There is something funny happening with COVID hospitalizations

Proportion of COVID pts getting hospitalized falling

A lot

Just recently

My theory?

As hospitals fill up, bar for admission rising

A patient who might have been admitted 4 weeks ago may get sent home now

Thread
So what do I mean "proportion of COVID pts hospitalized falling"?

For months, you could reliably predict new hospitalizations

How?

By taking cases 7 days prior, multiplying by 3.5%

That is

3.5% (1 in 29) of those diagnosed today will be hospitalized about 7 days later

2/10
LOTS of caveats to this formula

Not all states report new hospitalizations (formula takes that into account)

Could build a 10 day lag formula (3.7% hospitalized by 10 days)

All data I report are 7-day moving avgs from @COVID19Tracking

3/10
So let's look at data

On October 15, there are 53K new cases

On Oct 22, you’d expect 1844 hospitalizations (53K*0.035).

What was it actually? 1855

So far so good

4/10
On Nov 1, there are 80K new cases.

On Nov 8, you’d expect 2804 new hospitalizations (80K*0.035)

What was it actually? 2604. A little less. But fine

On Nov 15, there are 146K new cases.

On Nov 22, expect 5111 new hospitalizations (146K*0.035)

But there are only 3670

5/10
I’m ignoring data from last week b/c Thanksgiving messiness though this effect gets much more dramatic

But here’s the fact:

Over much of September and October, you could look at cases today and predict that 3.5% of that number gets hospitalized 7 days later

6/10
But in November, that number starts falling

Initially to 3.2% by Nov 8

By Nov 15, drops to 3.0%

By Nov 22, drops to 2.5%

And by Nov 29, down to 2.1%

So what’s going on? What does this mean?

7/12
May be more testing means we're picking up more cases?

No

% test + is rising from 5.1% (Oct 1) to 12.1% (Nov 15)

So number of cases being missed is climbing

Proportion of pts being admitted from identified cases should be RISING

That 3.5%? Should go up. Instead, its falling
So what’s happening?

What you’d expect – doctors’ threshold for admission is likely going up

In early Oct, with plenty of beds, a COVID patient who is clinically borderline likely gets admitted to ensure they do OK

By Nov 15, beds in short supply, that person likely goes home
Of course, critically ill patients always get admitted

But over time, marginal admissions start disappearing

And as hospitals get fuller, what is defined as “marginal” keeps changing

Until you only admit the sickest folks

10/12
And that’s what appears to be happening

Here's the bottom line:

1 in 3 people who would have been admitted on October 1 aren’t being admitted by November 22

That’s a big change!

And given big rise in test positivity – its likely much higher

11/15
Best guess? we're admitting half the COVID pts we would have admitted October 1

Is this a problem?

If those people don’t need hospitalization, is this bad?

Well, on Oct 1, we would have said that they need hospitalization

So “need” is fluid

But here's the other key issue
The evidence suggests that many of these pts now being sent home will likely do worse at home

Some may be OK -- but others will come back sicker or even die at home

And one more thing

When hospitals fill up, threshold for admission for everything goes up
Everything -- COVID and non-COVID is affected

So borderline admission for heart failure? Person now goes home

Pt with infected leg where you’d prefer IV antibiotics in hospital? They go home with oral antibiotics

This is what is happening in hospitals across America right now
This is not Doctors being cruel

Its that with fewer and fewer beds, bar for hospitalizing anyone is rising

And likely means more people are suffering, getting worse, or even dying at home

All because our hospitals are really full

15/17
So what's the bottom line?

We can see in data that we are likely admitting far fewer COVID patients than we would have just 2 months ago

Likely due to very full hospitals

Because political leaders have politicized mask wearing, social distancing

Its a travesty

16/17
So remember:

Hospitalizations are not static thing

When hospitals get full, by definition you can’t hospitalize all the folks you'd like to hospitalize

And many who would benefit from hospitalization suffer

Because only the very sickest get a bed

Everyone else goes home

Fin

• • •

Missing some Tweet in this thread? You can try to force a refresh
 

Keep Current with Ashish K. Jha, MD, MPH

Ashish K. Jha, MD, MPH Profile picture

Stay in touch and get notified when new unrolls are available from this author!

Read all threads

This Thread may be Removed Anytime!

PDF

Twitter may remove this content at anytime! Save it as PDF for later use!

Try unrolling a thread yourself!

how to unroll video
  1. Follow @ThreadReaderApp to mention us!

  2. From a Twitter thread mention us with a keyword "unroll"
@threadreaderapp unroll

Practice here first or read more on our help page!

