COVID-19 (artist formerly known as Wuhan strain novel Corona virus)

Just to clarify - are you speculating that a decent chunk of the population has some form of natural immunity to this thing due to past exposures?

That'd be neat.

 

I think the specificity of the assay is important, but IgG inside of 5 days is not abnormal.

Also, 1918 was only 33% give or take overall, so slowing at 25-30% seems right.

 

Depends, do you appreciate a standard curve?

 

Yeah, basically......some people - a decent amount of people - perhaps have an immune system that is much more ready to produce specific antibodies to this virus than they would be had their bodies never seen anything somewhat like it.

I'm reached out to my friend who, unlike me, actually is an immunologist to see if he sees any data to support it. But, something seems odd to me.

 

Shelby county

 

Little in the middle, but got much back.

 

Many of the estimates I had heard were 60-70% required for herd immunity, which really shifts things.

If you take an iFR of 0.26% (current CDC best-guess scenario) and a 33% herd immunity level, your US epidemic burns out at 300k deaths by next spring.

 

Way too high for herd immunity. I would look at other pandemics, when semi modern, and the flu seems like a pretty good one.

I would put immunity near 33%, same as 1918. The death rate will shift as we open more up. Our levels now are based on social distancing that is about to die.

 

What do you mean by death rate will shift - deaths per day or our understanding of iFR?

 

The best guess value will grow. Not to the point where deaths are 600k. But any value is now, at this moment, influenced by the percentage of kids out of school, and adults not working along side each other.

 

I'm not sure how much that number will grow. It is based on deaths and seroprevalence. The hard part is that we've been in a pretty steep slope on deaths/day, so error on where you want to call your cutoff between infected and developed antibodies will induce error (thus my questions about how fast can we expect to build immune response for the purposes of seroprevalence testing) in the total number. We need some really good, new antibody tests now that sero and total deaths are higher, which will decrease the error.

 

Serological is still lacking in specificity, I believe. I would be wary of it.

 

It depends on the test - there are some that are getting quite good, I think.

Roche's test is 100% sensitive and 99.8% specific across multiple antibody types. So 100% predictive on negative and 96.5% predictive on positive at a 5% seroprevalence. For places like New York that are closer to 30%, that becomes 99.5% predictive on positive. Abbott's test is pretty similar, but not quite as good at 99.6% specific. So, 92.9% predictive of positive cases at 5% seroprevalence, or 99% predictive of positive at 30% seroprevalence.

 

I doubt the vast majority of serological tests to date have used either of those.

Is there reason to believe that serological testing to date has used highly specific tests, since we lack a gold standard?

 

I agree that most of the testing has not used those, which is why I want to see a refreshed study.

 

I think we all want a refreshed study, as often as we can get it.

And maybe the values are high. But I think more likely the values are low, and we'll see an uptick.

I hope not, but that is my gut.

 

The large study in NY that showed 14% of the state was AB+ for CV used a test that had a sensitivity of 87.9% and a specificity of 99.75%. So, I'd say that 14% is fairly reliable. Converting that to an iFR results in about .3%; however, there is error because the days you go back to plant the flag for death count matters due to the timing of the study and the rate we were adding deaths at the time. Much more error from that than the actual AB test.

 

Question - IHME predictions spiked today
are they considered to be a reliable source?

 

I am seeing almost no one wearing masks anymore. I feel like I am some astronaut on a strange planet when I am out and in Weigels wearing mine.

 

It changes regularly as the underlying assumptions put into the model change. It reflects a sort of "if things stay like x, this is what one can expect." So every time a state opens up or changes a policy, the prediction will shift. E.g., it is very sensitive and its predictive power is thus limited since the state of knowledge on future conditions and behavior within populations is tied to the present without a precedent to fall back on.

JMHO based on my limited and armchair level understanding. Knoxville going from being a coronavirus ghost town to actually having a few clusters, for example, probably moves the needle up alone when plugging that into a projection with a given growth rate.