4 ms·

> Serological data from a random sample of the German town of Gangelt shows that even a robust testing regime missed about 90% of the cases. The results are di

> Serological data from a random sample of the German town of Gangelt shows that even a robust testing regime missed about 90% of the cases. The results are disputed to say the least. A random sampling study in Austria found that the true number of infected is likely to be only three times higher. > Analysis of this data suggests about 50% of people never develop symptoms At the time of the test asymptotic. They did not say never develop symptoms.

mlyle 7y ago
The odds ratio approach was backwards-looking, so it means "never developed symptoms [in a couple of weeks]". Seems Iceland is testing a lot more than NYC and that their focused testing (which is still broader than NYC's overall testing) misses ~90% of cases. Note that the Austrian study I've read shows that current infection as detected by RT-PCR is about 3x higher than present case count, which isn't quite the same thing I'm talking about with the antibody tests in Germany. If you're referring to something else, please share. In particular, no one knows the sensitivity for RT-PCR of asymptomatic people and it might be rather low--- it's not exactly wonderful in people who are quite sick and presumably shedding more virus. I'm hoping in the next week or so we have antibody data from a couple of US jurisdictions, including the SF Bay Area. It can really inform policy.
- the_mitsuhiko 7y ago
  > Note that the Austrian study I've read shows that current infection as detected by RT-PCR is about 3x higher than present case count, which isn't quite the same thing I'm talking about with the antibody tests in Germany. The reason Austria wasn't doing antibody tests yet is because the quality of these is not anywhere close enough to be reliable.
  - mlyle 7y ago
    There's good antibody tests at this point, that validate with no false positives on serum predating COVID-19. Just today, new data from Santa Clara County: https://www.medrxiv.org/content/10.1101/2020.04.14.20062463v1 https://www.medrxiv.org/content/10.1101/2020.04.14.20062463v... Stanford estimates 2.49% to 4.16% seropositivity, versus approximately 1000 reported cases as of that date-- so about 75x the actual reported cases. Of course, the way that Stanford recruited their sample means that people who were previously exposed might be more likely to report, so I think this overstates things a little, but...
  - mlyle 7y ago
    Holy crap, check out these guys' estimate for New York: https://www.medrxiv.org/content/10.1101/2020.04.13.20064519v1.full.pdf https://www.medrxiv.org/content/10.1101/2020.04.13.20064519v...
  - mlyle 7y ago
    Now we have this to add to the pile (though I can't find an actual preprint): http://archive.is/UDrho http://archive.is/UDrho
mlyle 7y ago
Now we can add the Netherlands to the mix. 29214 reported cases / 17280000 population = 0.17%. But apparently about 3% of blood donors in the Netherlands have antibodies. https://nltimes.nl/2020/04/16/3-dutch-blood-donors-covid-19-antibodies https://nltimes.nl/2020/04/16/3-dutch-blood-donors-covid-19-... If blood donors are representative of the population as a whole, that means about 94% of cases have been missed. Again-- with better testing than the US.

mlyle 7y ago

The odds ratio approach was backwards-looking, so it means "never developed symptoms [in a couple of weeks]". Seems Iceland is testing a lot more than NYC and that their focused testing (which is still broader than NYC's overall testing) misses ~90% of cases. Note that the Austrian study I've read shows that current infection as detected by RT-PCR is about 3x higher than present case count, which isn't quite the same thing I'm talking about with the antibody tests in Germany. If you're referring to something else, please share. In particular, no one knows the sensitivity for RT-PCR of asymptomatic people and it might be rather low--- it's not exactly wonderful in people who are quite sick and presumably shedding more virus. I'm hoping in the next week or so we have antibody data from a couple of US jurisdictions, including the SF Bay Area. It can really inform policy.

the_mitsuhiko 7y ago
> Note that the Austrian study I've read shows that current infection as detected by RT-PCR is about 3x higher than present case count, which isn't quite the same thing I'm talking about with the antibody tests in Germany. The reason Austria wasn't doing antibody tests yet is because the quality of these is not anywhere close enough to be reliable.
- mlyle 7y ago
  There's good antibody tests at this point, that validate with no false positives on serum predating COVID-19. Just today, new data from Santa Clara County: https://www.medrxiv.org/content/10.1101/2020.04.14.20062463v1 https://www.medrxiv.org/content/10.1101/2020.04.14.20062463v... Stanford estimates 2.49% to 4.16% seropositivity, versus approximately 1000 reported cases as of that date-- so about 75x the actual reported cases. Of course, the way that Stanford recruited their sample means that people who were previously exposed might be more likely to report, so I think this overstates things a little, but...
- mlyle 7y ago
  Holy crap, check out these guys' estimate for New York: https://www.medrxiv.org/content/10.1101/2020.04.13.20064519v1.full.pdf https://www.medrxiv.org/content/10.1101/2020.04.13.20064519v...
- mlyle 7y ago
  Now we have this to add to the pile (though I can't find an actual preprint): http://archive.is/UDrho http://archive.is/UDrho

mlyle 7y ago

Now we can add the Netherlands to the mix. 29214 reported cases / 17280000 population = 0.17%. But apparently about 3% of blood donors in the Netherlands have antibodies. https://nltimes.nl/2020/04/16/3-dutch-blood-donors-covid-19-antibodies https://nltimes.nl/2020/04/16/3-dutch-blood-donors-covid-19-... If blood donors are representative of the population as a whole, that means about 94% of cases have been missed. Again-- with better testing than the US.