I’m sure many have seen this, and I’m hoping Mark Sumner will diary it. I’m NOT an expert in this field, so hope that someone who is will provide more/better info. Just want to point to the info for those who are interested and especially those who can better analyze.
NYTimes piece sums up a big study of efficacy of serology antibody tests. Well, it’s “big” in terms of # of scientists involved, # of tests, …. but not that many known positive (up to 36 I think) or known negative (52) samples. It’s another preprint, as paper is undergoing review.
Some good-ish results, but...
"Other scientists were less sanguine... Four of the tests produced false-positive rates ranging from 11 percent to 16 percent; many of the rest hovered around 5 percent."
NYT: https://www.nytimes.com/...
Contains link to actual paper: https://www.dropbox.com/...
The "Premier" test used in the Stanford/USC studies found 2 false positives (one for each manner of testing, IgM or IgG) from the 52 samples, and scores well overall in the sensitivity range (97-99%) -- but, big enough false positive rate to be VERY concerning when their overall population rate was found to be basically within that false positive range, then only slightly higher when they "corrected" to make their sample more representative of population.
I don't know what test was used in the NY/NYC study that Cuomo cited.
ALL of the tests had a pretty high false-Negative results (missing cases of people who’d tested positive). I don’t know how the studies that have been reported deal with the false negatives — I assume they build it into their estimates of population prevalence, by I don’t know. Only one test reached about 100% detection, but that was only after 20 days post symptomology (VERY small sample). Which brings me to..
Another interesting aspect of the study was re-testing of sub samples of people at many points after symptoms appeared (tests administered w/in first 5 days, 6-10, 11-15, 15-20, or over 20). Unfortunately, declining number of samples were available for the longer test periods (only 11 after 20 days). Tentatively, the percent of tests detecting the virus rises throughout those 20+ days. I surely hope they’ll follow patients longer and get tests further out, to see how long antibody detection goes out.
To sum up, the tests are useful, presumably people are still working on getting better and better tests, but results are hardly iron-clad in either direction.