totm april 2020 Coronavirus

[tater]

There was a guy on space twitter who posted that he had just donated plasma as a COVID-19 survivor to help other patients (convalescent plasma therapy). He was sick very early January, and had traveled in December (when he thinks he got it). I'll look for the tweet. This has been around a while.

It;s like what my wife said when the WA cases were annouced, if they got it in the nursing home in mid Jan, it was community spread, and it must have been all over WA already, maybe even in the fall.

 

From the continuation of his linkedin post:

Quote

Lucky to be able to donate my plasma with Covid-19 antibodies to help UCLA treat folks. I was infected with the virus, my guess is on an airplane to the Bay Area in December and came down with the illness on January 4. It was basically a really bad flu that kept me up coughing all night for a month. I have been back to 100% since mid February. I knew it was Covid as soon as I saw the news reports, so when UCLA asked two weeks ago to sample my blood to see if I had the antibodies, I said yes and sure enough I tested positive for antibodies from more than three weeks in the past.

 

[sevenperforce]

2 minutes ago, tater said:

I suppose one possibility is that the spike in deaths occurs once you hit a critical mass of infections such that the most at risk from this virus get exposed, then it bumps them off, and once the most susceptible are gone, it continues as a more mild illness (gotta make sense of substantial breakouts like NY).

There's a paper on reproductive number up the thread that discusses very low R₀s with high variability being possible—generally low transmissibility, but with a subset of super-spreaders. Wonder if this could be a thing?

One of the things we have to make sense of is the number of healthcare workers who are getting infected and dying, including those in low-risk groups.

A possibility is genetic susceptibility to infection alongside genetic resistance to the disease's pathology. If anyone should have gotten it outside the healthcare industry, it would have been me -- I am in one of the most internationally-trafficked cities in the world and I was riding the metro daily until it shut down. I've had tightness in my chest and an occasional dry cough for weeks. Yet if I had it, there's a 100% chance I would give it to my wife and kids, none of whom have shown any signs of illness. If I had to wager, I'd say all of us have it (or have had it) and are simply asymptomatic or nearly so.

Here in the US, DJT Junior was panned in 2016 for comparing Syrian refugees to poisonous Skittles, but this is a situation where the metaphor is perhaps more apt. The elderly and those with pre-existing conditions are at high risk, but even among the low-risk groups there is evidently a significant portion of the population with an elevated risk of severe health outcomes. It doesn't matter how young or healthy you are, or whether you are a healthcare worker, or whether you take all kinds of precautions; one of the Skittles in this (relatively small) bowl of Skittles will put you in the ICU, and have no way of knowing which one. And that's one of the terrifying things here.

(statistically, "you have no idea which of these hundred Skittles will kill you" is equivalent to "you have no idea whether you are one of the 1% of the population who is deathly allergic to Skittles")

[tater]

18 minutes ago, sevenperforce said:

One of the things we have to make sense of is the number of healthcare workers who are getting infected and dying, including those in low-risk groups.

Viral infections are usually partially a function of dose. HCW get repeated doses, and sometimes large doses.

 

Quote

A possibility is genetic susceptibility to infection alongside genetic resistance to the disease's pathology. If anyone should have gotten it outside the healthcare industry, it would have been me -- I am in one of the most internationally-trafficked cities in the world and I was riding the metro daily until it shut down. I've had tightness in my chest and an occasional dry cough for weeks. Yet if I had it, there's a 100% chance I would give it to my wife and kids, none of whom have shown any signs of illness. If I had to wager, I'd say all of us have it (or have had it) and are simply asymptomatic or nearly so.

This is not strictly true, the attack rate for people in the same household is apparently on the order of 10-15% (least according to that Taiwanese paper I read and I think linked someplace). Course I am unsure on the time horizon on that, was it 10% every 2 weeks?

My wife had some bad crud in Feb, and lost her sense of smell/taste for a few days. Her office-mate was so sick he didn't come in for a week (spiked a high fever, etc). Everyone here in our house had some sort of crud, but not bad (my daughter the least cruddy of us). We all got flu vaccine.

 

Quote

Here in the US, DJT Junior was panned in 2016 for comparing Syrian refugees to poisonous Skittles, but this is a situation where the metaphor is perhaps more apt. The elderly and those with pre-existing conditions are at high risk, but even among the low-risk groups there is evidently a significant portion of the population with an elevated risk of severe health outcomes. It doesn't matter how young or healthy you are, or whether you are a healthcare worker, or whether you take all kinds of precautions; one of the Skittles in this (relatively small) bowl of Skittles will put you in the ICU, and have no way of knowing which one. And that's one of the terrifying things here.

