Coronavirus

[Daniel in SD]

Looks like the Abbott IgG antibody test is extremely accurate. Hopefully all future serological surveys will use it instead of questionable Chinese lateral flow tests.

Coronavirus disease-19 (COVID19), the novel respiratory illness caused by severe acute respiratory syndrome coronavirus 2 (SARS-CoV-2), is associated with severe morbidity and mortality. The rollout of diagnostic testing in the United States was slow, leading to numerous cases that were not tested for SARS-CoV-2 in February and March 2020, necessitating the use of serological testing to determine past infections. Here, we evaluated the Abbott SARS-CoV-2 IgG test for detection of anti-SARS-CoV-2 IgG antibodies by testing 3 distinct patient populations. We tested 1,020 serum specimens collected prior to SARS-CoV-2 circulation in the United States and found one false positive, indicating a specificity of 99.90%. We tested 125 patients who tested RT-PCR positive for SARS-CoV-2 for which 689 excess serum specimens were available and found sensitivity reached 100% at day 17 after symptom onset and day 13 after PCR positivity. Alternative index value thresholds for positivity resulted in 100% sensitivity and 100% specificity. We then tested 4,856 individuals from Boise, Idaho collected over one week in April 2020 as part of the Crush the Curve initiative and detected 87 positives for a positivity rate of 1.79%. These data demonstrate excellent analytical performance of the Abbott SARS-CoV-2 IgG test as well as the limited circulation of the virus on the West Coast. We expect the availability of high-quality serological testing will be a key tool in the fight against SARS-CoV-2.

https://www.medrxiv.org/content/10.1101/2020.04.27.20082362v1.full.pdf
One thing to note for the IFR obsessed is how soon after symptoms the test is able to detect antibodies.

 

[Norbert]

Looks like the Abbott IgG antibody test is extremely accurate. Hopefully all future serological surveys will use it instead of questionable Chinese lateral flow tests.

https://www.medrxiv.org/content/10.1101/2020.04.27.20082362v1.full.pdf
One thing to note for the IFR obsessed is how soon after symptoms the test is able to detect antibodies.
View attachment 538249

The graph shows only PCR, right? Or does IgG sensitivity coincide with symptom onset? From the text I read 53.1% at 7 days, 82.4% at 10 days, 96.9% at 14 days, and 100% at day 17.
(with a larger-seeming confidence interval). Is that test already used somewhere?

The sensitivity of the assay from the estimated day of symptom onset for the 125 patients included in our chart-review study was 53.1% (95%CI 39.4%-66.3%) at 7 days, 82.4% (51.0-76.4%) at 10 days, 96.9% (89.5-99.5%) at 14 days, and 100% (95.1%-100%) at day 17 using the manufacturer’s recommended cutoff of 1.4.

EDIT: No, IgG sensitivity does not coincide with symptom onset: The above times are _after_ symptom onset.
EDIT 2: It seems "symptom onset" in the graph is how long after symptom onset the IgG test becomes sensitive. Not the symptom onset itself.

 

[EinSV]

Looks like the Abbott IgG antibody test is extremely accurate. Hopefully all future serological surveys will use it instead of questionable Chinese lateral flow tests.

https://www.medrxiv.org/content/10.1101/2020.04.27.20082362v1.full.pdf
One thing to note for the IFR obsessed is how soon after symptoms the test is able to detect antibodies.
View attachment 538249

Wow, that's really impressive data -- even better than advertised.

99.9% specificity(!)

Plus 17 days after symptom onset -- or about 22 days after infection -- 100% sensitivity (95.1%-100% at 95% confidence interval).

If this holds up, Abbott hit it out of the park and it will help reduce some uncertainty around these tests.

Side note: redditors are calculating an IFR of 0.17% for the region tested -- similar to the Santa Clara County results from the Stanford researchers. Performance Characteristics of the Abbott Architect SARS-CoV-2 IgG Assay and Seroprevalence Testing in Idaho : COVID19

 

[dfwatt]

A point in time (today) in the Death sweepstakes.
The USA (orange bar) is climbing the ranks quickly, and with the help of the Southern states has high aspirations.