More from @ashishkjha

1 Dec
Scott Atlas resignation from @WhiteHouse comes not a moment too soon

His time was marred by barrage of misinformation, from promoting anti-mask quackery to falsehoods about testing

He repeatedly tweeted disingenuous "data" & promoted widespread infections as a strategy

Thread
He supposedly was a policy expert, but in my 2 decades of working on US health policy, he was never a serious player

Scott rose to prominence saying on TV what pandemic experts would never say

And found a buyer for his herd immunity strategy in @realDonaldTrump

2/3
.@ScottWAtlas elevated @SunetraGupta, Bhattacharya & @MartinKulldorff...herd immunity advocates who wrote shameful "Great Barrington Declaration"

They claimed to promote "protect the vulnerable" & let others get infected strategy...but didn't do much to protect vulnerable

3/4
Read 4 tweets
23 Nov
If its Monday -- it must be more vaccine news

@UniofOxford / @AstraZeneca vaccine results posted

This is largely more good news

Overall efficacy of 70% but one dosing regimen was 62%, other 90%

A key result reported was safety: no severe adverse events reported

Short thread
What do these results mean and is this actually good news?

Why yes it is

First, I don't fully understand why two dosing regimens were so different in efficacy

One possibility is vector immunity

This vaccine uses adenovirus vector (virus to deliver the genetic material)

2/4
Less efficacious arm used full dose of vaccine in both shots

More efficacious arm was half dose followed by full dose

It MAY be that the full first dose generated immunity against vector, causing second dose to be less effective

We don't know but will need to sort out

3/4
Read 5 tweets
23 Nov
Daily reports tell us what happened last week, 2 weeks ago, 4 weeks ago

This is why folks are worried about what's next

Today's 7 day moving avgs

167K cases, 3600 new hospitalizations, 1460 deaths

What does this tell us about upcoming weeks?

It says we have a problem

Thread
Today's 167K cases? infection happened last week

Today's 3600 new hospitalizations?

Infections happened about 12-15 days ago

Today's 1460 deaths?

Infections happened about 3+ weeks ago

And that's why we have a problem

2/4
About 3.5% of identified cases get hospitalized

CFR right now about 1.7%

So today's new cases will cause:

5000 new hospitalizations in 10 days

2900 new deaths in 3 weeks

Corollary

Today's hospitalizations were cases 10 days ago

Today's deaths were cases 3 weeks ago

3/4
Read 7 tweets
22 Nov
Slaoui, speaking to @jaketapper, lays out an aggressive vaccination timeline

If he's right, we'll see virus spread starting to slow from growing population immunity by end of January into February

And much better by March

How?

Short thread
By end of January, based on timeline laid out:

50M vaccinated

And I suspect 60M infected, recovered by then

Obviously overlap between groups

So likely at 30%+ population immunity by 1/31/2021

That'll slow spread

By end of Feb, probably 40%+

Not "herd immunity"

But

2/4
Herd immunity not like a light switch

As population immunity builds, spread slows

By 30-40% immunity, spread becomes meaningfully slower

When we get to HI threshold, infections won't "disappear" overnight

But they will stop being self-sustaining

3/4
Read 4 tweets
20 Nov
Today was a very, very odd day

I testified before @SenateHomeland

They held a hearing on hydroxychloroquine.

Yup, HCQ

In the middle of the worst surge of pandemic

HCQ

It was clear how our information architecture shapes questions of science and medicine of COVID

A thread
There were 4 witnesses.

3 who strongly supported HCQ

They believed thousands of Americans were dying from lack of HCQ

And then, there was me

This split was not a reflection of evidence or the consensus in medicine

It reflected ability of the majority to seat more witnesses
The hearing was a testament to how politicized science has become

I shared evidence of studies that have failed to find benefit of HCQ

3 other witnesses shared personal experiences

And suggested my testimony was reckless because it would deny people access to lifesaving HCQ
Read 9 tweets
18 Nov
Number of COVID infections is high, rising.

What should we expect between now and December 1?

How many more infections?

Looking at the last 6 weeks gives us some clues

But bottom line?

We will likely have 250K to 300K new daily infections by December 1

Thread
All data from @COVID19Tracking and 7-day moving avgs

Oct 1 to Oct 15

42K new cases --> 53K new cases

This was an increase of 24% off base of 42K cases

Oct 16 to Oct 31

53K --> 78K; Increased 46% off base of 53K

Nov 1 to Nov 16

78K-->150K; Increased 88% off base of 78K

2/4
So infections accelerating off of a higher and higher baseline

If things continue without intervention, expect around 300K or more new cases on December 1

And the implications for hospitalizations and deaths is awful

But a few things can alter this:

3/4
Read 6 tweets

Did Thread Reader help you today?

Support us! We are indie developers!


This site is made by just two indie developers on a laptop doing marketing, support and development! Read more about the story.

Become a Premium Member ($3/month or $30/year) and get exclusive features!

Become Premium

Too expensive? Make a small donation by buying us coffee ($5) or help with server cost ($10)

Donate via Paypal Become our Patreon

Thank you for your support!

Follow Us on Twitter!