The % of people with really bad outcomes who are not old or obese is really tiny. Only 0.8% of the deaths (not % of confirmed cases, % of deaths) have no preexisting conditions. Young, healthy people also mysteriously die of flu. This might get more bad outcomes with COVID than flu, but google "20 year old dies of flu" and see what you get. It happens, and it seems so random to see perfectly healthy 20 YOs dead from flu. And they get the kitchen sink thrown at them, too, docs really don't like losing 20 YOs.

 

Quote

(statistically, "you have no idea which of these hundred Skittles will kill you" is equivalent to "you have no idea whether you are one of the 1% of the population who is deathly allergic to Skittles")

The risk is very low. For the 21k deaths CDC has sorted so far, 2.9% are under 45, and there is no column for preexisting conditions (and from other data like NY I have seen, it's the huge majority of cases). That's % of deaths, not % of cases. That's 0.07% of confirmed cases in the US., so likely at most 0.007% of actual cases (possibly lower). So realistically, ~7:100,000 pop at large (possibly half that depending on what % has actually had this). (a little more than half the chance of death in a car accident).

 

Edited April 22, 2020 by tater

[tater]

The serological studies are getting attacked, but it's odd that we'd have something SUPER contagious, yet only have a tiny% of the population infected, particularly after being in the country for months. What surprises me about the studies that show 3%, 4%, etc is how LOW they are, not how high they are (Chelsea's 32% makes way more sense to me).

R₀ is a trainwreck. It's a very useful number to know—but no one knows it accurately, not for any pathogen. I keep seeing people (epidemiology people) now saying it's 3-5, but we had it unmitigated in the US since at least late December until March. For higher R₀, doubling time is 2-3 days (chart I saw went up to 2.5 (found it!)).

1 case late Dec, double every 3 days til mid March... that's 10% of the pop of the country by the time the shutdowns even started anywhere in the US. If it doubled in 2 days... everyone on Earth had it by then.

Seems like they could work backwards from cases estimated in mid March, and get R0 from that. US had ~1200 confirmed cases on March 11. If there were actually 100 times that number, ~120,000 cases, then that means the doubling time is on the order of 5 days from first case in late Dec. That's an R0 of ~1.4-1.5?:

That would be herd immunity at 1/3 the pop.

If the confirmed cases is actually more accurate than catching only 1%, R0 must be even smaller. 10% caught? Closer to 6 day doubling.

 

 

 

Edited April 22, 2020 by tater

[sevenperforce]

12 minutes ago, tater said:

This is not strictly true, the attack rate for people in the same household is apparently on the order of 10-15% (least according to that Taiwanese paper I read and I think linked someplace). Course I am unsure on the time horizon on that, was it 10% every 2 weeks?

I have several very young kids in close quarters so I suspect we are at an exponentially higher risk for intra-household transmission.

13 minutes ago, tater said:

Viral infections are usually partially a function of dose. HCW get repeated doses, and sometimes large doses.

Biggest factor IIRC is the size of the initial dose, because that impacts antibody production. Not sure which way this goes, tho.

18 minutes ago, tater said:

The risk is very low. For the 21k deaths CDC has sorted so far, 2.9% are under 45, and there is no column for preexisting conditions (and from other data like NY I have seen, it's the huge majority of cases).

Some data from roughly a week ago in NYC may prove instructive.

Look at the third column: cases per 10,000 for age group. Let's assume, for the sake of this analysis, that there's no age-dependence on susceptibility to infection, only on susceptibility to pathology. Let's also assume that the disease is infectious enough that all age groups have a roughly equal chance of exposure. Neither of those are strictly true, but there's a fair chance that they cancel each other out, Fermi-wise (those more vulnerable to infection are less likely to be exposed, all other things being equal).

