COVID-19: A closer look

Remarks on 2020-March-25

Understanding the new plots The graphs for # Confirmed and # Deaths are now relative to the same time: the time elapsed since day that # Confirmed passed 10 Confirmed in each data set, according to the regression line for that data set. A nice effect of this choice is that countries where the growth rate in # Confirmed is slowest are furthest to the right (Japan and Singapore), and those with the highest are the furthest left (USA). Another effect is that if you look at # Confirmed and # Deaths in say Canada at "20 days since ...", these are for the same calendar date.

Other minor changes As before, the straight lines drawn on top of the data are the linear regressions of log10(counts) vs time, from which the growth rates and doubling times in the legend are calculated. I've added one-sigma error bars on these estimates, though these are underestimated because this is a basic analysis and most of the error comes from choosing the segment of data on which to perform the regression, and how much data before/after to include.

Remarks on 2020-March-22

Confirmed There is consistency in growth rates across the different countries considered here. This suggests that any discrepancy in counting protocols (delayed reporting, increase in testing as case count increases) is perhaps not so significant. But Japan's slow growth rate is an outlier.

Deaths Again, there is consistency in the growth rates across the different countries. Here the outliers are Korea, and Japan again.

Confirmed vs Deaths There is good consistency in the growth rates between # Confirmed and # Deaths. This consistency, true also for Japan, suggests the low Japanese numbers might not be due to insufficient testing or faulty reporting, but reflects a genuinely slower growth rate. BUT we're still talking about relatively low number of deaths for Japan so the analysis is not too reliable.

Japanese numbers The numbers stand out but it isn't clear why [see this interesting article]. There is a sentiment that this won't last, and that Japan will eventually see the growth rate others have experienced. But I'm skeptical: I feel like it would have happened already if it was going to. Have a look at Japanese live cams or news reports... not much distancing, but a lot of disinfecting. Here are my leading hypotheses in no particular order:

  • Mode of transmission: Some have speculated Japan's low growth rate has to do with the natural physical distancing in Japanese culture (rare public displays of affection or hugs, bowing over handshakes, etc.). But if you look at news feeds and live cams, restaurants, bars, and even amusement parks are still opened. With cherry blossom viewing, the usual large groups will be/are having communal picnics and booze under the cherry trees. But what you see (and smell if you're there) is a lot of disinfectant, everywhere! Not something we're seeing as much (and certainly not smelling) here. The low growth rate in Japan could be the result of nearly eliminating transmission via indirect contact (touching contaminated surfaces), possibly aided by the low direct physical contact practices in Japan.

  • Genetics or pre-existing immunity: After all, China and Japan probably share a lot of fauna/flora and it's not unthinkable that Japanese people have had more exposure to these CoV over history. Even looking at historical records, a CoV outbreak probably wouldn't look much different from a very bad flu year if no one is being tested for virus specificity. So could there have been past, undocumented CoV outbreaks in the region that might have left Japanese partly protected, either via genetics (selection bias shaped by past local outbreaks) or some degree of pre-existing immunity (ongoing, low-level exposures). But then why no reduced growth rate in China, where the virus originated?