New Stanford preprint has serology data on 3300+. Claims 50-85X as many with antibodies as number of confirmed cases. If holds up, would be great news.
One q: was sample representative? Or was it folks who believed they'd been exposed and seeking testing? medrxiv.org/content/10.110…
One q: was sample representative? Or was it folks who believed they'd been exposed and seeking testing? medrxiv.org/content/10.110…
The other thing I looked at was calibration data.
Paper's conclusion would be at risk if test had a high false positive rate. But it looks like their test instead has a high false negative rate.
Still, need to do the math & think about possible sources of false positives.
Paper's conclusion would be at risk if test had a high false positive rate. But it looks like their test instead has a high false negative rate.
Still, need to do the math & think about possible sources of false positives.
*Possible* sources for too many positives:
- sample wasn't representative, enriched for COVID-19 cases
- test produced false positives
- test produced "true" positives but for antibodies to other coronaviruses, though pre-COVID sera results seem to rule this out
Need to think.
- sample wasn't representative, enriched for COVID-19 cases
- test produced false positives
- test produced "true" positives but for antibodies to other coronaviruses, though pre-COVID sera results seem to rule this out
Need to think.
If the paper doesn't hold up, it may be because 50 people in their sample knew they had symptoms or exposure and wanted a free COVID-19 test.
We can reject this hypothesis with a true representative sample, with people selected at random & compelled to be tested (like jury duty)
We can reject this hypothesis with a true representative sample, with people selected at random & compelled to be tested (like jury duty)
I'm glad they did the study and very supportive of serosurveys, but cautiously skeptical of the "everyone-already-has-it" hypothesis, because it would mean this virus tore through the world so much faster than past pandemics that took >1 year.
1) Many people with mild symptoms still couldn't easily get tests in Bay Area in early April. So perhaps some symptomatic people did study JUST to get tested?
2) Since it was on FB, link to study could have been posted in a COVID private group. Could also mean more positives.
2) Since it was on FB, link to study could have been posted in a COVID private group. Could also mean more positives.
*If* their study recruitment links could be shared, then subjects could recruit other subjects.
And if any of them recruited from a private FB group for people who suspected they had COVID-19, wanted testing, and couldn't get it any other way...that would spike positives.
And if any of them recruited from a private FB group for people who suspected they had COVID-19, wanted testing, and couldn't get it any other way...that would spike positives.
Three mechanisms that could lead to enrichment for positives:
- symptomatic people seeking tests they can't get elsewhere
- those people recruiting others.
- the nonzero false positive rate, comparable to the study's overall positive rate
- symptomatic people seeking tests they can't get elsewhere
- those people recruiting others.
- the nonzero false positive rate, comparable to the study's overall positive rate
