In this month’s Health Affairs, Seghal et al ask a damning question:
What is the association between a county’s partisan lean and COVID outcomes?
There has been significant descriptive, univariate work on this in the past, but the research team builds a much better model to tease out the plausible associations between a region’s politics and its associated information ecosystem, public health infrastructure and lean with COVID mortality.
In this observational study that captured data from a majority of US counties, we compared the number of COVID-19 deaths through October 31, 2021, among counties with differing levels of Republican vote share, using 2020 presidential election returns to characterize county political affiliation. Our analyses controlled for demographic characteristics and social determinants likely to influence COVID-19 transmission and outcomes using state fixed effects. We found a positive dose-response relationship between county-level Republican vote share and county-level COVID-19 mortality. Majority Republican counties experienced 72.9 additional deaths per 100,000 people relative to majority Democratic counties during the study period, and COVID-19 vaccine uptake explains approximately 10 percent of the difference. Our findings suggest that county-level voting behavior may act as a proxy for compliance with and support of public health measures that would protect residents from COVID-19….
This is part of a massive literature on the impact of political identity on health. In my world and research, Trachtman wrote an amazing paper that analyzed the effect of partisanship as a means of adverse selection on gross premiums in the ACA. Premium levels increased as a function of a rating area’s Republican lean through a mechanism where Republicans needed to be sicker/more expensive to insure before they would buy insurance from the ACA relative to Democrats. My first published paper looked at the end of the 2017 Open Enrollment Period where there was a huge (24% to 30%) decrease in enrollment on Healthcare.gov compared to expectations. One of our models found a notable partisan effect. There is plenty of other research that health behaviors are filtered and intermediated through the lens of identity including political identity.
Where I’m getting stuck as a researcher who occasionally does some work in geography and space is how to think about political identity and local geographies’ political structures in a public health context? We are seeing more and more evidence that this matters a lot on decision-making, especially in decision-making that requires people to act in ways that acknowledges the presence of externalities where one person’s behaviors influences other peoples’ probabilities and outcomes. But I’m not sure how to think about this beyond knowing that I need to think and far more importantly, read a lot about this over the next couple of years.
Thursday Morning Open Thread: We Begin
Betty
An area well worth your time to study.
MattF
One confounding factor is that there’s both a community effect and a personal effect- personal political alignment matters as well as the degree of redness or blueness of the community. How do you disentangle that?
lowtechcyclist
This is the sort of challenge that my professors would have described as an “interesting” problem, which I think translated to one that would require months or years to solve.
narya
I would love for there to be a way to use FQHC data, overlapped with the political data, to help with this analysis. Things I can tell you (and my apologies if you already know about these data sources):
1. There is a MASSIVE trove of health center data–every health center reports, every calendar year, on the exact same questions, and there is a bunch of demographic data in there (age, race, sex at birth, SOGI, income, insurance) as well as a lot of info on quality measures, with special focus on hypertension, diabetes, and prenatal care (those data are broken out by race & ethnicity). The UDS data are freely available. I will note that cost data is NOT fully reliable for this; they accumulate 340b money in the costs per patient that does not take into account patient populations with high incidence of diseases that result in high 340b income (e.g., HIV–the top 20 health centers in terms of # and % of patients with HIV have much higher apparent, but not necessarily actual, cost/patient in these data). The zip code table in the UDS data allows you to see what zip codes a health center serves (its service area). They added covid questions for 2021–but it only includes PATIENTS (i.e., someone who had a face-to-face interaction w/ a provider of some kind); what will be missing from this is all of the folks we tested and vaccinated in walk-in or pop-up sites but who otherwise did not come to us for care.
2. The UDS Mapper (udsmapper dot org) allows you to dig into zip code data, including health centers w/ the most patients in the zip code, disease incidence in the zip code, poverty in the zip code, etc.
