The preprint on #SARSCoV2 seroprevalence in Santa Clara County continues to make headlines. They estimate 2-4% of the population had #COVID19 by April 4 implying an infection fatality rate (IFR) of 0.1 to 0.2%.

But there are many reasons to be very skeptical. Thread...

What would these numbers imply for areas like New York, Madrid or similar?

There have been ~9k #COVIDー19 deaths in NYC. IFR 0.1 to 0.2% implies 50-100% of 8.4M New Yorkers were already infected even if there were no more deaths!

All of NCY -- not just heavily affected areas!

Numbers for Madrid and other places with high #COVID19 prevalence are similar.

To me, this suggests something is odd with the results from Santa Clara, at least in the way they are interpreted. Let's look at the data....

They tested n=3300 with k=50 positive tests. The test validation by the manufacturer and the study yielded

TP = 78 + 25
TN = 369 + 30
FP = 2 + 0
FN = 7 + 12

How do we turn this into an estimate of prevalence not knowing the true specificity s and sensitivity r?

Following @NimwegenLab's analysis, we can turn the results for test validation and testing in Santa Clara into a posterior probability for prevalence f by integrating over the unknown r and s. (Z being the normalization)

This results in a posterior distribution for the prevalence f which is peaked somewhere around 1% but compatible with any value below 2%. Hence the data are pretty uninformative.

I have not considered the demographic corrections they have done in the study as they seem tangential to the main point made here:

All existing evidence points to a high IFR>0.5% and the Santa Clara data does not change this.

This is Erik's tweet that inspired this thread: https://twitter.com/NimwegenLab/status/1251261262366900232