View attachment 538215

After all, didn't Donald J Clorox promise that once he got elected we'd be so tired of winning. We're number one. Well not yet, in total deaths yes, not in deaths per capita. Give him time. He and the other morons that both work for him and that follow him will eventually get us that incredibly dubious title. I'd love to be proven wrong on this, but I see nothing to be confident about in terms of any likely mitigating factors that will prevent that outcome.

 

[dhanson865]

All the stores around here have plexiglass partitions screening you from the cashier and bagger (those people are also wearing masks and gloves). And why should I deprive someone of the livelihood by using the self checkout? The store doesn't pay me to be a checkout person, so why should I work for them for free?

why should I stand in line for 20-30 minutes to have someone slowly check me out (taking another 10-20 minutes) when I can go through a self checkout in 5-10 minutes?

I value my time more than you I guess. Or maybe I'm just more physically fit and don't mind moving my own groceries around.

 

[EVNow]

why should I stand in line for 20-30 minutes to have someone slowly check me out (taking another 10-20 minutes) when I can go through a self checkout in 5-10 minutes?

I value my time more than you I guess. Or maybe I'm just more physically fit and don't mind moving my own groceries around.

Answer to both questions is pickup.

 

[powertoold]

 

[dhanson865]

Answer to both questions is pickup.

An eye doctor once told me the progression from better to worse was laser surgery > contacts > glasses (specifically in terms of clarity and field of view in the ideal case).

in the same vein you could say home delivery > store pickup > self checkout > cashier/bagger

There are some things I get from each of those methods but I avoid the line and direct interaction as much as possible.

The person I was responding to seems to prefer to stand in the line and deal with 2 or more people for the single transaction.

 

[TespaceX]

Looks like the Abbott IgG antibody test is extremely accurate. Hopefully all future serological surveys will use it instead of questionable Chinese lateral flow tests.

https://www.medrxiv.org/content/10.1101/2020.04.27.20082362v1.full.pdf
One thing to note for the IFR obsessed is how soon after symptoms the test is able to detect antibodies.
View attachment 538249

This is good news.

Would have been even better if the test was able to give a quantitative titer result. We have ideas, but not much concrete information about what having antibodies actually mean at this time. Having that quantitative information could help translate the result into daily behavior modification, once we know more about the virus.

 

[Norbert]

Apparently I am being asked to update my IFR calculations.

The May 2 New York City update for deaths is 13,156 + 5,126 = 18,282 for 8.4 million population.
COVID-19: Data - NYC Health

That is a mortality (deaths/population) of almost 0.22%. (Mortality is not IFR)

The antibody infection level update for NYC today was 19.9%. That results in a "naive" IFR of almost 1.1%.
"Naive" means without adjustments.

Personally I'd make 2 or 3 adjustments that I know of:

1) Like previous numbers for infection level, this is likely from people who were outside, and may have because of exposure, on average a higher infection level than those who stay home (although some stay home because the feel they might be sick. But not me, for example).
Previously I deducted 3%, this time I just want to take off 1.5%. Assuming an 18.4% infection level results in an IFR estimate of 1.18%.

2) Time lag from positive test to death. The current death number does not correspond to today's infection level, but to that from some time ago. I've previously estimated that this time is on average (not median) about 14 days. That was for PCR tests. Daniel's post above suggests that Abbott's antibody test responds about 4 days later than the PCR test, in so far as I understand. Although the NYC test is likely not Abbott's test, I'll just use that number, and take off another 3 days to be more cautious about the current decreasing dynamic. That leaves a lag of 7 days. In the last 7 days the number of cases was about 13,692 considering that the last few days have incomplete data (I took the sum of day -1 to day -7). The current number of cases in NYC is 166,883, so that's an adjustment factor of about 1.09x, resulting in, surprise, an estimated IFR of about 1.29%. Just to show that I am consistent with my previous estimates, which were also about 1.29%.

