data requires researchers to use to one of a few state based marketplaces. I’ve used California and Colorado data to answer some recent questions. My co-authors and I have reached out to Washington State in the past but the institutional details of the state put them into the study exclusion criteria for that particular question. Plenty of people at Harvard routinely publish on Massachusetts data. Some of my colleagues and I have reached out to other SBM states for other questions (we have our fingers crossed for either data access or the grant to score well…)
These studies from these states always have a substantial limitation that the data is from a single state and may not generalize well.
Establishing an SBM is one hell of a signal that a state’s political alignment and values are different than the value and alignment of the states which rely on Healthcare.gov.
California is great about sharing their enrollment data as a public records request (cheaply available with fast turn-around times). But California has weird institutional features in what plan designs are allowed and how the state regulates the market. Colorado has been an awesome research partner. However they were a year late to Silverloading and now have a funky 1332 waiver as well as a public option program. Massachusetts is Massachusetts. The Bay State has been pounding the drum on managed competition in the individual market for most of a decade longer than the rest of the country. Washington State has a non-Essential Health Benefit mandate as well as a public option program. All of these states expanded Medicaid. Until recently, Idaho was the only SBM state that had not expanded Medicaid. Idaho has recently expanded Medicaid.
More importantly, there is not a ton of enrollment. Enrollment through January 11th is mostly on Healthcare.gov. Over half of Healthcare.gov enrollment is from four states.
Enrollment on Healthcare.gov is mostly a Florida, Texas, Georgia and North Carolina story. Realistically, it is a Florida and Texas story.
Those four states have some commonalities. The biggest ones are Republican elite dominance and the lack of Medicaid Expansion.
The Medicaid expansion eligible population earning (or claiming to earn) over 100% FPL is a substantial component of enrollment in these states. These folks shape the market as more people earning under 138% FPL means the Silverload gap gets bigger and relative prices for not-Silver plans decline relative to Silver.
And we just don’t see research on these states using administrative individual level data sets. This is problematic. I don’t know what to do. My research agenda routinely requires individual level data. Right now means I’m looking for data partners who are willing to share that type of data. Those partners are a subset of SBMs who routinely share data.
We have an increasing generalizability challenge for ACA research.
Fake Irishman
You are correct this is a huge problem.
And it’s one of the challenges of operating in a federal system with 50 different standards.
Cervantes
Do any of these states have an APCD? Why would you need data from the SBM or Healthcare.gov in that case?
David Anderson
@Cervantes: There are some APCDs in Healthcare.gov states. We would need the Healthcare.gov or SBM data to determine FPL, plan design, networks and a few other things. APCDs are good about claims, they’re not good about other things ( given their name, I hope they are really good at claims.)
I was spitballing a project yesterday with a colleague. Our dream data sketch needed 5 linkages. The minimal viable data sketch had 3 linkages (APCD to Medicaid to State Based Marketplace) to protected personal information as well as few dozen linkages to public use area data sets.
Ruckus
We are never going to have the kind of information that you need widely available. Because many states relish being different, in not going with everyone else, with being their own entity. Yes we are a big country but healthcare, while different for different age groups is basically the same for all humans. Yes, some need a lot more, some get by on less, but as we age we all require more, even if we are completely independent assholes. I know people that will only get healthcare when it is almost too late. I’m not sure why, like it’s some level of independence, or an unwillingness to admit problems. I have no idea if this is an American thing or a human thing. And I know that our healthcare system’s most likely response, unless you are actively bleeding is some sort of medication, something that as often as it is good, it is bad or at least not successful. And we charge out the ass for all of it. I’ve had, because of the last federal maladministration, the VA raise the cost of some of my drugs to triple what I used to pay because the drug is an over the counter item that comes 120 to a bottle rather than 90. And I can purchase the same drug over the counter for half of what I was paying before that raise. What actually amazes me is that the healthcare system actually works as good as it does, considering that most of our healthcare system seems to be about filling bank accounts rather than trying to figure out how to better help people that actually need it. The concept that when we are seen by our healthcare system we look like a dollar sign rather than a human being is a telling reflection on our society.
Cervantes
“We would need the Healthcare.gov or SBM data to determine FPL, plan design, networks and a few other things.” Isn’t this all publicly available? I mean, you know the FPL, and the plan design and networks are published by the insurers. They have to be or you couldn’t choose a plan. Am I missing something?
Ruckus
@Cervantes:
I’m betting that David needs to dive into the data at a deeper level, to understand how decisions are made on both sides of the transactions.
Big R
-ponders-
a couple of research thoughts and a policy thought:
1. How do the systematic differences between SBM and HC.GOV states affect what you predict will happen? Is there another policy space where these differences would have similar effects with better data availability? That may be a hook for you to test the generalizability of your findings by proxy. Spitballing off the top of my head, thinking about food stamp eligibility or unemployment seems like there might be some space for informed speculation.
2. The HC.GOV data sets absolutely exist because the same information is getting collected as in SBMs. Could a FOIA request shake them loose? You would need to be very specific about what you wanted and what format you wanted it in and would probably have to pay a hefty fee but at least you’d have the data. Seems like something that a grant could pay for.
3. I recognize that we’re not in a policy space where this matters until 1/3/2025 (we hope), but could we begin advocating for a statutory provision in an omnibus HHS bill that requires HC.GOV to release the kind of data that would be useful for you? Policy implementation review is an important component of public administration and I would think you’d be able to make a compelling case for using it to improve on HC.GOV and the ACA more generally.
chopper
NERD ALERT
David Anderson
@Cervantes: I would need to know exactly what plan Bob chose and his income level, plan menu which might vary by county and zip code and time of the Open Enrollment Period, network (which also might have meaningful variation that I want to exploit) and cost sharing… and then compare Bob’s choice to what Jo did for their household….. That is one of the questions that I’m excited to sketch out for a likely 2025 or 2026 paper.
Xantar
So Maryland hasn’t gotten back to you? I work for the Maryland Department of Health although not for the Exchange. In other parts of the Department, people are usually eager to share data with researchers.
David Anderson
@Xantar: Maryland Exchange has been great — just trying to find the right set of questions to match to their institutional details and our bandwidth.
WeimarGerman
I understand that CMS will be adding more data to the Qualified Entity program which is accessible via application through RESDAC. Perhaps there’s a way to urge the Office of Minority Health (OMH) to add this level of specificity to a data set that they could also publish via the QE program? OMH seems to want to do more to address disparities and you work on plan choice, and complexity of that choice could fall within their domain.
I’m no expert in the programs but have had the pleasure of participating in analysis that used QE data many times.
David Anderson
@WeimarGerman: Can we talk offline