Read on Twitter
(TINY THREAD) I want to explain in a bit more detail the reason there was concern about the (now disproven) claims made by the IHME authors last week that last week would be the peak of COVID-19 deaths. We don't know when the peak will be—but today has been the deadliest day yet.
1/ We have been told by medical experts—and everything I write about comes right from experts—that most of those recorded as having died from COVID-19 are dying after being placed on a ventilator, and that the time from infection to a fatality of this sort is roughly 18.5 days.
2/ Public data reveal when sudden bumps in daily infection rates came:
March 21–March 22: +4,500
March 25–March 26: +4,000
March 30–March 31: +4,000
April 1–April 2: +3,500
That final bump put the U.S. over 30,000 new infections daily, which is roughly where we've stayed since.
March 21–March 22: +4,500
March 25–March 26: +4,000
March 30–March 31: +4,000
April 1–April 2: +3,500
That final bump put the U.S. over 30,000 new infections daily, which is roughly where we've stayed since.
3/ Of *course* some percentage of these four bumps were due to increased testing: people who weren't *very* sick, and wouldn't have been tested at all prior to the US testing regimen expanding. But some of these folks had to be the newly very sick who would end up on ventilators.
4/ What the IHME model authors never publicly explained is why these bumps wouldn't show up in increased death tolls approximately 18.5 days later, as experts say fatalities are a "lagging indicator" due to the 18.5-day course of the virus for someone who ends up on a ventilator.
5/ Adding ~19 days to the 4 bumps would suggest the following:
March 21–March 22
April 9–April 10 death toll rise
March 25–March 26 April 13–April 14 death toll rise
March 30–March 31 April 18–April 19 death toll rise
April 1–April 2 April 20–April 21 death toll rise
March 21–March 22
April 9–April 10 death toll riseMarch 25–March 26
April 13–April 14 death toll riseMarch 30–March 31
April 18–April 19 death toll riseApril 1–April 2
April 20–April 21 death toll rise
6/ If you count 18 days instead, you'd expect the first death-toll bump around April 8 (just a rough estimate—surely it could be April 6, 7, 10, 11).
And what happened on April 7? A massive death-toll spike—up 700 in daily deaths—that looked tied to the earlier infection spikes.
And what happened on April 7? A massive death-toll spike—up 700 in daily deaths—that looked tied to the earlier infection spikes.
7/ You'd then expect a *second* death-toll spike—per the data—to come somewhere between April 12 and April 16.
And what's happened today, April 14? A second massive death-toll spike—up yet again (so far) about 700 deaths—and once again it looks tied to earlier infection spikes.
And what's happened today, April 14? A second massive death-toll spike—up yet again (so far) about 700 deaths—and once again it looks tied to earlier infection spikes.
8/ Meanwhile, the IHME model—which constantly self-deletes earlier projections and so always appears "correct"—originally proposed a peak in deaths last week that would (unlike what we've seen everywhere else) *not* "plateau" but steadily decline. That disparity wasn't explained.
9/ Over and over and over again I've said simply that *the experts need to better explain things*. I'm not an expert. But I watch the experts avidly, and none are explaining why infection spikes don't cause death spikes and why or how we are different from Italy (re: plateauing).
10/ So far the basic data experts are putting out on daily infections/deaths and the infection-to-case closure timeline is *predicting* fatalities—whereas the IHME model isn't. All I want is for experts to explain things better to those of us—including me—who aren't experts. /end
