Read on Twitter
I hope so but I don't think so. Who thinks they have a shareable expert opinion on Sweden? https://www.cnbc.com/2020/04/22/no-lockdown-in-sweden-but-stockholm-could-see-herd-immunity-in-weeks.html
Back in Sweden for a moment, estimating population immunity. First, its an open question whether infection, survival and antibodies really do provide immunity. That's a heck of a bet, but let's assume/hope that it does.
How might one estimate disease spread? 1/...
How might one estimate disease spread? 1/...
Let's take a VERY simple model. This was used in the Red Dawn emails but its basically common sense with heroic assumptions.
SUPPOSE the disease follows a predictable course. A person is infected at Day 0. The disease incubates, gets worse (or not), the person lives or dies. 2/
SUPPOSE the disease follows a predictable course. A person is infected at Day 0. The disease incubates, gets worse (or not), the person lives or dies. 2/
IF they die, let's assume/estimate its at Day 21. If they live, assume detectable antibodies at Day 28.
And will they live? We need an Infection Fatality Rate, or IFR. We don't know that. Of course, lacking widespread testing we don't infections at Day 0 or antibodies at 28. 3/
And will they live? We need an Infection Fatality Rate, or IFR. We don't know that. Of course, lacking widespread testing we don't infections at Day 0 or antibodies at 28. 3/
Rather than catalog our ignorance (I have free time, but not THAT much), what DO we know? Well, deaths, sort of - reporting issues are well documented but it may be our best handhold.
So work backwards. If we know Deaths on 4/22, by assumption they were infected on 4/1. 4/
So work backwards. If we know Deaths on 4/22, by assumption they were infected on 4/1. 4/
Add another Heroic Assumption: Use an IFR of 0.5%. That's five times as deadly as the regular flu, so its a scary number. 1 out of 200 that are infected go on to die.
So if 100 people died on 4/21 they were infected on 4/1 along with 199 x 100 (19,900) who go on to live. 5/
So if 100 people died on 4/21 they were infected on 4/1 along with 199 x 100 (19,900) who go on to live. 5/
But don't test those 19,900 for antibodies yet! They won't test positive for that until 4/28. Weird, and an assumption researchers will be checking, obvi, when the real modelers crunch this sort of thing.
The gist: knowing deaths by day and total deaths, we can estimate dates 6/
The gist: knowing deaths by day and total deaths, we can estimate dates 6/
of infection and antibody presence. Put those deaths within a population and we can estimate population immunity.
For example, I worked that out for the NY Metro area (NYC and the nearby counties), which account for 95% of the deaths in NY State. Lots of numbers, here goes: 7/
For example, I worked that out for the NY Metro area (NYC and the nearby counties), which account for 95% of the deaths in NY State. Lots of numbers, here goes: 7/
And the Too Many Numbers chart, which I've snipped to save space and probably create confusion.
And what does it mean? From memory here, the controversial Stanford Santa Clara County study estimated an IFR of 0.25% and about 3% population with antibodies. People applauded the low IFR but I was worried about the low spread - it means COVID-19 has room to run in CA.
With an assumed IFR of 0.25% and bodies stacked like cordwood, NYC Metro would have about 34% showing antibodies today. Hard to believe Stockholm is at that level.
From the chart using 0.5%, antibodies on 4/21 are based on deaths on 4/14, so 17.1% in NY Metro should test pos.
From the chart using 0.5%, antibodies on 4/21 are based on deaths on 4/14, so 17.1% in NY Metro should test pos.
So just at a glance I'm surprised Stockholm's body count is high enough for them to be talking about herd immunity. Sweden pop. is 10 MM. Stockholm County pop. is 2.3 MM. 1,765 deaths (let's put them all there).
4/21 Stock. Deaths/MM ==>> 765. NY Metro==>> 1178.
Well, hmm...
4/21 Stock. Deaths/MM ==>> 765. NY Metro==>> 1178.
Well, hmm...
