Thread by @Bob_Wachter

1/ Covid (@UCSF) Chronicles, Day 34

Long thread today (sorry); lots to cover after that thing we used to call a weekend.

@UCSFHospitals, we’re on downslope, w/ 17 cases, 3 intubated. Still only 1 death since start. ZSFG also stable: 26 cases; few admits from homeless outbreak

2/ SF also stable. Total of 1216 cases; new case # s zigging/zagging, avg ~40/d in past wk. Last 4 d: no new deaths in SF (vs. NYC, still ~2000 new cases & >150 deaths/day; both improving) (Reminder: NY is 10x population of SF.) SF hospitalizations also going down (note 2d lag.)

3/ @UCSFhospitals, starting to test all hospital pts this wk (for virus); already testing procedure/surgery pts. As we start testing more asymptomatic pts, expect very few positives. Of our last 1000 viral PCRs @ucsf (mostly in pts WITH symptoms), just 10 (1%) have been positive.

4/ I’ll spend most of today on antibody testing. But first, steel yourself and read Donald McNeil’s @nytimes masterpiece on what the next few yrs of Covid may look like nyti.ms/2VFqeuB Also listen to #TheDaily, where McNeil adds color commentary nyti.ms/2RU83jB

5/ Remember McNeil (#TheDaily on Feb 27 nyti.ms/2RU83jB) was 1st to predict need for national stay-home, ubiquitous masks, & deaths on scale of Spanish Flu (below). In other words, he got it utterly right, so his new projections, while a lot to absorb, are highly credible

6/ Some takehomes: a) “Hammer & Dance”, w/ 2 steps forward, 1 step back, maybe for yrs (term from @tomaspueyo bit.ly/34Ryhsk); b) Social strife as people get angrier about hassles & economics of distancing; c) Dystopian “immunity passports” & even possible Covid Parties;

7/ … & d) how to make/distribute 300M vaccine doses? He notes fastest vaccine in history took 4yr, & max vaccine capacity in US now <10M/yr.

Strikes me as realistic (& painful) portrayal of likely road ahead. We should gird ourselves, while of course hoping for better news.

8/ OK, now topic d’jour: antibody testing. This is confusing! Before diving into @Stanford study that got ton of attention this wkend, I’ve tried to distill what u need to know about Ab testing onto 2 pages (no way it'd fit in a tweet) – hope it’s helpful. Page 1 below (& Fig):

9/ And page 2 of my Covid Antibody Testing Primer is below:

10/ If you want an even deeper dive into antibody testing, here are three useful resources: @nytimes nyti.ms/2Kx9ssB (on problems with the current crop of tests), @statnews bit.ly/2KgWFds (on Abs & immunity) & a general one @JAMA_current bit.ly/2zicIFL

11/ Now let’s consider the @Stanford pre-print (meaning a study that's not yet been peer-reviewed) bit.ly/2yyGtBy. Lots of press attention – here’s @nbcnightlynews on it bit.ly/2VXwfmH (story is at 7:50, I’m on at 8:40).

12/ My bottom line on study (which concludes that true # of Covid infectns is 50-85x higher than diagnosed cases): I'm skeptical. Major critiques:

a)Ab test may be less reliable than cited specificity of 95%. Slight dip (to ~98.5%) could mean many of the positives are false +s

13/
b)Recruiting via Facebook may bias to nonrandom sample (? more w/ prior symptoms)
c)Results conflict w/ evidence from prior studies, like Diamond Princess (bit.ly/3cynsOv); Iceland (bit.ly/2xBPpq5); Vo, Italy (bit.ly/2VErcas) & prior pandemics

14/
d)If the true # of infected people was really 50-85x more than reported cases, then everyone in NYC would be infected (current NYC cases 133K; x65=8.6M; NYC population=8.4M). That’s clearly wrong: if they were, how would 1000s of new infections be occurring each day in NYC?

15/ For more detailed critiques (fascinating if you're methods nerd) try:

Natalie Dean @nataliexdean: bit.ly/3anFRvW
Trevor Bedford @trvrb: bit.ly/2VmbqlF
Balaji Srivivasan @balajis: bit.ly/3eC7qFm
Andrew Gelman @statmodeling: bit.ly/34SOtd6

16/ Personally, I’d go with current estimate ->[true cases = ~5-10x diagnosed cases]. In SF, ~1200 dxed cases, implies ~9000 actual cases, or ~1% of city. In NY (133K dxed cases) implies ~1M, or 12% of city. If case-fatality rate ~1%, expect ~10K deaths in NYC, which is ~correct.

17/ Even as I write, another prelim study (from @USC & looking at LA) released bit.ly/3aofUfK & showed similar results (4.1% w/ antibodies). So let’s keep an open mind, though I remain skeptical for reasons above, & would love to see both after they go thru peer-review.

18/ In any case, this kind of antibody sero-surveillance will be important for the reasons above, and you can expect to see many more in the coming weeks. Getting the methods right might seem arcane, but it’ll be crucial to ensuring good decisions.

19/ Whew, pretty heavy stuff. Let’s end with something fun: I present to you (drumroll) The Quarantine Machine! Not clear where they got all that toilet paper, but that’s not the most amazing part bit.ly/3cvrHuk (Thx David Rueben @UCLA for heads up.)

More Wednesday…

Sorry, meant 99.5%, not 95% (as stated elsewhere).