Inferring the effective fraction of the population infected with Covid-19 from the behaviour of Lombardy, Madrid and London relative to the remainder of Italy, Spain and England

TL;DR

This study estimates the total infected fraction of Covid-19 in Lombardy, Madrid, and London using a simple model and death data, suggesting higher infection rates and lower fatality rates than previously assumed.

Contribution

It introduces a straightforward deterministic model to infer infection prevalence from regional death data, providing new estimates of infection rates and fatality rates.

Findings

Infection fraction in key regions is around 30%.

Infection fatality rate is estimated between 0.25% and 0.5%.

Higher infection rates imply lower fatality than commonly believed.

Abstract

I use a very simple deterministic model for the spread of Covid-19 in a large population. Using this to compare the relative decay of the number of deaths per day between different regions in Italy, Spain and England, each applying in principle the same social distancing procedures across the whole country, I obtain an estimate of the total fraction of the population which had already become infected by April 10th. In the most heavily affected regions, Lombardy, Madrid and London, this fraction is higher than expected, i.e. $\approx 0.3$. This result can then be converted to a determination of the infection fatality rate $ifr$, which appears to be $ifr \approx 0.0025-0.005$, and even smaller in London, somewhat lower than usually assumed. Alternatively, the result can also be interpreted as an effectively larger fraction of the population than simple counting would suggest if there is a variation in susceptibility to infection with a variance of up to a value of about 2. The implications are very similar for either interpretation or for a combination of effects.