Interpreting, analysing and modelling COVID-19 mortality data

Didier Sornette; Euan Mearns; Michael Schatz; Ke Wu; Didier Darcet

doi:10.1007/s11071-020-05966-z

Interpreting, analysing and modelling COVID-19 mortality data

Nonlinear Dyn. 2020;101(3):1751-1776. doi: 10.1007/s11071-020-05966-z. Epub 2020 Oct 1.

Authors

Didier Sornette^{1

2

3}, Euan Mearns³, Michael Schatz³, Ke Wu^{2

3}, Didier Darcet⁴

Affiliations

¹ Tokyo Tech World Research Hub Initiative (WRHI), Institute of Innovative Research, Tokyo Institute of Technology, Yokohama, 226-8502 Japan.
² Institute of Risk Analysis, Prediction and Management (Risks-X), Academy for Advanced Interdisciplinary Studies, Southern University of Science and Technology (SUSTech), Shenzhen, 518055 China.
³ Department of Management, Technology and Economics, ETH Zurich, Scheuchzerstrasse 7, 8092, Zurich, Switzerland.
⁴ Gavekal Intelligence Software, 75016 Paris, France.

Abstract

We present results on the mortality statistics of the COVID-19 epidemic in a number of countries. Our data analysis suggests classifying countries in five groups, (1) Western countries, (2) East Block, (3) developed Southeast Asian countries, (4) Northern Hemisphere developing countries and (5) Southern Hemisphere countries. Comparing the number of deaths per million inhabitants, a pattern emerges in which the Western countries exhibit the largest mortality rate. Furthermore, comparing the running cumulative death tolls as the same level of outbreak progress in different countries reveals several subgroups within the Western countries and further emphasises the difference between the five groups. Analysing the relationship between deaths per million and life expectancy in different countries, taken as a proxy of the preponderance of elderly people in the population, a main reason behind the relatively more severe COVID-19 epidemic in the Western countries is found to be their larger population of elderly people, with exceptions such as Norway and Japan, for which other factors seem to dominate. Our comparison between countries at the same level of outbreak progress allows us to identify and quantify a measure of efficiency of the level of stringency of confinement measures. We find that increasing the stringency from 20 to 60 decreases the death count by about 50 lives per million in a time window of 20 days. Finally, we perform logistic equation analyses of deaths as a means of tracking the dynamics of outbreaks in the "first wave" and estimating the associated ultimate mortality, using four different models to identify model error and robustness of results. This quantitative analysis allows us to assess the outbreak progress in different countries, differentiating between those that are at a quite advanced stage and close to the end of the epidemic from those that are still in the middle of it. This raises many questions in terms of organisation, preparedness, governance structure and so on.

Keywords: COVID-19 epidemic; Life expectancy; Logistic equation; Mortality; Outbreak progress; Stringency of confinement measures.