Stayin' Alive - Covid-19 #10

[Mudguard]

6 hours ago, Ran said:

This is probably a very generous assumption.  Imperial College's modelling suggests the number is much higher. In fact, there's the first results (Reddit thread because it provides a translation of the key facts) of antibody reporting in Denmark  is useful because it uses March 28th breakpoint that Imperial College did. At that time, there were 917 detected cases. Imperial College estimated that with  1.1% (95% credible level) of the population was infected at that point with a range from .40 to 3.1%. Denmark's antibody test suggests to them that on March 28th 3.5% were actually infected.

So more people were infected then even the most extreme of the estimate. The most extreme of the Italy and Spain estimates based on their models were 26% and 41% respectively, so... yeah. We've likely detected substantially less than anyone thinks.

Also saw the NY Times believes that NYC has had cases as far back as mid-February.

As I mentioned before, the modeling is very unreliable for numerous factors.  If you've been following some of the Imperial Colleges modelling, the predictions can change by orders of magnitude week to week.  A massive shift in their model's prediction is what caused Britain to go from light mitigation to lockdown.  Same thing with the US models.  Massive changes in predictions week to week.

Also, I would not trust any antibody testing to accurately report low rates of infection due to the problem of false positives. In the US, we recently approved an antibody test for emergency use only (not for diagnostic use by itself, but only in conjunction with other testing), and it has around 95% sensitivity and 95% specificity.  That means about 5% false negative and 5% false positive.  Sounds OK?  Depends on what you want to do.  If 0.1% of the population is actually infected, that antibody test is still going to identify about 5% of the population as infected, with virtually all being false positives.  Britain was claiming that they would have an antibody test soon, but they backtracked after actually testing the performance.  So going back to the Denmark antibody test, if the sensitivity of the test was 96.5%, that test would tell you that at least around 3.5% of the population was positive, even if the actual percentage was around 0.1%.  I'd need to see that their antibody test was properly validated, and not just based on the claimed sensitivity by the manufacturer.  Many manufacturers are claiming 100% sensitivity (0% false positives), but I'm extremely skeptical of those claims.  It's either based on extremely limited testing, or outright fabrication.  

26% to 41% of the population infected is evidence that that model shouldn't be relied on.  If that was the case, they should reach herd immunity in a matter of weeks, at the longest.  That's what I mean by the models generating predictions that vary by orders of magnitude based on relatively small tweaks in the values of the model's parameters or assumptions.  An antibody test with 5% false positives would work OK to confirm whether 40% of a population was infected though, because at that point, the number of real positives greatly outnumbers the numbers of false positives.

I think it's wildly optimistic to believe that we are only counting about 1 out of every 200 cases, which is what the Denmark antibody test results are saying.  Current worldwide CFR is now about 6% (will continue to go up as active cases are resolved).  6%/200 is 0.03%, which is much less fatal than influenza.  This number goes against an overwhelming amount of data that we currently have, and it's a clear indication that the results of the antibody test are not reliable.  I'm sure that we've undercounted to some degree, but I doubt it's anywhere near that amount.

[Mudguard]

5 hours ago, JoannaL said:

Today a study was pulished which may help determine the number of infected without symptoms. In a county with a major outbreak (1500 infected, 44 dead),  600 randomly selected persons were examined (antibody-testing). They found in  2 % active cases of a coronavirus infection  and in 15% of all persons antibodies. Which means the mortality rate is substaintely lower than expected previously ( something like 0,37 %)

As I mentioned in my reply to Ran, I would be very skeptical that an antibody test can accurately detect a low percentage of infected in a population, due to false positive problems.  I can't read German, but I think the article claims that the test has a specificity of greater than 99%.  If that's from the manufacturer, that's worthless.  I'd need to see the specificity confirmed through a rigorous validation process before I believe these numbers.  

If the specificity is actually only about 98%, the test would report that 2% of the population is positive even if the real percentages is zero.

[JoannaL]

5 minutes ago, Mudguard said:

As I mentioned in my reply to Ran, I would be very skeptical that an antibody test can accurately detect a low percentage of infected in a population, due to false positive problems.  I can't read German, but I think the article claims that the test has a specificity of greater than 99%.  If that's from the manufacturer, that's worthless.  I'd need to see the specificity confirmed through a rigorous validation process before I believe these numbers.  

If the specificity is actually only about 98%, the test would report that 2% of the population is positive even if the real percentages is zero.

As I understood it, they did not use a commercially available manufactured test kid, but did it themselves in their labs. If done correctly it should be rigorous validated. The German study is a prepublication. It will be published soon (and in English), and then we can see all the details.