These assumptions suggest that total cases are roughly equal across age groups and so the third column actually reflects the percentage of symptomatic cases. If the actual number of cases is at least 210.4 per 10k across all age groups, then you recalculate the estimated number of cases and estimated hospitalization rate. Moreover, death rate as it stands is rather useless; what you want is the death rate per actual hospitalization:

NYC COVID-19 Cases

AGE	EST. CASES (min)	SYMPTOMATIC	HOSPITALIZATIONS	EST. HOSP RATE	DEATHS	DEATHS/HOSPS
0-17	36,818	5.5%	190	0.52%	3	1.58%
18-44	70,648	56.2%	4,304	6.09%	284	6.60%
45-64	43,258	87.5%	10,182	23.54%	1449	14.23%
65-74	14,717	89.2%	6,147	41.77%	1511	24.58%
75+	11,477	100.0%	6,853	59.71%	2935	42.83%

 

Taken in aggregate, this suggests 176,918 cases and 104,158 symptomatic cases, for a symptomaticity rate of 59%. That number will drop with serological testing, but it is reasonably close to the 48.3% symptomaticity found on the Diamond Princess. If the true symptomaticity rate is 48.3%, then the actual odds of COVID-19 putting you in the hospital in the 18-64 aggregated age group (i.e., most American workers) is 10.4%, and your odds of death from hospitalization is 11.9%. That's a pretty damn small bowl of Skittles.

[tater]

2 minutes ago, sevenperforce said:

I have several very young kids in close quarters so I suspect we are at an exponentially higher risk for intra-household transmission.

Maybe, but the contact tracing in Taiwan was families, too. Dunno.

 

2 minutes ago, sevenperforce said:

Some data from roughly a week ago in NYC may prove instructive.

Look at the third column: cases per 10,000 for age group. Let's assume, for the sake of this analysis, that there's no age-dependence on susceptibility to infection, only on susceptibility to pathology. Let's also assume that the disease is infectious enough that all age groups have a roughly equal chance of exposure. Neither of those are strictly true, but there's a fair chance that they cancel each other out, Fermi-wise (those more vulnerable to infection are less likely to be exposed, all other things being equal).

How do they get a death rate of 0.15% for 3 deaths out of 10,000 in the age group? Oh, it's 3 out of the number tested. So divide that number by 10? 100? Pick a number. Ditto the hospitalization rate. Divide by at least 10. For NY it's important to remember that ED docs I have read flat out said they didn't test anyone they were not admitting after a while. They'd say, "You have COVID, go home and isolate, come back if you are short of breath."

 

2 minutes ago, sevenperforce said:

These assumptions suggest that total cases are roughly equal across age groups and so the third column actually reflects the percentage of symptomatic cases. If the actual number of cases is at least 210.4 per 10k across all age groups, then you recalculate the estimated number of cases and estimated hospitalization rate. Moreover, death rate as it stands is rather useless; what you want is the death rate per actual hospitalization:

NYC COVID-19 Cases

AGE	EST. CASES (min)	SYMPTOMATIC	HOSPITALIZATIONS	EST. HOSP RATE	DEATHS	DEATHS/HOSPS
0-17	36,818	5.5%	190	0.52%	3	1.58%
18-44	70,648	56.2%	4,304	6.09%	284	6.60%
45-64	43,258	87.5%	10,182	23.54%	1449	14.23%
65-74	14,717	89.2%	6,147	41.77%	1511	24.58%
75+	11,477	100.0%	6,853	59.71%	2935	42.83%

 

Taken in aggregate, this suggests 176,918 cases and 104,158 symptomatic cases, for a symptomaticity rate of 59%. That number will drop with serological testing, but it is reasonably close to the 48.3% symptomaticity found on the Diamond Princess. If the true symptomaticity rate is 48.3%, then the actual odds of COVID-19 putting you in the hospital in the 18-64 aggregated age group (i.e., most American workers) is 10.4%, and your odds of death from hospitalization is 11.9%. That's a pretty damn small bowl of Skittles.

Diamond Princess was old people. More than half were over 60. Those 60+ people were 75% of those who tested positive. Anyway, it was a Petri dish, hardly normal living in the world levels of contact.

All the serological testing is showing vastly more than 50% of cases not seeking care/testing (doesn't mean asymptomatic, just means they thought they had a cold, allergies, whatever and ignored it as we all do every other year). Your chance of hospitalization is nothing like 10%, I bet it ends up closer to 2% per infection (dominated by the elderly, lower for people younger). Seasonal flu has symptomicity of ~25% (posted that UK NHS paper about the rate that people even report flu as a function of their antibody testing up the thread that showed 3/4 asymptomatic).

We lack good data.

That said, look at what I was saying about R₀. There were no meaningful closures before ~March 12 in the US. There was certainly at least 1 case in the US in December. How can we have so few cases by mid March? Tested cases at 100% ascertainment would make R₀, extremely low. If we only detected 10% of cases, R₀ is also low, in fact about like flu. If testing caught 50%? Lower than flu.

