Are there Covid-19 death maps locally adjusted for population density?

As a first motivating example, according to an Oxford CEBM map [even] Sweden has more deaths per capita from Covid-19 (217 deaths/mil.pop.) than the US as a whole (166 deaths/mil.p.) On top of that, according to Wikipedia the US has higher population density (34 per km2) than Sweden (23 per km2). So in that view, Sweden is doing "doubly worse" worse (320 = 217 * 34/23) than the US.

As another data point, Spain for example has 496 deaths/mil.pop. but their population density is 93 per km2. So that makes Spain seen closer to the US--their number would roughly be a third (181 = 496 * 34/93) if (crudely) adjusted by population density to compared to the US.

For Italy, their population density is twice that of Spain (200 per km2) and their deaths per mil.pop. (441) are even less than Spain's. So that makes Italy look good in comparison to Spain in this view. Even compared to the US, Italy fares better (75 = 441 * 34/200) in this perspective.

My calculations are fairly crude adjustments though, ignoring the fact that within countries (esp. US or Sweden) there are large internal variations in population density. So are there some studies/maps that plot the local Covid-19 death rate relative to the local population density (worldwide)? (Of course one needs to do some clustering to produce such maps, so I'm leaving fairly open-ended what "local" means.)

As more motivation for this q, I found a paper on the 1918-1919 flu deaths (per capita) vs population density (in the US):

Investigations of possible links between population density and the propagation and magnitude of epidemics have so far proved inconclusive. There are three possible reasons (i) A lack of focus on appropriate density intervals. (ii) For the density to be a meaningful variable the population must be distributed as uniformly as possible. If an area has towns and cities where a majority of the population is concentrated its average density is meaningless. [...] Here we show that when these requirements are properly accounted for, the size of epidemics is indeed closely connected with the population density.

Relationship between population density d and the size µ of the influenza epidemic of September-December 1918. In the graph m means million. The data are for Indiana, Kansas and the city of Philadelphia in Pennsylvania. Influenza and pneumonia deaths are counted together. It can be seen that the relationship between population density holds only on a broad density scale. Inside of the three groups of data points the background fluctuations are strong enough to override the power law. The regression reads (the confidence interval is for a confidence probability of 0.95): µ = Cdα, α = 0.22 ± 0.08, C = 3.5. Source: Bureau of the Census (1920).

So yeah, the correction for pop density should probably be on a power law, not (linearly) how I've done it in the first part of my question. Doing this (power law correction) instead however, would "disadvantage" the low-density areas/countries even more than how I've done it in the first half of my question!

So, to repeat my question: are there any published models/maps of this kind (adjusting the death rate for population density) for Covid-19?

• @divibisan: contagious diseases spread easier in areas with higher population density (although there may be a ceiling effect). And yes, pop density has been considered in peer-reviewed publications on such models e.g. doi.org/10.1016/j.mbs.2013.04.013 – SX welcomes ageist gossip Apr 27 at 23:53
• the problem is that the average population density doesn't mean all that much. what you want is something like a median population density instead. Canada for example has a low population density, but a lot of people do live in big cities, especially in BC. there are lots of places like that, where most of the population is clustered in a few cities and the rest of the country is fairly unpopulated. those look like low density countries, but each person is likely to be living in close proximity to a lot of other people – Italian Philosophers 4 Monica Apr 28 at 0:15
• @ItalianPhilosophers4Monica: I agree, my correction is pretty crude, that's why I'm asking for better models. – SX welcomes ageist gossip Apr 28 at 0:16
• @divibisan: Even the CDC reckons that's a factor: "Geographic differences in numbers of COVID-19 cases and deaths, cumulative incidence, and changes in incidence likely reflect a combination of jurisdiction-specific epidemiologic and population-level factors, including 1) the timing of COVID-19 introductions; 2) population density; 3) age distribution and prevalence of underlying medical conditions among COVID-19 patients (1–3); 4) the timing and extent of community mitigation measures; 5) diagnostic testing capacity; and 6) public health reporting practices." – SX welcomes ageist gossip Apr 28 at 0:17
• ... and none of the events nor skiing holidays that were responsible for large spread are tied to high population density. – cbeleites unhappy with SX Apr 28 at 7:00

1 Answer

This is not quite a map, but a scatter plot of log deaths/pop vs pop. density. The US is the red dot next to "North America" and Sweden (USA label doesn't show for some reason unless you hover over it in the interactive version)

I'll accept a better answer that does go to a more local (i.e. below-country) level.

I found some data for the US like that (i.e. below country-level), but aggregated by county type (with a non-obvious figure of actual pop density), and it's also pretty dated (~1 month old by now):

As to the question of density itself: Kolko’s analysis finds density to be significantly associated with Covid-19 deaths across U.S. counties. But density is not the only factor at play. His analysis also finds that Covid-19 death rates per capita are higher in counties with older populations and larger shares of minorities, and colder, wetter climates. It’s important to remember that this analysis only looks at the U.S., and in other parts of the world, denser cities have had more success controlling the spread.

• In case anyone is wondering about Mongolia, it probably doesn't appear because it has had zero deaths, which doesn't compute with log graphs. – Andrew Grimm Apr 28 at 11:55
• @AndrewGrimm: well, a lot of the third world countries probably have under-reported deaths from Covid-19 (due to not enough testing), so one should probably not try to infer too much from such cross-country comparisons between countries with vastly different GPD/capita etc. In the "column" (vertical band) that US and Sweden form in that scatter plot you also have Angola etc. – SX welcomes ageist gossip Apr 28 at 12:27
• @AndrewGrimm: another (3rd world) outlier: Burundi, with 1x-2x the pop density of Italy (sources don't even agree much on the pop density in Burundi...). Burundi apparently has 1 death reported. – SX welcomes ageist gossip Apr 28 at 12:37
• Is it just me or is the dot on the scatter plot representing the USA not labelled? I'm guessing logically it's the orange one to the right of the North American average. – Brian Z Apr 28 at 15:56