Is there a political divide in how the COVID-19 pandemic has moved across the US?
To summarize: There are still no convincing country-level associations between COVID-19 deaths and either social vulnerability or the proportion of individuals over age 60 in U.S. counties on exploratory analysis, but I expect differences to emerge over time. Cases are more concentrated in urban areas, but death rates per population are similar in urban and rural areas. Hospital and ICU bed capacity is concentrated in urban areas, which will likely present a logistical challenge as the epidemic progresses in rural areas. 2016 election results largely track with population density, ie urban/rural analysis.
And here’s the actual analysis:
Aaron and I were wondering if we could see differences in counties’ experience of the COVID-19 pandemic based on which candidate won the county in the 2016 election. These are purely exploratory, no formal stats, and again using publicly available NY Times data. Data are current as of April 3rd.
We wanted to revisit the idea that population density plays a role in how the virus spreads in a region, figuring this will become more clear as time goes on. This plot graphs population density against cases, both on a log scale.
Blue dots are counties that went for Hilary Clinton in the 2016 election, red dots for Donald Trump. Many blue counties are both more dense and have more cases at this stage (the upper right of the graph). It isn’t really surprising to learn that more densely populated areas (so…cities) tend to be both more democratic and have more cases. The colors look like they separate out even more when deaths rather than cases are considered, with (roughly) more densely populated blue counties on the right and less densely populated red counties on the left :
BUT what is interesting here is that deaths per population are much more similar across more and less densely populated areas compared to cases. In other words, the death RATE is similar in urban and rural areas at this stage, figuring that ascertainment of deaths is pretty good compared to the limitations of general population-level testing in each area. What this also suggests to me is that we are seeing the baseline death rate here when resources are generally NOT overwhelmed in most places. My concern for areas with fewer resources to begin with is that they will get more overwhelmed earlier in their community’s experience with the virus moving through.
As a next step we pulled in general medical and surgical hospital beds, followed in the next graph by ICU beds, for each county. The size of the dots in the following two plots corresponds to the # of available hospital and ICU beds, respectively:
A lot of the hospital bed capacity, particularly ICU beds, is located in urban areas. Again, this isn’t surprising, but as cases ramp up in rural areas the logistical challenges of transporting patients to available beds may complicate more rural regions’ response.
We looked at the Social Vulnerability Index again, and found there doesn’t appear to be a pattern between political result in the 2016 election, COVID-19 deaths, and level of social vulnerability in a county.
Meaning thus far, this has been an equal opportunity epidemic across the breadth of the US, which I’m sure isn’t precisely true everywhere. I have a working theory that any underlying vulnerability in a community will be exacerbated as time goes on.
For one final topic, we again looked at the percentage of a community over age 60. If there were a perfect correlation between this % and COVID-19 impact, we would expect to see the dots clustering along a 45 degree line running diagonally towards the upper right of the graph. Instead, there are counties with both younger and older populations, some that have fewer deaths and some with more deaths. The red dots do look to trend somewhat older on this analysis.
To summarize: There are still no convincing country-level associations between COVID-19 deaths and either social vulnerability or the proportion of individuals over age 60 in U.S. counties, but I expect differences to emerge over time. Cases are more concentrated in urban areas, but death rates per population are similar in urban and rural areas. Hospital and ICU bed capacity is concentrated in urban areas, which will likely present a logistical challenge as the epidemic progresses in rural areas. 2016 election results largely track with population density, ie urban/rural analysis.