So if the doubling time is longer that people thought (lower R₀), the hospitalization/fatality is worse (closer to actual CFR numbers you posted above), but herd immunity is also closer to 30% (lower the higher the % of cases that actually got caught). 23% for R₀=1.3, 28.5% for R₀=1.4, 33% for R₀=1.5). So if the risks are closer to what you think, then we're far closer to herd immunity (which is a plus).

Bottom line is that I still think a vaccine is wishful thinking (fingers crossed, obviously), so we need to hit herd immunity one way or another. The other thing is that if immunity to this is not long term durable, then we not only need herd immunity, we need to have it as quickly as possible without overwhelming healthcare. That way there is no wave of previously infected getting infected again in X years. There will be a basal rate of susceptible people getting it going forward since it's now endemic, but with herd immunity it can stay fairly contained, and outbreaks dealt with. If we flatten the curve over years, then we might end up with previously immune people becoming susceptible again (even if they are partially protected, they can still spread it, increasing R_e).

[tater]

If the total US cases were ~3% of the pop by mid March, the doubling time is ~3.7 days (R₀~1.7?), but this is all VERY sensitive to the date of the first import, obviously (since every 3.7 days is a doubling). If it's ~3% NOW, then in mid March (much slower doubling since lockdowns, call it 4 doublings in the last 6-7 weeks), then the doubling time was over 4 days (so R₀ close to 1.5).

Note I'm thinking of "real" R₀ here, to get herd immunity where "real" is functionally the reproductive number with people behaving voluntarily as they like (so some people knew about COVID before March, and changed behavior themselves, but no "shut downs"). Everything is really R effective, when it comes down to it.

Going forward (which is what matters, IMO), I think we need to have a good sense of this so we know what to do. A vaccine in a year is not impossible, but it seems pretty unlikely. Both from a safety standpoint (proper testing takes time), and just from the fact that coronaviruses have not had vaccines made (they've wanted some in veterinary medicine for a long time, no luck). So what is the minimum, long term change in infrastructure/behavior that can mitigate this and go on as long as it takes? Clearly "shut down" is not sustainable for years (or even months). A rational return to normalcy is needed.

Isolate the elderly, and have people wash their hands like civilized people already did? Closer attention to disinfecting at risk areas periodically? Sensible norms around being sick (stay home if sick, with incentives to encourage this rather than all those that already exist to discourage it)?

 

[tater]

Does that link to an actual paper? Vox is as bad a source as Fox, frankly. (Vox had been saying the flu was worse and SARS-CoV2 was not going to be a pandemic, then they later deleted all their tweets)

Here's what we need to get back to normal? (clearly OSHA would require an organic vapor filter mask for the chimp here, but the shirt is the sort of PPE I prefer, myself)

 

Edited April 22, 2020 by tater

[tater]

 

[tater]

^^^Seems like a good paper. They have IFR at ~0.5%, heavily skewed towards the elderly.

 

 

Edited April 22, 2020 by tater

[TheSaint]

[tater]

This is interesting when compared to the analysis in that previous paper about Italy, and the implications globally. Of that 0.5% average IFR, we could expect to see as many as 50% coming from elder care facilities.

[tater]

 

[kerbiloid]

In Sklifosofsky Institute they are trying the 92° hot helium to weaken the virus illness.
The patient inhales it for 15 minutes to restore the blood acidic balance.

https://translate.google.com/translate?hl=ru&sl=ru&tl=en&u=https://www.interfax.ru/moscow/705637

[KerikBalm]

That just sounds like bad science

Also, a word of caution on the seroprevalence studies, and the reports of asymptomatic rates.

Some excerpts from an e-mail chain from a COVID19 researcher google group:

Quote

I am quite confused by the wide range of estimates of the asymptomatic proportion. Does anyone know more studies than these? Does anyone have an explanation or possible hypotheses? Of course this variance can partly be explained by pre-/oligo-symptomatic cases, and at least the last two studies maybe partly by selection bias.

... goes on to cite some studies

a response

Quote

Thanks, ______ for assembling this set of papers. A range from 8-100% for ostensibly the same condition is certainly thought-provoking. I agree with ______ that it seems likely that many of the allegedly asymptomatic individuals in some of the studies would have become symptomatic later. To have any real validity, studies like these should have at least one follow-up examination, say 7-10 days after the first to check if symptoms subsequently developed.

and another response:

Quote

Thanks _______.

I raise the issue of the sensitivity and specificity of the various serology assays that were used in the below studies. That really seems to be infrequently considered when discussing these population estimates. There are others in the this forum who are the real lab experts who can comment on this with some authority. But as an epidemiologist, let me comment on test sensitivity and specificity as one aspect of this issue.

