Coronavirus – possible trajectories

I guess the UK line in the Johns Hopkins chart, reported earlier, might well flatten at some point soon, as some other countries’ lines have.


But if we continue at 3 days for doubling of cases, according to my spreadsheet experiment, we will see over 1m cases after 40 days. See:
https://docs.google.com/spreadsheets/d/1kE_pNRlVaFBeY5DxknPgeK5wmXNeBuyslizpvJmoQDY/edit?usp=sharing
and the example outputs attached for 3, 5 and 7 day doubling.

A million cases by 40 days if we continue on 3 day doubling of cases


If we had experienced (through the social distancing and other precautionary measures) and continue to experience a doubling period of 5 days (not on the chart but a possible input to my spreadsheet), it would lead to 25,000 cases after 40 days.

25600 cases at 5 day doubling since case 100


If we had managed to experience 7 days for doubling of cases (as Japan and Singapore seem to have done), then we would have seen 5000 cases at 40 days (but that’s where we are already, so too late for that outcome).

Not a feasible outcome for the UK, as we are already at 5000 cases or more


So the outcomes are VERY sensitively dependent on the doubling period, which in turn is VERY dependent on the average number of people each carrier infects.


I haven’t modelled that part yet, but, again, assumptions apart, the doubling period would be an outcome of that number, together with how long cases last (before death or recovery) and whether re-infection is possible, likely or frequent. It all gets a bit more difficult to be predictive, rather than mathematically expressing known data.


On a more positive note, there is a report today of the statistical work of Michael Levitt (a proper data scientist!), who predicted on February 21st, with uncanny accuracy, the March 23rd situation in China (improvements compared with the then gloomy other forecasts). See the article attached.

Michael Levitt article from The Times 24th March 2020

Coronavirus – forecasting numbers

A few people might have see the Johns Hopkins University Medical School chart on Covid-19 infection rates in different countries. This particular chart (they have produced many different outputs, some of them interactive world incidence models – see https://coronavirus.jhu.edu/map.html for more) usefully compares some various national growth rates with straight lines representing different periods over which the number of cases might double – 1 day, 2 days, 3 days and 7 days. It’s a kind of log chart to base 2.

Johns Hopkins University national trends, log base 2 chart

I’ve been beginning to simulate the outcomes for 2 input data items:

your chosen number of days (x) since the outbreak (defined at 100 cases on day zero to give a base of calculation);

your chosen rate of growth of cases, expressed by an assumed number of days for doubling the cases number (z), and then;

the output, the number of cases (y) on day x.

This spreadsheet allows you , in the last columns, to enter x and z in order to see the outcome, y.

Of course this is only an output model, it knows nothing about the veracity of assumptions – but the numbers (y) get VERY large for small doubling periods (z).

Try it. Only change the x and z numbers, please.

https://docs.google.com/spreadsheets/d/1kE_pNRlVaFBeY5DxknPgeK5wmXNeBuyslizpvJmoQDY/edit?usp=sharing