GSAC Analytics

Notes about covid

Notes about covid

With the advent of Covid-19 in our lives, I developed a new habit. In addition to regularly browsing the archive of preprints (arXiv) on high energy physics (which is part of my duties as “duty”), I began to look from time to time to the archive of preprints on medicine (medRxiv). The other day I came across a couple of interesting works that relate to our situation (I will leave links in the first comment):

  1. Number of tests required to flatten the curve of coronavirus disease-2019. In the work, Korean scientists estimate the number of tests that need to be carried out to “flatten the curve” of the number of infections. It turns out that it is very important to conduct a sufficient number of tests during the lockdown period. Otherwise, the meaning is lost in it. The minimum coefficient that they consider acceptable is 5.7 tests for each positive case. For reference, in Georgia, such a ratio has not been reached since October either before or during the lockdown. What happens during the holidays defies any explanation at all. Conducting only a few thousand tests a day, they doom us to quarantine until the third wave in March (while we can only dream of a vaccine).

2. Simulating the Impacts of Interregional Mobility Restriction on the Spatial Spread of COVID-19 in Japan. The Japanese scientist generalized the well-known SEIR model, taking into account the geographical distribution of the inhabitants of Japan and their movements. It simulates the scenarios of the spread of Covid-19, taking into account the restrictions of different degrees of severity (hard and soft localization of regions). He takes the daily and weekly movements of the Japanese around the country from open sources provided by mobile operators. Interestingly, the simulation results are quite ambiguous (which he himself admits). That is, under some conditions and for some cities, the ban on movement between regions gives a significant decrease in the number of cases, while under other conditions there is no effect at all for other cities. The results are very interesting. There is a model web application that you can play with. Why did I pay attention to this article? I am still haunted by the fact that the primary focus of the spread of the virus in Georgia was Batumi. If the government in October isolated the outbreak for at least a couple of weeks, then in December we would have figures for the whole of Georgia an order of magnitude lower (not only diseases, but also deaths).


1. I really liked the Japanese model. For me, this is the most logical generalization. So far, I’ve seen a ton of generalizations that have taken into account different groups in addition to Suspended, Exposed, Infectious and Recovered. But I didn’t like all of them, because, in my opinion, they were small-order corrections that influenced the simulation results outside of exponential growth. The Japanese introduced a population density matrix for 47 regions of Japan and actually took into account the difference between Tokyo and Yamagata at once. That is, having in hand the statistics of geolocation of mobile phone users (count 99% of the population), you can expand the density matrix to enormous sizes. According to my estimates, limited computing resources will be enough to calculate the grid of necessary measures for each micro-region of the country. Moreover, tourists can be allowed in without any problems, having previously discussed their possible routes. I like this idea much more than the actually failed initiative to develop a mobile application for keeping track of contacts with potential carriers.

2. Also from the Japanese I spied a good definition of what the government does to us – Non-pharmaceutical interventions (NPIs). And we bet, lockdown / not-lockdown. Interventions!

3. In addition to the past. Nowhere until now have I met a serious discussion of the curfew as NPI. Because this is utter crap. Its sole purpose is to train us and in the future to love the intervention of personal freedom.

Tamaz Khunjua, GSAC