Information on US COVID-19 seroprevlance surveys can be found on the CDC website [0]. There are also a number of more locally sampled sero surveys going on concurrently right now by state/city/county public health departments. 'Sero' surveys are a commonly used tool in infectitious disease research to understand the difference between identified and total case numbers. They have tons of pros and cons (see [1] for some examples). There are also different types [2]
They aren't exact, but the goal with this type of research isn't perfect accuracy, that is impossible, it's better information than we currently have. The prevalence number that you get from a sero survey is hugely important to accurately estimating the effectiveness of medical tests. However, getting that data into a model of test accuracy isn't necessarily easy. [3]
This is an example of a COVID sero survey from Santa Clara County last April [4]. Over two days they collected around 3500 samples over two days. There results summary is basically this:
The raw prevalence of antibodies to SARS-CoV-2 in our sample was 1.5%. Weighted for population demographics of Santa Clara County, the prevalence was 2.8%. Using those prevalence numbers, the unweighted estimate of case numbers was 23,000. Using the weighted prevalence number, the estimate was 54,000. In early April, there were approximately 1,000 confirmed cases in Santa Clara County.
From my epidemiologist friends, they think 50% diagnosed cases is if anything a massive overestimate of the diagnostic rate and something like 1 in 4 diagnosed cases is probably accurate (this is hearsay that I can't cite from people in the field). The problem is that the 23,000-54,000 people who 'had' (air quotes) COVID last April may or may not be immune at this point. The numbers may be high or low ...either way they are BAD
Not even getting into what this means about vaccine policy...just talking about what the data and numbers are here. I'm still trying to process what the original post means/is trying to say honestly.
Wasn't [4] widely criticized for messing up basic statistics? Either way, numbers from April might as well not exist for what they tell us about today.
I haven't seen a particular critique of that paper so can't comment.
Bigger picture, I wasn't commenting on the particular validity of that paper as much as what this type of study could tell us...hindsight I should have picked an example from another disease.
Unfortunately that paper became my litmus test for people arguing from bad faith as it was such a darling of the "COVID is just the flu!" crowd. My memory of the main problem was that the confidence interval of the false positive rate of the test used included values higher than the entire positive rate of the sampled people. I.e. they didn't have enough data to rule out the claim that no one had COVID antibodies, let alone to justify that such a shockingly high percentage of the population could have had them already. Here's a random overview of the critiques that I found on google just now:
Anyway, I'm sure better data exists at this point. The CDC survey data you linked to looks very interesting and seems to (with caveats) put the percent-immune at 16% or less as of a few months ago. It's sort of a middle ground with what the article here was claiming - 100 million immune Americans still seems to high but not by that much so given how cases have continued since the last serological survey results.
They aren't exact, but the goal with this type of research isn't perfect accuracy, that is impossible, it's better information than we currently have. The prevalence number that you get from a sero survey is hugely important to accurately estimating the effectiveness of medical tests. However, getting that data into a model of test accuracy isn't necessarily easy. [3]
This is an example of a COVID sero survey from Santa Clara County last April [4]. Over two days they collected around 3500 samples over two days. There results summary is basically this:
The raw prevalence of antibodies to SARS-CoV-2 in our sample was 1.5%. Weighted for population demographics of Santa Clara County, the prevalence was 2.8%. Using those prevalence numbers, the unweighted estimate of case numbers was 23,000. Using the weighted prevalence number, the estimate was 54,000. In early April, there were approximately 1,000 confirmed cases in Santa Clara County.
From my epidemiologist friends, they think 50% diagnosed cases is if anything a massive overestimate of the diagnostic rate and something like 1 in 4 diagnosed cases is probably accurate (this is hearsay that I can't cite from people in the field). The problem is that the 23,000-54,000 people who 'had' (air quotes) COVID last April may or may not be immune at this point. The numbers may be high or low ...either way they are BAD
Not even getting into what this means about vaccine policy...just talking about what the data and numbers are here. I'm still trying to process what the original post means/is trying to say honestly.
[0] https://www.cdc.gov/coronavirus/2019-ncov/cases-updates/geog...
[1] https://www.who.int/immunization/monitoring_surveillance/bur...
[2] https://www.cdc.gov/coronavirus/2019-ncov/covid-data/seropre...
[3] https://academic.oup.com/aje/article/190/1/109/5893084?login...
[4] https://www.medrxiv.org/content/10.1101/2020.04.14.20062463v...