3) Most sources that I notice (but am not keeping track of) assume an under counting of deaths based on above-average deaths compared to the dynamic of previous years. However these numbers are not getting updated, AFAIK, so I am just leaving that as a reserve argument.

 

[AZRI11]

Listen to Jack "not" Chris Cuomo

 

[Daniel in SD]

Side note: redditors are calculating an IFR of 0.17% for the region tested -- similar to the Santa Clara County results from the Stanford researchers. Performance Characteristics of the Abbott Architect SARS-CoV-2 IgG Assay and Seroprevalence Testing in Idaho : COVID19

I’m pretty sure that sample was people willing to pay $100...
Crush the Curve | Coronavirus (COVID-19) Testing in Idaho

 

[NicoV]

Since you've probably got 8-10% of the population infected, it's possible you're seeing some effects of herd immunity in this decay

I doubt we have 8-10% herd immunity. Here’s a map of infections per community: UPDATE. Bekijk het aantal coronabesmettingen in uw gemeente
Where I live, we have 2.7 confirmed infections per thousand. We ‘d need 30x unconfirmed infections to reach that herd immunity number. Our neighbour city has 8.28. The peak communities have 18 infections per thousand. Those regions may have 10% herd immunity, but most regions are like mine and probably have minimal herd immunity. Note that the high infection rate in our neighbour city is most likely caused by a giant outbreak in a nursing home (the biggest one in the region, it has it’s population decimated by Covid-19). The infection rate in the general population in that city is probably a lot less. We have systematically tested all personnel and residents in nursing homes (see section 2.2 of https://covid-19.sciensano.be/sites/default/files/Covid19/Meest recente update.pdf) and for entire Belgium, 3% of personnel (83K people) is infected. That 3% is probably representative of the entire population, and also corresponds to the outcome of antibody tests on random blood samples.
IIRC as of next week more systematic antibody testing will start, so we’ll probably know more about herd immunity levels in a week or 2.

 

[cwerdna]

We need widespread active virus tests and widespread antibody tests. Both of which are accurate. We need to cull the counterfeit tests from the market ASAP. This relies on competent governing top to bottom, which we don't have right now. If we had an idea of how many people had gotten over the virus and how many were currently infected, leaders could make educated decisions about opening up the economy. Right now most leaders in the US are blind to how many active infections are out there and how many have likely immunity.

There are issues with false positives in some (many? most?) antibody tests in populations w/low overall infection rates.

The below was written by three MDs.

Beware of Antibody-based COVID-19 ‘Immunity Passports’

Accurate tests do, however, exist. The first of four antibody tests that was approved on an emergency basis by the FDA, one manufactured by Cellex, boasts a sensitivity of 94 percent (the proportion of people with the disease who will test positive) and a specificity of 96 percent (the proportion of people without the disease who will test negative). Those numbers might seem impressive, but, as former defense secretary Robert McNamara once said, in the fog of war, “belief and seeing are both often wrong.”
...
Over 860,000 Americans have been diagnosed with COVID-19, roughly 0.25 percent of the population. The true total is surely higher, possibly around 1 to 2 percent, because most infections are asymptomatic (50 percent in Iceland) or undetected (86 percent in China). Random population sampling has indicated various numbers for prevalence in different areas, including 0.5 percent in San Miguel County, Colorado, 0.8 percent in Iceland[OO1] , 1.8 percent in South Korea and 2.8 percent in Santa Clara County, Calif.

What would happen if we used the FDA-approved Cellex test to issue COVID-19 passports in a hypothetical society with 10,000 people where the prevalence is 1 percent? If all 100 people who had COVID-19 also had antibodies, passports would be correctly issued to 94 (94 percent of 100) of them who test positive. Of the 9,900 uninfected, 9,504 (96 percent of 9,900) would correctly test negative. However, the remaining 396 (4 percent of 9,900) would falsely test positive and get passports incorrectly. Thus, 396 out of the 490 (94 plus 396) passports issued (four out of every five) would be illegitimate!