3. Each health center has been reporting a bunch of covid data every week or two weeks since about April 2020–this includes number of folks tested, number of positives, number vaccinated, and then folks tested/vaxxed/treated w/ HRSA-provided materials. These data are a little squishier and aren’t tied to either of the above, but it should be available by health center.
Another thing I saw (when preparing our original FQHC NAP application) was poverty data over time for specific zip codes. I don’t know how this would fit, if at all, but a health center is a building–and the poverty levels around that building evolve over time, such that 30 years later, a specific site may no longer have low-income folks in the near neighborhood. The HCH program where I used to work was odd in in a different way, too: it was blocks from some of the richest real estate in the city, but also close to a lot of SROs, so the poverty level of the zip code was very misleading in terms of actual need.
Sorry to go on so long–cannot wait to see what you do w/ this.
p.s. we have a great epidemiology/data team, w/ a lot of statistical expertise, in case you’d want to work with them, and a research team as well.
JCJ
If possible it might interesting to go beyond county level. I am looking at a map of Waukesha County where all of Brookfield and some other areas are >70% vaccinated while the area around Eagle is 30-39% with most of the county 50-69%. It is no surprise to me that Eagle’s numbers are so low.
Kent
Personally I think Covid death rates are a symptom of a MUCH larger phenomenon. I have a lot of evangelical MAGA family living in rural red MAGA-land. I visit them and have a pretty good sense of their lives. The fact of the matter is that red vs blue in this country also corresponds to rural vs urban/suburban. You have all of the following to disentangle.
Red counties and red states are less likely to have rural hospitals and rural health care for a variety of reasons no the least of which is the Republican refusal to expand Medicaid which resulted in the closure of many rural hospitals in red states.
Rural areas have zero walkability so people are forced to drive EVERYWHERE and just get less exercise. In many rural areas you can’t even walk around the block. Homes are on busy 2-lane highways with no shoulders.
Rural people are older
Rural diets are crappier. Visit just about any rural restaurant to see what I mean. They also tend to be food deserts. The town my mother grew up in used to have a nice IGA, now just a Dollar General. One has to drive 20 miles to the next town to find a Wal-Mart.
Rural people tend to be more obese at least in my experience.
And that is before you even get into things like differential vaccination rates. Even if you had identical vaccination rates I expect red counties would have a higher death rate because they start out so far behind the curve.
WeimarGerman
Have you considered collaborating with Andrew Gelman at Columbia, or his colleagues on his blog (maybe Jessica Hullman)? I consider him a leading expert on multi-level, Bayesian modeling especially in politics and voting. Jessica is less a social science modeler than Andrew, but also an expert on complex modeling and computational methods.
David Anderson
@narya: I believe my advisor would (justifiably) strangle me if I said yes to this project
David Anderson
@WeimarGerman: I believe my advisor would (justifiably) strangle me if I massively expanded the scope of my curiosities and questions at this time.
Brachiator
@Kent:
Some excellent observations. I also wonder whether older people, without regard to political ideology, were more likely to get vaccinated than younger people. Also, I don’t know if information about vaccines and the medication itself got distributed as well in rural communities as it did elsewhere.
Also, can you measure whether the public health authorities were pro or anti vaccine (to exaggerate slightly).
Jerzy Russian
@David Anderson: You study some really weird shit. I respect that.
different-church-lady
When you rebuild a party brand on a foundation of anti-science paranoia, this is what you get.
Redshift
@MattF: And of course they aren’t completely disentangled. Some of the factors causing worse health outcomes in red counties undoubtedly affect people who aren’t personally conservative. It would be useful to know, but it would be tough since the available data are likely aggregated only with information that’s collected in a health care context.
J R in WV
Perhaps your “advisor” is taking himself too seriously?
Isn’t newer/bigger always better than older/smaller? ;~)
Jesse
@Jerzy Russian: +1 from me! I really appreciate your stuff. It makes BJ special.
