COVID-19 Epidemiology, Part 2

COVID-19 Epidemiology, Part 2

I continue to track the UK data each day, recording now the reported 'deaths' as well as 'cases' ❋❋. And I continue to plot the Log(10) and find a straight-line; i.e. there is a good fit to 'logarithmic growth'.  But, in order to make my curves responsive to good news in the form of a slowing in the rate of transmission, I plot only data from the last 7 days. Currently (data to 29/3/2020) I find a doubling time of 3.8 days for cases, but 3.3 days for deaths. 

John Burn-Murdoch (@jburnmurdoch) Tweets on Sunday 29th that the doubling time for the death toll in 2.8 days. I wonder why the discrepancy. (And would appreciate discussing this in detail: cawstein@gmail.com)

(❋❋ I realize that there is some doubt as to 'cause of death' when COVID-19 is present, and some doubt about 'cases' when tests are not used, but those problems are inherent in the data, and not my problem here.)

COVID-19 Epidemiology, Part 1

COVID-19 Epidemiology, Part 1

    I have been tracking the numbers of reported cases in the UK each day since the 4th of March, and plotting the logarithms on a spread sheet. On my semi-log plot, the slope of 'line of best fit' on 21st March was 0.1035 which corresponds to a doubling time of 2.908 days.  Assuming that the doubling time remains 2.908 days, the extrapolation to 14th April (41 days from the 4th March) indicates 1,573,620 cases of COVID-19.  

    There are odd points that lie off the line, probably showing how difficult it is to get data pertaining to a standard time of day. There may be a slight amelioration in the rate of infection during the last week, for when I calculated the 'line of best fit' on day 12 the doubling time was 2.77 days, extrapolating to 1,906,777 cases by 14th April. 

[Keep an eye on this space for I shall add updates,

 as the data accumulate.]

    The naif average death rate in UK on 20th March was 177 deaths per 3,983 cases cases, i.e. 4.44%. That is likely to be an underestimate as some 'cases' may become 'deaths'; but, if that ratio of deaths to cases is maintained for the extrapolated 4 weeks, we would have, on 14th April, a UK death toll of 69,869.

     To put that in context, there were 1,793 road deaths in the UK in 2017. We have not experienced a pandemic of this scope and gravity since the 'flu epidemic of 1918/9 which claimed 228,000 lives in the UK (when the population was 43.9 million).

---.---.---.---

    The UK Government is now trying to lower the current rate of spread. From 23rd March, schools and universities are to be closed, and train services are to be curtailed. There is a clamp down on social mixing in pubs and restaurants. The measures are stern but not 'draconian'. We shall look for a change of slope on the semi-log plot, but only after the incubation lag of perhaps 7 days.

     The doubling time is presumably determined by the number of transmissions per case. Modelling of this process is useless without accurate data; the number of unknowns is considerable. Suppose that an uninfected person (A) internalises a virion on day 0, which multiplies internally for 6 days when the host starts to shed virions, i.e. becomes infectious. Suppose symptoms appear on day 7 and continue till day 14, while shedding of virions continues till day 21.  The most dangerous period is clearly between days 6 and 7 for 'cases' will probably become more careful when they know they are infected.

    But this is idle guesswork. We need data.