Let’s take a test like the Cellex serology assay that has stated the test has a sensitivity of  93.8% and specificity of 95.6% For the sake of calculation ease I’ll call it 95 and 95% respectively. And let’s assume that the population has an actually prevalence of 5% seropositivity for SARS-CoV-2 antibody (I know of no study that has estimated more than that for a population-based estimate other than what might been found post-outbreak for a closed population like a cruise ship, long-term facility or prison.) Even the infamous Santa Clara study which has been challenged found an upward bound estimate of 4.2%. In our theoretical 5% population positive and with the performance of the Cellex serology you can expect to find for every one thousand persons tested, 47.5 (i.e.48) test positive (i.e. 2.5, or 3 false negatives) and 47.5 (.i.e. 48) false positives. That means for every one person you actually find positive, one out two are false positives. If the actual prevalence is lower than 5%, the proportion of false positives goes up even more. This issue can surely account for some of the findings of asymptomatic persons positive for SARS-CoV-2

I fear that most everyone advocating widespread serology testing in low prevalence populations do not really understand this issue of using a screening test in low prevalence populations. Imagine telling a nurse, she or he is positive for SARS-CoV-2 antibody but in actuality you have a one in two chance that you really don’t have antibody. Also I believe few serology assays currently available for use have anyway near the sensitivity and specificity of the Cellex assay. For those other assays the numbers I just shared with you would be “even worse!”

FWIW, I now work in a hospital virology research group assisting them in writing papers and reviewing literature.

Edited April 23, 2020 by KerikBalm

[sevenperforce]

3 hours ago, KerikBalm said:

That just sounds like bad science

Also, a word of caution on the seroprevalence studies, and the reports of asymptomatic rates.

Some excerpts from an e-mail chain from a COVID19 researcher google group:

a response

and another response:

FWIW, I now work in a hospital virology research group assisting them in writing papers and reviewing literature.

We really are flying blind in so many ways with this.

The strong genetic dependence of hACE2 expression would provide a basis for widespread genetic resistance to SARS-CoV-2 pathology. Its role in hypertension, for example. Concentrations of genetically-restricted populations would also explain a wide range of tested asymptomaticity. 

[tater]

Yeah, serologic testing in low prevalence populations will give lousy results. Places like Bergamo (or NYC) might be another matter entirely. Or Chelsea, MA (small section of Boston, but large outbreak there).

[sevenperforce]

4 minutes ago, tater said:

Yeah, serologic testing in low prevalence populations will give lousy results. Places like Bergamo (or NYC) might be another matter entirely. Or Chelsea, MA (small section of Boston, but large outbreak there).

It's particularly bad when you don't have any good way to actually test the tests. 

[VoidCosmos]

I just want to say something to every doctor on Earth:

Thank you for risking your lives and loved ones to save us from this terrible Pandemic! Thanks a lot!! We really really really really really really appreciate this! Thank you doctors once again!

[tater]

19 minutes ago, sevenperforce said:

It's particularly bad when you don't have any good way to actually test the tests. 

The good studies test the tests vs known positive and known negative samples as far as I know to get the sensitivity and specificity.

[sevenperforce]

Just now, tater said:

The good studies test the tests vs known positive and known negative samples as far as I know to get the sensitivity and specificity.

I guess you can get true known positive and known negative by individual microscopic exam? 

[DDE]

14 minutes ago, VoidCosmos said:

I just want to say something to every doctor on Earth:

Thank you for risking your lives and loved ones to save us from this terrible Pandemic! Thanks a lot!! We really really really really really really appreciate this! Thank you doctors once again!

Or else

[Codraroll]

6 hours ago, kerbiloid said:

In Sklifosofsky Institute they are trying the 92° hot helium to weaken the virus illness.
The patient inhales it for 15 minutes to restore the blood acidic balance.

https://translate.google.com/translate?hl=ru&sl=ru&tl=en&u=https://www.interfax.ru/moscow/705637

Well, I would say that if a patient were to breathe an inert gas at close to boiling temperatures for a quarter of an hour, the chance that they will die of the coronavirus becomes extremely slim. 

[VoidCosmos]

Just now, DDE said:

Or else

Yeah

[tater]

2 minutes ago, sevenperforce said:

I guess you can get true known positive and known negative by individual microscopic exam? 

Blood sample from a known COVID patient vs a blood sample from so many months ago, no cases.