Covid19 mortality analysis
Based on the analysis here by Tomas Pueyo, I have put together an sheet based on recent daily data.
Based on this methodology, millions are currently contracting the virus daily, and in a few weeks tens of thousands will be dying every day and cumulatively 250,000 will have died, regardless of what we do.
Here is a snapshot of the analysis:
I’ll explain the table and show some results. The first two rows show Deaths and Cases for the the world (excluding China). China has been excluded as their initial outbreak is under control and not contributing to present cases. (Data was taken from the John Hopkins dataset). From this, we calculate daily reported cases and deaths.
We assume that it takes 20 days from contraction to death, so can estimate the date of contraction. Obviously not everyone did get it 20 days previously, but we assume this for simplicity.
Then we estimate how many people must have had it at that date of contraction, based on a 1% mortality rate. This means that it takes 100 contracted cases to give 1 death. So if 10 people died today, 1000 people must have contracted it 20 days ago.
However if 1000 people contracted it 20 days ago, they will have been giving it to other people. We assume that the amount of people who contract it increases exponentially (at a growth rate of 20%). So if 1000 people contracted the disease today, we assume that 1200 will contract it tomorrow, so in 20 days time 1000 * 1.2²⁰ = 38337 will contract it daily (this is the Estimated daily new cases today row).
Note: the bold numbers above are all variables in the sheet — make a copy and you can adjust these.
You can see on March 19 there were 1126 world-wide deaths. We therefore estimate that 100 times this number — 112,600 — had it 20 days ago (February 28th). At a 20% growth rate, then this grows to 4,316,813 by March 19. Note that only 27,744 new cases were observed. This means for every person in quarantine, there could 150 people who don’t know they have the disease and may well be spreading it.
We can then use this to project future deaths. At a 1% mortality rate, this means that 43,168 people may die on or around the 8th of April.
We can graph this sheet, to show the trends up to 5 days into the future. Note that the y axis is log scaled. There are 2 methods for calculating cases, one based on today’s deaths, calculating the cases and growing it for 5 days, and then a second one looking at the amount of deaths in 5 days time and calculating the cases (this second method doesn’t predict, but gives a second way of showing current observed cases). Deaths are calculated by growing the estimated current case amount by a further 5 days and applying the mortality rate. It seems to track pretty closely with actual numbers.
We can then forecast not 5, but 20 days into the future (which, remember, is the time from contraction to death). Of the 4 million people who may have caught Covid19 today, we expect 1% to succumb to it, or 40K people.
Changing it to a linear scale makes it easier to see the growth that is projected to happen at the end of the month.
Note that this is based on what has already happened — there isn’t much we can do to prevent it. In fact things may get worse, as we may not be able to cope with such an influx of sick people. At that point cumulatively 250,000 will have died, or more if mortality rates increase.
All we can really do now is to make sure that the spread of the disease is reduced so that this number stops climbing.