Do you know of any work — either at the state level, as policy at the insurer level, or even at the national level, of simplifying these things? One pet project of mine is about reducing the complexity of the state, overall, at the level of interaction of ordinary individuals. (This is NOT an interest on Norquist-style “drown the government in the bathtub. If the state needs to be big for the interactions to be easily accessible, so be it.)
David Anderson
@J R in WV: the best dissertation is a defended dissertation.
Omnes Omnibus
@David Anderson: Brett Favre was once asked what his favorite play was, long pass, improv genius, etc. His response was the kneel-down at the end of a win.
Ohio Mom
@Kent: As a native New Yorker, my automatic reaction when driving through the hinterlands is always, Why do people live here? What do they do — for work, for entertainment, for any sort of errands (car repair, dry cleaning their fancy occasion clothes, etc.), for shopping, for anything?
I am going to have to add health care to my list, for some reason that escaped me.
West of the Rockies
Fascinating material, David! I wonder if conservative regions tend to be identifiably patriarchal–men in authority. Thus more likely to restrict abortion and reproductive rights and health.
Is there more alcohol abuse, less scientific literacy?
Covid outcomes certainly seem in great part dictated by a community’s nature.
narya
@David Anderson: I get that :-). OTOH, you could pick some counties or zip codes with divergent politics and look at the data just for those–you could look at the health centers serving those zip codes and/or look at UDS mapper data for those zip codes . . .
Barbara
Lots of pre-pandemic research has captured this trend in other areas. E.g., Dying of Whiteness: How the Politics of Racial Resentment Is Killing America’s Heartland. This book came out in 2020 and reflects research that took place pre-pandemic. Link to publisher
But research studies on the disparate incidence of reduced longevity and chronic disease, suicide and substance abuse rates by demography and geography go back further. E.g., https://medicalresearch.com/author-interviews/decrease-in-life-expectancy-concentrated-in-rust-belt-and-appalachia/52263/
Refusal to expand Medicaid, putting roadblocks in the path of the ACA, opioid abuse, etc. are either symptoms or causes, depending on who you talk to.
Kent
@Brachiator: I don’t think rural healthcare workers were undermining public health campaigns. I think they were just overwhelmed to be going against the mindless partisan conspiracy theory bullshit that took hold in their communities and spread by social media.
scav
Ok, I’m spitballing it a bit as it’s a long time since I was medical geography adjacent. I’m not quite sure it addressees exactly the issue you’re thinking if, but it’s seems to be getting at an issue that could be described in such a fashion (words being slippery things). I’m sort of stealing an approach from studying cancer hotspots, basically it’s a way of teasing out if there’s more geographic structure in the data.
Presumably you’ve got a population of individuals with known characteristics and can build a model (build a surface) of what the expected distribution of disease would be. This model can even be a geographic model that incorporates known (and measurable) geographic influences. Then, compare your expected map(s) of disease outcome to the observed one, is there still unexplained geographic patterning that might be due to (non-modeled) community influences (in your exact case). The individuals political risk factor being already accounted for, any additional community risk factor in theory should be bobbing in the residuals (I think they’re called) showing up as remaining geographic structure. Ok, it gets messy as you introduce the specific geographic factors into the model as some of the community signal influences things like medical infrastructure which you’ve just pulled from the residual but oh well, life is messy.
Clearly, this is an exploratory statistical approach and comes with all sorts of those caveats. Well, those were the statistical seas I was mist swimming in at the time — there’s probably some Luc Anselin in there as well as cancer hotspots.
I made that as clear as I could with no benefit of morning coffee. I hope at least the gist is there.
PaulB
I think this is going to get more interesting in the years ahead. This is just the beginning. With Republican politicians gutting the authority and budget for health departments, with the growing Republican opposition to basic protective measures, including vaccinations, I suspect the blue/red divergence will grow markedly.
One other thing to watch out for: Republican politicians have also demonstrated their antipathy for testing and numbers. I think that, for their own political protection, this is going to get worse, as well, and that getting accurate data out of red cities, counties, and states will prove to be a challenge in the coming years.