[Mudguard]

7 minutes ago, JoannaL said:

As I understood it, they did not use a commercially available manufactured test kid, but did it themselves in their labs. If done correctly it should be rigorous validated. The German study is a prepublication. It will be published soon (and in English), and then we can see all the details.

We shall see.  Validation would require testing against a panel of common coronaviruses, other types of viruses, and hundreds of blood or plasma samples from early 2019 or earlier.  It's possible for an academic lab to do this, but it would be difficult.

[Zorral]

http://www.bbc.com/travel/story/20200407-turkeys-unique-hand-sanitiser

 

Quote

 

[...]

Meaning “cologne”, kolonya has been a treasured symbol of Turkish hospitality and health since the Ottoman Empire, and it’s often described as Turkey’s national scent. Traditionally, this sweet-scented aroma made with fig blossoms, jasmine, rose or citrus ingredients is sprinkled on guests’ hands as they enter homes, hotels and hospitals; when they finish meals at restaurants; or as they gather for religious services. But unlike other natural scents, this ethanol-based concoction’s high alcohol content can kill more than 80% of germs and act as an effective hand disinfectant.

[...]

 

 

[Fez]

Not much info here yet, but the South Korean CDC is suggesting that, rather than people getting re-infected, some people may have the virus re-activating in them. https://www.bloomberg.com/news/articles/2020-04-09/coronavirus-may-reactivate-in-cured-patients-korean-cdc-says?sref=yYYRek8e

Quote

 

About 51 patients classed as having been cured in South Korea have tested positive again, the CDC said in a briefing on Monday. Rather than being infected again, the virus may have been reactivated in these people, given they tested positive again shortly after being released from quarantine, said Jeong Eun-kyeong, director-general of the Korean CDC.

“While we are putting more weight on reactivation as the possible cause, we are conducting a comprehensive study on this,” Jeong said. “There have been many cases when a patient during treatment will test negative one day and positive another.”

 

Although this is possibly an issue with testing sensitivity instead of anything else. But if this the virus rather than the testing, I wonder if/how this squares with the issue of some recovered people having low antibody counts.

[Ran]

Here is a press conference from today regarding the Heinsberg study:

 

Skimmed through it using auto-translate, and he was asked about false positives/negatives as a predecessor at Bonn University pointed out the issue with some of the tests. While he did not get into why they were confident in the >99% specificity, he did say that they have worked hard on the subject, and that an earlier assessment they made had indicated 20% immune, not 15%, and I gather that that one had issues. Will have to wait and see what the full publication reveals. I've seen expectations that they used ELISA rather than any sort of new rapid techniques. 

Some more details about the people and labs involved via Reddit, from someone who appears to know one or more of the authors. The lab that ran the tests was apparently the national reference laboratory for coronavirii up to a few years ago, when it's head (Christian Drosten) moved to Berlin, but they still have a lot of expertise. 

[Jen'ari]

19 hours ago, Isis said:

TB gets imported to the UK from places where the healthcare infrastructure isn't quite as good. It's still around. 

Treatment is five different drugs for six months. Compliance for that is a massive issue, as you can imagine. 

Yeah this is what I thought, it gets bought in when people visit from other Countries where it’s more prevalent. I guess what I meant was it’s not really something most of the indigenous population in developed Countries with good healthcare would think was still around a great deal if that makes sense, sorry I worded it badly .

 

[Mudguard]

22 minutes ago, Ran said:

Here is a press conference from today regarding the Heinsberg study:

 

Skimmed through it using auto-translate, and he was asked about false positives/negatives as a predecessor at Bonn University pointed out the issue with some of the tests. While he did not get into why they were confident in the >99% specificity, he did say that they have worked hard on the subject, and that an earlier assessment they made had indicated 20% immune, not 15%, and I gather that that one had issues. Will have to wait and see what the full publication reveals. I've seen expectations that they used ELISA rather than any sort of new rapid techniques. 

Some more details about the people and labs involved via Reddit, from someone who appears to know one or more of the authors. The lab that ran the tests was apparently the national reference laboratory for coronavirii up to a few years ago, when it's head (Christian Drosten) moved to Berlin, but they still have a lot of expertise. 

They are going to have to provide all the testing they did to validate their test in order for people to trust the data.  Hopefully that is included in the full paper, or provided separately.

Generally speaking, all the antibody tests are ELISA tests, and should be very rapid, at least when compared to RT-PCR methods.  The Heinsburg study test should be similar to all the other tests in structure and methodology (there's two different types of structure for the test, which depends on whether you immobilize an antibody or antigen on the testing substrate).  Where the design comes in that differentiates all these antibody tests is what antigen (or antibody) is immobilized on the substrate. 