Even if the prevalence were 2 percent, two out of three passports would be issued to people who didn’t have antibodies. This false positive problem is what bedevils mass population screening when the prevalence is low.

 

[AlanSubie4Life]

I doubt we have 8-10% herd immunity

I based this on your number of deaths, 8000, which suggests about 900k to 1 million cases to date. (Deaths would go to 9-10k and IFR will go to about 1% eventually, if no further infections occurred.).

And also the model I linked to uses the same numbers.

It’s possible IFR is higher for some reason in Belgium. And that would mean somewhat fewer cases but I can’t imagine fewer than 600k.

Population is 11 million...so.

It is possible that Belgium has done a really good job of counting deaths relative to other countries, which would also mean a lower number of actual infections.

 

[NicoV]

So, I’ve been trying to follow the developments of the San Francisco public tracing program.
The way they are going about it is:

1. Infected person is interviewed by a tracer and asked for the identity/phone number of everyone who the patient came in close contact with.
2. The tracer then text messages all these exposed contacts with a number to call for further follow-up to inform them of next steps. (The identity of the infected patient is never disclosed, but obviously can be deduced in certain circumstances).
3. The exposed individuals are asked to self-quarantine for 2 weeks, and a follow-up call is made at 7 and 14 days to see whether any symptoms develop.
4. 2nd degree of contacts (by the exposed individuals) are not contacted unless an exposed individual develops symptoms and tests positive themselves

One worker in the program said that trying to reach out and speak with exposed individuals to explain their situation can take an entire day for a single case. One tracer said they’ve had some patients with just a couple contacts to call, and some with 30 contacts to call.

That’s similar to what Belgium is planning to do starting from 15/5, except that all contacts will be asked to go take a test too.

 

[NicoV]

It is possible that Belgium has done a really good job of counting deaths relative to other countries, which would also mean a lower number of actual infections.

Yes we have. There was a lot of critique because in the initial phase we included suspect deaths from nursing homes, and that was supposedly too pessimistic. But now that we have tested the entire nursing home population, it turns out that 80% of the deaths in nursing homes are confirmed infections. From those 8000 deaths, over 4000 are from nursing homes. Only 16% of those are confirmed cases, but that’s because we didn’t test those cases early on in the epidemic. With the same way to evaluate the death cause, the last 24 hour we had 80% confirmed deaths, so it is reasonable to assume that we had the same percentage when we didn’t test as much.
I think when all is said and done, Belgiums death rate will have turned out to be accurate, and other country’s death rate will have turned out too optimistic

 

[renim]

Let's put some numbers to your conjecture...
---
My opinion:
Since I anticipate mass vaccination in 12 - 18 months I reject the notion of natural vaccination given the attendant increase in morbidity and mortality.

Why a coronavirus vaccine could take way longer than a year

'The mumps vaccine—considered the fastest ever approved—took four years to go from collecting viral samples to licensing a drug in 1967. Clinical trials come with three phases, and the first stages of the current COVID-19 trials aren’t due for completion until this fall, spring 2021, or much later. And there are good reasons to allow time for safety checks. Some preliminary vaccines for the related coronavirus SARS, for instance, actually enhanced the disease in model experiments.'

There are different levels of confidence for differing vaccine scenarios, but at baseline it should be considered as an open ended question with low confidence either way.

 

[Singer3000]

Lots of excitement about contact tracing.
Singapore had one of the most advanced contact tracing models in the world. It also had since the start of Feb onwards quite strict social distancing measures, work from home requirements, a culture fully at ease with masks getting centrally distributed masks (use is now legally mandated). If you are contact traced as exposed and test positive, you are put in a centralised government quarantine facility (my friend was for 30 days despite being largely asymptotic throughout). Borders progressively restricted to the point of being closed. Only one adult per household allowed out at any time. Criminal penalties for non compliance with orders etc...

And yet it’s R0 still slipped over 1 requiring a hard lockdown.

