1. Here's quick summary of the main takeaways from our brand new analysis with @CharlottadcM on the Geography of COVID-19 in Sweden: swopec.hhs.se/cesisp/abs/ces…
2. @CharlottadcM was able to get very fine-grained and unique data on the geographic spread & variation of COVID-19 across Swedish cities (municipalities) and neighborhoods (34 of them in Sweden's 3 largest cities).
3. Sweden makes a useful case study because it did not implement a lockdown or have regional variations in public health policies that might impact the geographic spread of COVID-19 in some locations versus others.
4. We looked at factors like density, population size, overcrowding, global connectivity, socio-economic status, presence of at-risk nursing homes, timing of first infection and more. And ran regressions at the city (municipal) and neighborhood level.
4. First off, even without any lockdown, there was tremendous variation in the spread of COVID-19 across Swedish city-regions ... with smaller regions actually hit hardest on a per capita basis.
5. So what factors seemed to matter? Our results are in line with some existing studies and counter-intuitive at times ...
6. Density & size are only weakly associated with COVID 19. Density is associated with COVID-19 cases across municipalities, but not directly with deaths. Density is also not associated with COVID-19 cases across neighborhoods where it is likely to vary more.
7. Population size is weakly associated with COVID-19 cases but not deaths, and it is not associated with COVID-19 across neighborhoods. These results are in line with other studies like the recent LSE study of the US.
8. Our measure of overcrowding based on household size is positive and significant in relation to COVID-19 cases at both the municipal and neighborhood levels, also in line with other research.
8. Our results for income, education, front-line workers and neighborhood disadvantage are also weak and at tomes intuitive.
9. Two factors that seem to matter more are global connectivity (airport connections) and the timing of first infections. Places that were more connected and hit earlier were hit harder by the virus.
10. The variable that is by far most closely associated with COVID-19 across Swedish municipalities is the presence of at-risk nursing homes. This variable outperforms all others in the regression analysis for COVID-19 deaths across Swedish municipalities.
11. That said, all these variables explain only a small amount of the geographic variation of COVID-19 across Sweden—in fact, roughly only 10% of the variation in COVID-19 cases across municipalities and just 3% across neighborhoods.
12. Our biggest takeaway is that the geographic variation in COVID-19 in Sweden & the vulnerability of certain places exhibits a great deal of randomness ...
13. We find little evidence that place-based characteristics mattered to the spread of COVID-19 in Sweden, but rather we find the spread to be more random.
14. Places that got hit earlier got hit harder Places with high-risk nursing homes bore the brunt. But there is little evidence that factors like population size or density played any significant role at all.
Here is a link to the full study with @CharlottadcM: swopec.hhs.se/cesisp/abs/ces…
• • •
Missing some Tweet in this thread? You can try to
force a refresh