Ksmiami
@Ohio Mom: hmm- you’ve just identified the source of the meth and the opioid scourge…
Ksmiami
@PaulB: As I told my spouse, I don’t need to fight Republicans directly – I just need to outlive them as their stupidity and anti-social behaviors will lead them to early deaths
Chief Oshkosh
I’m in another meeting right now, so apologies if this was covered, but here’s another pertinent research paper just published in the British Medical Journal. A very good rundown of many factors red/blue, urban/suburban/rural, etc.
ETA: editorial/summary at: https://www.statnews.com/2022/06/07/us-death-rates-politicization-public-health/
https://www.bmj.com/content/377/bmj-2021-069308
Political environment and mortality rates in the United States, 2001-19: population based cross sectional analysis
BMJ 2022; 377 doi: https://doi.org/10.1136/bmj-2021-069308 (Published 07 June 2022)
Cite this as: BMJ 2022;377:e069308
Haider J Warraich, assistant professor of medicine123, Pankaj Kumar, internal medicine intern4, Khurram Nasir, chief of Division of Cardiovascular Prevention and Wellness56, Karen E Joynt Maddox, co-director of Center for Health Economics and Policy7, Rishi K Wadhera, section head of health policy and equity research38
NotMax
Not within shouting distance of a scientific survey yet would be (I think) fascinating to obtain aggregate data (month by month) of gross sales of masks via Amazon on a state by state, per capita basis.
Baud
Apologies if this has been posted, but I just saw this on Reddit and thought it fit here.
https://sph.umd.edu/news/deadly-price-pandemic-politics
StringOnAStick
@Baud: I was just about to go find that link; I saw it yesterday. My first thought was “hurry up!”
scav
@NotMax: Just as easy to get ZIP level data and play with multiple geographies. We’re absolutely going to need to pull out rural / urban divides within a state. Plus, need to tease out spatial differences in who is ir isn’t shopping on AMZN and for what. But, would be interesting. Would also be fun to get analogous data from Walmart & Costco & Target (big big boxes) — start teasing out clicks and bricks, plus stuff about who shops where.
JaneE
I have seen this in action, when the ACA went in. Even when you know insurance was needed because medical care was needed, using the ACA to get insurance was not good. Unacceptably so for at least one woman I knew. I saw several people in my exercise class gang up on a woman and try to shame her for using the exchange to get insurance so she could get needed care. They kept asking her why she did it. Doh. Because she had to. OTOH, not going to doctors because you don’t want your pre-existing condition on record until you can’t stand it and then lying to an insurance company to get coverage, and then lying to the doctor to tell them it just happened, that is perfectly fine. Also left with some permanent uncorrectable damage, but them’s the breaks.
Barbara
@PaulB: There is a durable streak in human nature — prevailing strongly or weakly over time depending on many factors — that prioritizes concepts of “inner” purity or cleanliness or strength over “external” means of maintaining health. Christian Science is the obvious epitome of this, but whether the trigger is traditional religious (“all you need is prayer!”) or new age spiritual (“my immune system was meant to handle these kinds of diseases without the addition of toxic chemicals”) these people have always been with us.
What combination of personal and demographic traits makes a person believe that they will survive the plague? In this regard, as many have noted, white people became markedly less concerned about Covid when reports began noting that minority groups were disproportionately affected. So reporters are probably as if not more innumerate than others and don’t seem to understand that — let’s say — 1/3 greater “chance” of dying in a relatively small group still means that a hell of a lot more people in the dominant group are going to die.
For all of that, the idea that you personally were not at risk even if others are is a real temptation for many people. In this case, ideas about “rugged independence” and bravado over not needing the government’s help probably fed into this as well, amplifying an already deadly set of circumstances, such as higher rates of chronic disease.
