Thread by @Farzad_MD

1/ It's time to look at week 16 (April 12-April 18) #COVIDView

Are things getting better or worse overall in the US?
Are there pockets of concern or good news?
What new data have we gotten in the past week?

How would an epidemiologist look at the data?

https://twitter.com/Farzad_MD/status/1251272054986416128?s=20

2/ Let's run through the big 5:

A) Case growth
B) Hospitalizations/Deaths among Cases
C) Test positivity
D) Non-specific Influenza-Like Illness (ILI) trends- Ambulatory/ ED visits
E) Non-specific Pneumonia/Influenza trends- deaths

& new data:
F) Serosurveys
G) Symptoms surveys

3/ CASE GROWTH

Case growth is pretty flat- ~30k new cases a day

But that's no cause for complacency

30k diagnosed cases now is probably >300k infections a day!

+ some heavy hitters (NY, MI) are down, so that means that a lot of other places are flat (MA, NJ worrisome) or up

4/ DEATHS ...are delayed

This paper got a ton of attention ("87% of all ICU patients die!") but @DrJohnScott nailed it in this string:

https://twitter.com/DrJohnScott/status/1253284899752009728?s=20

"Among 1,151 pts who were mechanically ventilated, 3% had gone home, 25% had died, & 72% were still in the hospital"

5/ This is what worries me the most about this reliance on the IHME models- they rely on deaths to fit their model curves and make projections

Makes sense cause it's less affected by under-reporting, but IT ASSUMES YOU KEEP DOING WHAT PRODUCED THOSE DEATH CURVES

if we reopen..

6/ Then we've changed the course of the outbreak-

sure, they can always refit the models when deaths rise after the outbreak flares again...but that will take weeks to show up, and the number of ensuing infections could be catastrophic

good quotes here:
statnews.com/2020/04/17/inf…

7/ TEST POSITIVITY

to control for changes in testing influencing what we see on diagnosed cases, makes sense to look at percent of tests that are positive

this piece from a while back is still highly relevant statnews.com/2020/03/24/we-…

But I have to say-WE NEED MORE TESTING TOO

8/ It would have ben unfathomable to me that we could be sitting here a month after that article, and be seeing that the total number of tests reported by CDC went DOWN to 575k in week 16, compared with 680k in Week 15.

Now, maybe public health isn't getting all the lab tests?

9/ It could well be!

"Results from testing where patient addresses were not provided or available are being held to allow @QuestDX ... to provide this information, before issuing the public health reports"

It's essential for public health to get complete AND TIMELY lab results

10/ Another problem could be those WONDERFUL point of care test device that give you results within minutes?

surely you've heard of them.

They don't seem to have a way to electronically report to public health yet.

(also some safety concerns)
khn.org/news/abbotts-f…

11/ I am dwelling on the testing piece because in order to reopen relying on contact tracing, we need to have a very high percent of all infections diagnosed. If half are asymptomatic, then we better be doing a bang-up job on the rest.

Are we?

The positivity rate is still 19%

12/ If you're panning for gold (+ cases) and every 5th time you took your pan out of the river you find a gold nugget, you know that there's a rich vein where that came from. That's what this is saying.

We'll know we're doing a good job when it becomes HARD to find a + case

13/ Now, those are official CDC results.

@covidtracking has more updated data (though probably less reliable) scraped from each state-

they show an increase in testing in Week 17, and a dip in positivity rate 🤞

NY (39%->28%)
MI (28%->18%)
are continuing to see drops 😅

14/ But there are also a number of states on that seem to be having increases in percent positivity.

DE 25%->34% ??
OH 15%->19%

IA 15%->20%
NE 10%->19%
KS 9%->16%
Do you see a pattern in those supposedly safe homeland states?

Not time to reopen

15/ Speaking of reopening- here are some notes on who's announced reopening plans:

Most famously, @GovKemp who's been roundly criticized, including by Trump, and rightly so (no decline in cases, several worrisome counties, >15% positivity rate)

thehill.com/homenews/state…

16/ Which brings us to syndromic surveillance (be still my heart)

The national rates of ED visits (NSSP) and ambulatory visits (ILINet) with symptoms compatible with COVID19 continue to drop. That's good.

But averages can be deceiving- what about the state/local level?

17/ The Sentinel ILI sites seem to be dropping, though still a few hotspots

parenthetically, it's nuts that we don't have lab testing for COVID at these sentinel sites so we could actually know what percent of those "ILI" visits were actually COVID.

Give em tests+PPE, & Pay em!

18/ The other problem w Sentinel ILI is that there are too few sites to get good geographic granularity- you don't get much info on the people w ILI, like their zip codes. (also hard to investigate spikes)

This is where ED syndromic shines. But guess what?

No state maps/graphs

19/ If we're going to rely on this data for closures and reopening, then every state should have this on their websites every day

some do:

MI
michigan.gov/coronavirus/0,…

AZ
azdhs.gov/preparedness/e…

NC
ncdhhs.gov/divisions/publ…

FL floridadisaster.org/globalassets/c…

ID public.tableau.com/profile/idaho.…

20/ Finally, instead of looking at just COVID-ILI recorded deaths, we should also be looking at "Pneumonia and Influenza" deaths, in case COVID is under-diagnosed.

There's good evidence (public soon!) of increases in P&I deaths prior to recognized outbreaks in several states

21/ So what kind of forecast would I give the US right now?

Americans' extraordinary physical distancing has -for the time being- calmed down the storm substantially, but there are millions of infectious people, and pockets of real concern if reopening happens too soon, too fast