It's theoretically possible that they have engineered a set of antigens that results in a test with >99% specificity, but I'm skeptical.  Normally, this requires tweaking parts of the antigen, which requires genetic manipulation, then producing the antigen in a host cell (i.e., bacteria, yeast, human, etc.), purifying the antigen, and then testing and validating it in the test.  You have to do all this for each tweak you make.  To speed things up, you can do some of this in parallel, but it's a lot of work.  If you are lucky, you might hit upon the right antigen(s) right away, but generally, it will take many months to optimize the selection of antigens (or antibodies) for the test.  I think most or all of the SARS-CoV-2 antibody tests are immobilizing antigens, because it's the easiest and quickest thing to do.  Test design can begin immediately after the sequence of the virus is known, which was in early January.

[Mr. Chatywin et al.]

The dog and I went for a decent run/walk. About five minutes before we get home she starts freaking out.

It's literally either snowing or hailing sideways right now. I guess that's what I get for falling asleep without finishing the Ten Commandments. 

[Mudguard]

8 hours ago, JoannaL said:

Today a study was pulished which may help determine the number of infected without symptoms. In a county with a major outbreak (1500 infected, 44 dead),  600 randomly selected persons were examined (antibody-testing). They found in  2 % active cases of a coronavirus infection  and in 15% of all persons antibodies. Which means the mortality rate is substaintely lower than expected previously ( something like 0,37 %)

I misread your initial post.  Didn't realize that 15% of the samples tested positive.  I misread it as 2% positive.  15% is more interesting because if the false positive rate is relatively small, say less than 5% which is reasonable, the actual number of infected shouldn't be that far off from the test results.  The higher the % infected, the better the test should perform.  Concern is greatly magnified when infection rates are very low.  Their test would have to be very poor for the 15% result to be significantly off, so I'm much more willing to give them the benefit of the doubt here.  Still want to see the test validation results though.

[Aemon Stark]

18 hours ago, The Anti-Targ said:

If you are talking about the Netherlands I think you mean 20,000 not 200.

Well, I know where I live so...

10 hours ago, Hereward said:

Newfoundland and Labrador I believe.

Pays to be on an island sometimes. We're up to 236 cases now. Some speculation that there will be a big bump after Easter Weekend. 

5 hours ago, The Anti-Targ said:

TIL.

Google was no bloody help, it just kept throwing the Netherlands at me.

Well, I rather enjoy the idea that we could be confused with the Dutch. Kinda like confusing Iceland with Italy. 

[rotting sea cow]

11 hours ago, Ran said:

This is probably a very generous assumption.  Imperial College's modelling suggests the number is much higher. In fact, there's the first results (Reddit thread because it provides a translation of the key facts) of antibody reporting in Denmark  is useful because it uses March 28th breakpoint that Imperial College did. At that time, there were 917 detected cases. Imperial College estimated that with  1.1% (95% credible level) of the population was infected at that point with a range from .40 to 3.1%. Denmark's antibody test suggests to them that on March 28th 3.5% were actually infected.

So more people were infected then even the most extreme of the estimate. The most extreme of the Italy and Spain estimates based on their models were 26% and 41% respectively, so... yeah. We've likely detected substantially less than anyone thinks.

Also saw the NY Times believes that NYC has had cases as far back as mid-February.

I've seen a number of these studies. Maybe is the same source? I tend to be skeptical because it would mean that a huge number of cases are completely asymptomatic or very very mild, which goes against clinical studies. It would also mean that the disease is extremely contagious but actually not very dangerous, even for the risk population. I've seen more credible estimates that put the true number of cases in Italy around half a million or so.

 

 

[rotting sea cow]

9 hours ago, JoannaL said:

Today a study was pulished which may help determine the number of infected without symptoms. In a county with a major outbreak (1500 infected, 44 dead),  600 randomly selected persons were examined (antibody-testing). They found in  2 % active cases of a coronavirus infection  and in 15% of all persons antibodies. Which means the mortality rate is substaintely lower than expected previously ( something like 0,37 %)

There are two problems I see.

1) Estimating the mortality rate during an ongoing outbreak is tricky because of the delay between case onset and decease

2) There is no "natural" mortality rate, specially during a pandemic. The mortality rate here depend critically on the capability of the health care system to absorb the serious cases and treat them accordingly (also protocols change as our understanding progresses). But once you health care is overwhelmed the death rate will sharply rise.

[The Anti-Targ]

2 hours ago, Aemon Stark said:

Well, I know where I live so...