Now while there are idiosyncratic domestic reasons why this might have happened (guest worker dorms), how hard have people thought about what this all means? Even with restrictions that many see as incompatible with western norms (e.g. forced central quarantine for even asymptomatic patients), they still needed a hard lockdown that’s so far scheduled to last 2 months. Far looser restrictions have been called “fascist” by Musk.

For the market bulls, what do they expect the world to look like in 6 months I wonder?

 

[Economite]

Why do you think the employees would tolerate this sort of employer-imposed isolation? How much extra do you think Tesla will pay them? Or is your notion that people will be given a choice between (I) leting Tesla impose extreme isolation on them outside of work hours or (II) losing their jobs and unemployment benefits?

I disagree. The value of Bluetooth contact tracing I think is a bit overstated. It's probably helpful, when combined with public health authorities standard methods of contact tracing (don't really see a downside to implementing Bluetooth contact tracing as long as it doesn't create extra work that is a waste of time).

But I don't think Tesla's going to have the expertise to monitor and control contacts in a factory environment. I could be wrong. I guess they could make "group" restrooms (employees are only allowed to use certain bathrooms), so that employees within the factory are sequestered in contact-free groups. That would reduce spread & risk of everyone getting infected. Not sure the value though - since if there is an exposure, then the whole factory shuts down since that group would have to be removed. To be clear, contact tracing is something that SHOULD be done - I'm not opposed. I'm just not sure it's the first thing that should be done, as it's kind of a "close the barn door after the horse has flown" solution.

As far as testing capacity - that's a reason to wait. Testing capacity is increasing. Do the surveys. Force strict mandatory quarantines on employees NOW who are low risk. For employees in higher risk categories, find some instant testing capacity and go get it. Get those employees tested.

There are things that can be done to at least TRY to make sure there are no infectious people at the factory. They should at least try.



As mentioned above - not necessarily. It needs to be communicated to employees that there is a zero tolerance policy for infection, and institute support measures for the employees that allow them to meet that goal (as outlined above). For some workers, it might be quite possible to have them operating in a "bubble." Home->work->home. People who live with those workers might be able to be classified as low risk (as long as they comply with mandatory quarantine as well). But going outside of that bubble - there can't be any exceptions. Anyone who goes outside the rules needs to be able to freely admit that (get a pass on coming to work for a week or whatever with paid leave with no impact on paid leave balance - so incentivize admission of violating the quarantine rules), followed by an instant test after a few days (to minimize chance of a false negative).

At this point, it should be pretty easy to assess all the exposure pathways - employees shouldn't be going to the grocery store, going to parks, beaches, going to get gas, going to Home Depot, etc. Any activity where they would normally put on a mask, they should not be allowed to do. Nothing other than Home->Work->Home. And Tesla should have a delivery service which services all their needs at home.

Those that don't fall neatly into these categories - give them a couple months off, or continue their furlough for now (they are receiving all the benefits of unemployment and the government supplementals through July as I understand it). Or give them the option of living at the Fremont factory.

Again, this is not meant to be a coherent plan - I don't work for Tesla. I'm just spitballing and presenting a few elements of a plan that actually tries to meet the objective of zero infectious workers at the Tesla factory. Which I think has to be the goal.

Coronavirus has Elon Musk acting like just another used car salesman


@EinSV - the Stanford and LA folks should have used a good old-fashioned American antibody test, rather than that Chinese junk:

Greninger Lab on Twitter

I'd like to see a repeat of their study with this test. Caveat still applies, though - I'd like to see the test repeatedly verified on OTHER groups of blood samples - it's entirely possible that certain populations will yield extraordinary sensitivity and specificity results, while other groups might perform poorly. So checking more populations and making sure none of this behavior exists is a valuable check. (They found about 1.6% prevalence in Idaho from their sample, BTW. But they have limited data on what that sample represented, exactly. Looks like Idaho has 0.6% overall prevalence from models, so seems like it's in the ballpark depending on where the sample was taken from.)