The data doesn't support that. The growth rate has not changed. If that were true we'd see the growth rate slowing, but it's not.
@FractalAudio, has anyone yet devised a way to tell the difference between...
(a.) increase in the number of known cases because more people are infected; and,
(b.) increase in the number of known cases because more people are being tested
...?
I ask because I'm having trouble figuring out how reliable these reported increases in "the number of known cases" actually
are, when we try to use them as a proxy for
actual spread of the virus.
The scenario I keep running through my head goes like this:
It's
Monday morning in Whoville, and medical professionals would like to test sick Whos for coronavirus. But there's limited testing available, so all the testing is rationed: Only the very ill get tested. (Occasionally, once the very-ill person tests positive, other people in the same house will also be tested, but not always. If they other folks are already symptomatic, they're merely presumed to have it, to save resources.)
As a result of testing performed prior to end-of-business on Monday, the numbers released on Tuesday morning show 250 active cases of COVID-19 in Whoville.
At noon on Tuesday, a lot of testing gear and extra personnel show up, so that there is suddenly an excess capacity for testing. Different groups of health professionals respond in different ways:
Group 1, "the randomizers": One group goes out and starts working to conduct entirely-random testing of the Whoville population. They test 100 of the Whos by 5pm, and confirm that 22 of them have COVID-19. Some of these 22 people are asymptomatic, some have a sniffle, some are very sick and feverish, and one is about to be hospitalized
Group 2, "the symptomizers": Another group goes out and randomly tests only persons who have symptoms. They find 50 persons with symptoms by 5pm, and test them, resulting in 35 test-positives.
Group 3, "the hospitalizers": Finally, there's the group that tests people who show up at the E.R. requiring hospitalization. 14 of them show up on Tuesday with symptoms that might be COVID-19. They're tested, and 9 of them have COVID-19; the other 5 have other problems.
All groups are careful not to double-count someone who's already been tested by another group.
On Tuesday night, they tally up the figures (the test-positives provided by all 3 groups), and on Wednesday morning they report that the number of known cases of COVID-19 in Whoville has grown by 66 (22 + 35 + 9) cases, from 250 on Tuesday to 316 on Wednesday.
Whoville's most popular morning-show,
Good Morning Whoville, has a Pretty Blonde Who With Blinding Lip-Gloss, who reports this as a 26.4% daily increase in COVID-19 infections.
But that reporting's complete nonsense, isn't it?
If they hadn't stopped rationing testing, it would have been only reported as an increase of 9 cases, which is only a 3.6% increase! Big difference from 26.4!
But then, on Wednesday, Group 1 ("the randomizers") goes out and finds a new sample-set of 100 people to test. Unlike the set on Tuesday, the 100 people tested on Wednesday produce 24 positive tests.
Group 1 also goes out on Thursday, and finds yet another 100 people to test. This time, they get 26 positive tests. Looking at their numbers across 3 different days (22, 24, 26) Group 1 announces that, in their opinion, the actual COVID-19 daily growth rate is between 8% and 9%.
Which group is right?
I think Group 1 is right, because they're using a statistically valid methodology.
But is anyone in the U.S. doing it that way?
Group 2 is using a methodology that helps us identify people who need to self-quarantine. That's good practical information, but it's unhelpful for guessing the spread of infection.
And the folks in the hospitals (Group 3) ...what useful, actionable information does
their testing give us? As far as I can tell, its sole usefulness is to put all the COVID-19 people in a separate hospital wing, and to tell the nurses to garb up, so that they don't infect others. Or, if there's a known treatment that's good for COVID-19 but useless for other diseases, it'll indicate for which patients that treatment should be used. But it's only distantly, indirectly related to the question of how fast the infection is spreading.
So I'm not saying all these groups shouldn't be doing their tests. Each group's approach has its uses.
But what I fear is that the publicly announced numbers are produced by
mixing numbers from all 3 groups!
The result is a big bowl of hen's teeth, snake-legs, and unicorn flatulence, and
that's what's being reported.