It wasn’t that long ago that I suspect if someone claimed there would be 10 million reported cases of COVID-19 most people would be skeptical. Now, maybe not so much.
It looks like we’re heading there—perhaps much higher.
Doing just a real dirty, back of the paper napkin type calculation looks like about 45 days to go from .5 million around March 26 to 4 million on May 10. So something like an 8 fold increase in 45 days.
That kind of puts us on track for 10 million in maybe 15 days—end of May?
But will the linear looking global trend keep the same slope?
There was a Metaculus question that opened in early April about “How many COVID-19 deaths will be recorded in the month of April, worldwide?” The community prediction was 210k (50% CI: 165k – 288k), which seemed little different from just extrapolating the trend of reported deaths. I saw that countries had all gone into lockdown a while back, so I predicted 75% that the numbers would end up below 193k. The resolution was 184k and I won a lot of points.
Trend extrapolation is only half of what’s important. If the trend is foreseeably going to break because circumstances are changing, we need to factor that in. If avturchin is right about the recent numbers being linear with 100K cases a day (I didn’t look this up), then we can say that it’ll probably take longer than 60 days until 10M confirmed cases. In the majority of locations, R0 is below 1 and many people are recovering (and PCR tests only catch active infections). Of course, case numbers may go up again, which can happen surprisingly fast. Still, I think the mark for 10M confirmed cases is unlikely to be hit before August. Unfortunately, I suspect that we will hit it at some point later in the year when cases go out of control again in some parts of the world where there’s extensive testing.
UPDATE June 12th: Seems like I got this one really wrong. Daily new cases are at 135k now, so a substantial increase in cases.
The growth is liner from the beginning of April and is around +100K a day in the world. In that linear rate it will be 60 days before 10 millions from now.
Yes. Should have just written the equation y = mX + b and filled in some blanks.
At that m=~100000 (division of the numbers says 77,7777.78) doubles from 4 million in about 52 more days. Silly mistake to look at the 45 days to get an 8 fold increase and then take the implied rate of increase as useful. If it was doubling at some rate it’s not linear!
Seems sometimes I need to not think out loud to actually start thinking....