Another Scott
I’m sure you and others have thought about this much more deeply than I have, and I haven’t read your links, but it seems to me to be a difficult data-quality problem. That is, you can get lots of data by census tract or zip code or Nielsen Ratings or whatever for groups of various sizes. But health insurance is ultimately an individual decision. Does the local Jiffy Lube having Fox News on all day as the default channel mean anything about the political views of the people there? How does one tease out political views as being determinative vs just associated-with? (There are probably more blue-eyed mass murderers than brown eyed, but being a young angry dude is probably more determinative…) Averages and trends over time can show associations and correlations – and that’s important!, but not causality.
It’s probably impossible to find a small yet representative sample that could be asked detailed questions about what they chose and why, and their political views, in a sufficiently unbiased way to get meaningful answers. Humans are suspicious and complicated! Figuring out what persuasion methods work best is probably highly regional – where you stand depends on where you sit…
Thanks.
Cheers,
Scott.
scav
@Another Scott: Even asking them questions about their decisions won’t necessarily get you there. They may be entirely unaware of how much their individually made decisions are being altered / warped by being in the community / gravity well of their surroundings. That’s partially why I find the method of looking at the difference of what we’d expect a person of such a bundle of understood characteristics to behave like (or ail of) and the pattern of the actual behaviors or ailments. Purely individual deviations should be uniformly distributed across space. Community based influences (or unmodeled geographic influences) shoukd show up as geographically structured.
Feathers
The first study I saw that looked at this sort of effect was back in the 2000s. It looked at an energy company that sent out mailers telling customers how much energy they were consuming compared to their neighbors. What they found was that households containing registered Democrats maintained or lowered their energy consumption, while Republican households began using more.
Worked in the polisci dept of a business school at the time. The professors were very surprised.
Scout211
I’m late to this thread, but I find it very interesting. Politics and geographic location as a factor in healthcare choices and outcomes leaves out another factor—income level. I do think all three are a factor.
We live in a rural, very Republican county in California. Healthcare providers seem to be few and far between. I’m not sure about right now but for years there were no Medicare Advantage plans in our county and good luck finding a primary care physician if you had Medi-Cal (Medicaid). There simply were not enough physicians and healthcare options to support a Medicare Advantage plan and while there were Medi-Cal providers, they could not accept any new patients.
The solution for so many (if you have transportation and money for a quality Medicare supplement) is you drive to either one of the three nearest counties for all of your healthcare. For Mr. Scout and me, our doctors and all other healthcare services are typically a 30-60 minute drive, each way. One of our neighbors actually rents an apartment in the neighboring county as their official residence so they can be Kaiser members.
I was shocked last year at statistics by zip code for vaccinations for our zip code (extremely low) when I realized that most people in my area were vaccinated in another county and likely were counted as vaccinated in that county.
So politics, geography and income are all factors. But in my personal observation and experience, politics is the least important of the three.
Another Scott
@Feathers: I received a postcard years ago that seemed to be part of an academic voting study.
It was short and to the point. As I recall it was something like:
“Your neighbors at 123, 125, 126, 128 Evergreen Terrace voted in the last election.” (or maybe it was “… in the last GOP primary.”)
I assume that they wanted to see if a gentle nudge that people could tell if you voted or not, or maybe to see if information about who voted in primaries affected the next election in some way.
Maybe a similar study could be done regarding health insurance, but it would be hard because so much of the information is (rightly) protected by privacy laws. “Your neighbor Mr. Simpson signed up for Obamacare and a year later his BP and cholesterol each dropped by 20 points.”
Cheers,
Scott.
RevRick
White male supremacy kills.
Ethan Grey has a great Twitter thread on what Republicans mean by freedom. He says it has, for them, two basic components:
1). Nobody gets to tell them what to do;
2). They get to tell others (whom they deem lesser beings) what to do.
This explains, Grey says, for example, why they’re up in arms over having their kids wear masks in school, but are indifferent to kids getting slaughtered with AR-15 like rifles.
It also explains their stance on abortion. Or why they come to oppose policies that might benefit them. Or why they resist any efforts to mitigate climate change (because they’re being told to care for others and share the Earth).