Note: This thread is related to #Coronavirus #COVID19

Follow the World Health Organization's instructions to reduce your risk of infection:

1/ Frequently clean hands by using alcohol-based hand rub or soap and water.

2/ When coughing and sneezing cover mouth and nose with flexed elbow or tissue - throw issue away immediately and wash hands.

3/ Avoid close contact with anyone that has fever and cough.

Natalie E. Dean, PhD+ Your Authors @nataliexdean Assistant Professor of Biostatistics at @UF specializing in emerging infectious diseases and vaccine study design. @HarvardBiostats PhD. Tweets my own. She/her. May. 18, 2020 2 min read + Your Authors

The LA County serosurvey results are now published in JAMA. I'll offer a few quick comments, since I guess that is what I do. 1/n 

The authors used a market research database to select participants. One adult per household was invited to visit a testing site, or offered in-home testing. The in-home testing component is useful because not everyone may have a convenient means of transportation. 2/n

The biggest challenge is consent. 50.9% of individuals invited to participate were tested. So this is why the authors acknowledge that "selection bias is likely." 3/n

What stuck out to me is that 43% of participants had a household income >$100K. Median income for the county is about $68K. This suggests a skew towards wealthier participants. LA County is also nearly half Hispanic, but only 22% in the study. 4/n 

I am not sure how much any of these factors influence the results. There is an unusual U shape with income and seropositivity (last column), with highest income comparable to lowest income. But the numerators are small, so I caution against over-interpreting differences. 5/n

(As an aside, at the bottom, I find the result that of 113 individuals that reported fever with cough in the last two months, 8.85% were seropositive - about 2.5x more than those without. So if you had a fever and cough, still more likely than not it was not COVID.) 6/n

The unweighted proportion positive for IgM or IgG was 35/863 = 4.06% (95% CI 2.84 - 5.60%). Unlike the original Santa Clara analysis, the results did not change a huge amount after reweighting for population characteristics and sensitivity/specificity. 7/n

I'll let others chime in on the sensitivity and specificity of the test and the confidence intervals (adequately wide?). They used sensitivity of 82.7% and specificity of 99.5%, referencing their Santa Clara preprint for the validation procedure. 8/n

Their main conclusion is that 4.65% infected translates to roughly 367,000 adults in LA county. We can crudely estimate 45X infected versus detected, though this math doesn't account for lags in seroconversion. If we take their lower bound of 2.5%, the factor is about 24X. 9/n

My take. It is hard to know how consent bias plays into the results. With 50% consent, anything is possible. What does it mean that the sample includes so many wealthy white adults? It's notable that the wealthiest group does not the have lowest prevalence. 10/n

I'd like to hear from others whether they think the lab validation problems have been addressed. Personally, I'm still skeptical about the 45X (these large extrapolation factors are unstable), but remain open to something larger than the 10-20X I have quoted. 11/END

One extra comment (slightly more controversial). Their stated conclusions are very mild. "Fatality rates based on confirmed cases may be higher than...on number of infections." Notably, they don't even report an IFR. This is peer review plus an editor. Less sensational results.

You can follow @nataliexdean.


Tip: mention @threader_app on a Twitter thread with the keyword “compile” to get a link to it.

Enjoy Threader? Sign up.

Since you’re here...

... we’re asking visitors like you to make a contribution to support this independent project. In these uncertain times, access to information is vital. Threader gets 1,000,000+ visits a month and our iOS Twitter client was featured as an App of the Day by Apple. Your financial support will help two developers to keep working on this app. Everyone’s contribution, big or small, is so valuable. Support Threader by becoming premium or by donating on PayPal. Thank you.

Follow Threader