5 ms·

Antibody tests suggest that coronavirus infections vastly exceed official counts

31 points by sohkamyung 7y ago

skybrian 7y ago
Already being discussed here: https://news.ycombinator.com/item?id=22899272 https://news.ycombinator.com/item?id=22899272
troughway 7y ago
And who knows how long this thing has been in circulation for. Questioning the narrative gets you downvoted on here, however (https://news.ycombinator.com/item?id=22862003 https://news.ycombinator.com/item?id=22862003)
pdonis 7y ago
The title, while it is the actual title of the Nature article, is arguably linkbait. What it should be is something more like: "Antibody tests in one California county, if taken at face value and naively extrapolated to everywhere else, imply that COVID-19 infections vastly exceed official counts". Which is a lot weaker than the claim the actual title is making.
- busymom0 7y ago
  I remember last week reading something similar coming from Chicago too. And Italy had similar reports where some 50% of people who came to donate blood already had antibodies. Would be interesting to hear how other places score in the antibody testing.
- pdonis 7y ago
  Update: It looks like the title of the Nature article now has "suggest" instead of "reveal". So the title of this thread should be changed to correspond.
  - dang 7y ago
    Changed now. Thanks!
    - pdonis 7y ago
      Btw, I used the hn@ycombinator.com email to bring this to your attention, but it would be nice if there were a way to do that from HN itself (maybe a simple web form that sent an email to that address when submitted?).
      - dang 7y ago
        Good idea! I'll add it to our list.
busymom0 7y ago
dang, there was a comment here by someone else which was linking to their comment on another HN post and how it got downvoted (even though now they are proven right). I was replying to that comment until HN told me that comment has been flagged and now it's not even visible on this post. This is quite sad from a community which is supposed to be of scientists (who are usually supposed to be a bit skeptic). This was my reply to that comment: Similar to how Twitter, FB and IG banned several accounts (I won't name them as I don't want to get in political wars on HN) which had been questioning the origins of the virus since January. Some people who were saying back in January that China was lying and this could end up becoming a pandemic were called names and fringe conspiracy theorists and banned. Now even Taiwan is showing emails from December that they had been warning about China's lies to WHO but were ignored. I really dislike how being skeptic now a days gets you labelled with bad names and removed.
- dang 7y ago
  This community is not "supposed to be of scientists" and it's certainly not a scientific forum. It's an internet watercooler. People speculate. If you mean https://news.ycombinator.com/item?id=22904356 https://news.ycombinator.com/item?id=22904356, that comment was flagged by users; rightly so since it broke the site guidelines. (You need to set 'showdead' to 'yes' in your profile to read dead comments.)
DenisM 7y ago
> An analysis of the blood of some 3,300 people living in Santa Clara county found that, by early April, one in every 66 people had been infected with SARS-CoV-2 So it's 1.5% then, according to this study. With effect size so small one has to look at the test's accuracy. This one lists 20% false negative and 0% false positive, however the sample size for the test's evaluation was in low dozens. According to the article evaluating a test properly requires a sample size of hundreds. So we don't know with any certainty how good this test is because it was not sufficiently evaluated. If the false positive rate is over 1.5% the results would become invalid. For a sense of scale consider that the Bodysphere antibody test has 9% false positive rate, and some other tests are at 4.5%.
robocat 7y ago
Here’s another population test: Andover Subacute has beds for 700 patients. “The 17 [deaths] were among 68 recent deaths linked to the long-term care facility, Andover Subacute and Rehabilitation Center I and II, including two nurses, officials said. Of those who died, 26 people had tested positive for the virus. For the others, the cause of death is unknown.”. 2 nurses dead, what was the total staff numbers? 68 other deaths. What was normal rate? Can you then project the death rates back to a country population (even given the deaths were obviously elderly unwell people).
bufferoverflow 7y ago
That article title is absolutely awful. They did a tiny test in some under-tested area and extrapolated it to everywhere. What we really need is random sampling tests from large areas, not tiny counties. And you don't even need that many of them to have a statistically significant result. I'd say 10-50 thousand random tests across the US will give you a very accurate figure of how many people have had the virus. I also highly doubt the claims that the mortality of this virus is low. It's demonstrably not. We have hospitals that used to have a few deaths a week (from all causes), and now have multiple deaths per day (mostly corona). With only around 1% of the population infected.
- goalieca 7y ago
  With only 1% of the population infected? Nope. It’s definitely a lot more than that. You know where the claim that up to 80% of people can be asymptomatic came from? It was when they tested all pregnant women for delivery at a hospital in New York. I recall that they found 15% were infected.
Gys 7y ago
Research in the Netherlands shows 3% of blooddonors have anti-bodies [0] (or [1] for translation). The Netherlands has about 17 million people. Assuming the 3% translates to only 1% of the population, that still would be 170.000 Officially they found 30.000 positively tested cases [2]. So even with this (I think) conservative estimate there are at the very least 5 times more infected people than tested cases. So proably most of the infected people really do not seem to have any sympthoms. [0] https://www.nu.nl/coronavirus/6045092/rivm-ongeveer-3-procent-bloeddonoren-heeft-antistoffen-tegen-coronavirus.html https://www.nu.nl/coronavirus/6045092/rivm-ongeveer-3-procen... [1] https://translate.google.com/translate?hl=en&sl=auto&tl=en&u=https%3A%2F%2Fwww.nu.nl%2Fcoronavirus%2F6045092%2Frivm-ongeveer-3-procent-bloeddonoren-heeft-antistoffen-tegen-coronavirus.html https://translate.google.com/translate?hl=en&sl=auto&tl=en&u... [2] https://www.rivm.nl/coronavirus-covid-19/actueel https://www.rivm.nl/coronavirus-covid-19/actueel