Pays to be on an island sometimes. We're up to 236 cases now. Some speculation that there will be a big bump after Easter Weekend. 

Well, I rather enjoy the idea that we could be confused with the Dutch. Kinda like confusing Iceland with Italy. 

236 isn't great for your population, now that I know where you are and what the population base is. 1:2200-ish is better than some places but worse than others. It's roughly 7 times higher rate than New Caledonia. But it's roughly 10 times better than Iceland. The real question for everywhere is what is the infection curve doing.

I hope people aren't stupid during Easter. Jesus won't congratulate them for spreading disease among fellow worshipers. Cops are out in force here making sure people traveling across country have legit reason to do so.

We have had some absolute a-holes going around coughing and spitting on cops, supermarket workers and people shopping in supermarkets. I think most have been uninfected (or should I say asymptomatic) pranksters, but I think one or two might be people who actually thought they had the disease.

[Fragile Bird]

14 minutes ago, The Anti-Targ said:

236 isn't great for your population, now that I know where you are and what the population base is. 1:2200-ish is better than some places but worse than others. It's roughly 7 times higher rate than New Caledonia. But it's roughly 10 times better than Iceland. The real question for everywhere is what is the infection curve doing.

I

You may have missed my previous comments, but one guy who came home from travels is responsible for at least 143 of those cases. Just after he came back two people he knew died and he attended both funerals and wakes. And just like that, wham!

I was going to link an article, but I see now the story has been updated - there is the possibility a woman who came in from another province to attend a different funeral could have been the super-spreader.

In the meantime nobody knows who it was and there is deep fear and suspicion aimed at every person who attended the main funeral, that of a woman in her 30s who died from complications from diabetes. Some people have even received death threats. It’s a small community of 6,000 people.

[The Anti-Targ]

Looks like India is about to blow up in official numbers. I imagine it's already blown up in unconfirmed numbers. Such a crowded chaotic country, we might soon be seeing how things might have gone in China if it hadn't put those draconian measures in place. I was looking at a short term posting to our embassy (high commission I guess since we are both commonwealth countries) in New Delhi in the middle of this year. It may be wishful thinking to plan for it to happen early next year.

Musing a bit more about seasonality. There's been a lot of talk recently about vitamin D having an influence on viral infections. Assuming most people don't take supplements for Vit D this could be a factor in why viral illnesses like cold and 'flu peak in winter. During winter we will all have lower levels, on average, of Vit D than in summer simply because we are out in the sun less and what sun we do get is far less intense. So I think the message to people is that regardless of lockdown situations, try to get out in the sun as much as possible while maintining social distance. And take Vit D supplements for those of us in the Southern Hemisphere heading into winter.

[Fury Resurrected]

14 minutes ago, The Anti-Targ said:

Looks like India is about to blow up in official numbers. I imagine it's already blown up in unconfirmed numbers. Such a crowded chaotic country, we might soon be seeing how things might have gone in China if it hadn't put those draconian measures in place. I was looking at a short term posting to our embassy (high commission I guess since we are both commonwealth countries) in New Delhi in the middle of this year. It may be wishful thinking to plan for it to happen early next year.

Musing a bit more about seasonality. There's been a lot of talk recently about vitamin D having an influence on viral infections. Assuming most people don't take supplements for Vit D this could be a factor in why viral illnesses like cold and 'flu peak in winter. During winter we will all have lower levels, on average, of Vit D than in summer simply because we are out in the sun less and what sun we do get is far less intense. So I think the message to people is that regardless of lockdown situations, try to get out in the sun as much as possible while maintining social distance. And take Vit D supplements for those of us in the Southern Hemisphere heading into winter.

I hyperdose vitamin D (20,000 iu first month then 10,000iu thereafter) from October to May for SAD and it does nothing at all to keep me from getting all the cold and flu symptoms everyone else does.

[williamjm]

2 hours ago, The Anti-Targ said:

We have had some absolute a-holes going around coughing and spitting on cops, supermarket workers and people shopping in supermarkets. I think most have been uninfected (or should I say asymptomatic) pranksters, but I think one or two might be people who actually thought they had the disease.

I saw a news story recently that some people doing similar things in America were being charged under anti-terrorism laws, which sounded slightly surprising at first but on reflection it does feel appropriate.

[Fragile Bird]

18 minutes ago, williamjm said:

I saw a news story recently that some people doing similar things in America were being charged under anti-terrorism laws, which sounded slightly surprising at first but on reflection it does feel appropriate.

Some of them are they same kind of people who used to make jokes about bombs in their suitcases at airports.