31 points by sohkamyung 7y ago

Already being discussed here: https://news.ycombinator.com/item?id=22899272 https://news.ycombinator.com/item?id=22899272

troughway 7y ago

And who knows how long this thing has been in circulation for. Questioning the narrative gets you downvoted on here, however (https://news.ycombinator.com/item?id=22862003 https://news.ycombinator.com/item?id=22862003)

pdonis 7y ago

The title, while it is the actual title of the Nature article, is arguably linkbait. What it should be is something more like: "Antibody tests in one California county, if taken at face value and naively extrapolated to everywhere else, imply that COVID-19 infections vastly exceed official counts". Which is a lot weaker than the claim the actual title is making.

busymom0 7y ago
I remember last week reading something similar coming from Chicago too. And Italy had similar reports where some 50% of people who came to donate blood already had antibodies. Would be interesting to hear how other places score in the antibody testing.
pdonis 7y ago
Update: It looks like the title of the Nature article now has "suggest" instead of "reveal". So the title of this thread should be changed to correspond.
- dang 7y ago
  Changed now. Thanks!
  - pdonis 7y ago
    Btw, I used the hn@ycombinator.com email to bring this to your attention, but it would be nice if there were a way to do that from HN itself (maybe a simple web form that sent an email to that address when submitted?).
    - dang 7y ago
      Good idea! I'll add it to our list.

busymom0 7y ago

dang, there was a comment here by someone else which was linking to their comment on another HN post and how it got downvoted (even though now they are proven right). I was replying to that comment until HN told me that comment has been flagged and now it's not even visible on this post. This is quite sad from a community which is supposed to be of scientists (who are usually supposed to be a bit skeptic). This was my reply to that comment: Similar to how Twitter, FB and IG banned several accounts (I won't name them as I don't want to get in political wars on HN) which had been questioning the origins of the virus since January. Some people who were saying back in January that China was lying and this could end up becoming a pandemic were called names and fringe conspiracy theorists and banned. Now even Taiwan is showing emails from December that they had been warning about China's lies to WHO but were ignored. I really dislike how being skeptic now a days gets you labelled with bad names and removed.

dang 7y ago
This community is not "supposed to be of scientists" and it's certainly not a scientific forum. It's an internet watercooler. People speculate. If you mean https://news.ycombinator.com/item?id=22904356 https://news.ycombinator.com/item?id=22904356, that comment was flagged by users; rightly so since it broke the site guidelines. (You need to set 'showdead' to 'yes' in your profile to read dead comments.)

DenisM 7y ago

> An analysis of the blood of some 3,300 people living in Santa Clara county found that, by early April, one in every 66 people had been infected with SARS-CoV-2 So it's 1.5% then, according to this study. With effect size so small one has to look at the test's accuracy. This one lists 20% false negative and 0% false positive, however the sample size for the test's evaluation was in low dozens. According to the article evaluating a test properly requires a sample size of hundreds. So we don't know with any certainty how good this test is because it was not sufficiently evaluated. If the false positive rate is over 1.5% the results would become invalid. For a sense of scale consider that the Bodysphere antibody test has 9% false positive rate, and some other tests are at 4.5%.

robocat 7y ago

Here’s another population test: Andover Subacute has beds for 700 patients. “The 17 [deaths] were among 68 recent deaths linked to the long-term care facility, Andover Subacute and Rehabilitation Center I and II, including two nurses, officials said. Of those who died, 26 people had tested positive for the virus. For the others, the cause of death is unknown.”. 2 nurses dead, what was the total staff numbers? 68 other deaths. What was normal rate? Can you then project the death rates back to a country population (even given the deaths were obviously elderly unwell people).

bufferoverflow 7y ago

That article title is absolutely awful. They did a tiny test in some under-tested area and extrapolated it to everywhere. What we really need is random sampling tests from large areas, not tiny counties. And you don't even need that many of them to have a statistically significant result. I'd say 10-50 thousand random tests across the US will give you a very accurate figure of how many people have had the virus. I also highly doubt the claims that the mortality of this virus is low. It's demonstrably not. We have hospitals that used to have a few deaths a week (from all causes), and now have multiple deaths per day (mostly corona). With only around 1% of the population infected.

goalieca 7y ago
With only 1% of the population infected? Nope. It’s definitely a lot more than that. You know where the claim that up to 80% of people can be asymptomatic came from? It was when they tested all pregnant women for delivery at a hospital in New York. I recall that they found 15% were infected.

Gys 7y ago

Research in the Netherlands shows 3% of blooddonors have anti-bodies [0] (or [1] for translation). The Netherlands has about 17 million people. Assuming the 3% translates to only 1% of the population, that still would be 170.000 Officially they found 30.000 positively tested cases [2]. So even with this (I think) conservative estimate there are at the very least 5 times more infected people than tested cases. So proably most of the infected people really do not seem to have any sympthoms. [0] https://www.nu.nl/coronavirus/6045092/rivm-ongeveer-3-procent-bloeddonoren-heeft-antistoffen-tegen-coronavirus.html https://www.nu.nl/coronavirus/6045092/rivm-ongeveer-3-procen... [1] https://translate.google.com/translate?hl=en&sl=auto&tl=en&u=https%3A%2F%2Fwww.nu.nl%2Fcoronavirus%2F6045092%2Frivm-ongeveer-3-procent-bloeddonoren-heeft-antistoffen-tegen-coronavirus.html https://translate.google.com/translate?hl=en&sl=auto&tl=en&u... [2] https://www.rivm.nl/coronavirus-covid-19/actueel https://www.rivm.nl/coronavirus-covid-19/actueel