Results adjusted in LA antibody test

Researchers behind a large-scale COVID-19 antibody study conducted in Los Angeles County in April have provided more specific estimates for county-wide antibody prevalence in a revised version of the study published in the Journal for American Medicine earlier this week.

The study was conducted by the University of Southern California and the Los Angeles County Department of Public Health on April 10-11 and, unlike a similar study conducted in Santa Clara County where participants were recruited via Facebook ads, the participants in the Los Angeles County study were recruited by market services firm LRW Group, which used a large proprietary database to create a random sample of the county population.

Of the 863 individuals tested in early April, 4.1 percent were found to have COVID-19 antibodies. When adjusting for statistical margin of error, researches originally estimated that 2.8 percent to 5.6 percent of the county's adult population has already been infected, which translates to 221,000 to 442,000 people. In the new updated JAMA version, researchers settle on a more precise prevalence figure of 4.65 percent, or 360,000 residents infected by early April. For reference, there were nearly 8,000 confirmed cases when the study was conducted.

"The number of confirmed COVID-19 cases is a poor proxy for the extent of infection in the community,” said lead investigator Neeraj Sood, professor at the USC Price School for Public Policy and senior fellow at the USC Schaeffer Center for Health Policy and Economics. “We need to update models and forecasts based on new evidence. We’re still far away from herd immunity or the end of the epidemic. We need to look at a longer time horizon when evaluating policy decisions.”

The Stanford researchers behind the Santa Clara study recently adjusted their own findings to estimate that 2.8 percent of the county had been infected in early April, which translates to 54,000 actual infections compared to just 1,000 confirmed cases. The researchers use a 19-day lag from infection to death, and when using the 94 deaths in the county reported on April 22, are able calculate a "true infection fatality rate" of .17 percent.

However, the Santa Clara study's weighting system and methodology have been criticized by other scientists. Immediately after the initial version of the study was released, Dr. George Rutherford, an epidemiologist at UCSF, expressed skepticism of the study's weights and suggested that people should look at the raw antibody prevalence percentage of 1.5 percent. An unweighted antibody prevalence of 1.5 percent translates to a total of 28,920 residents in the county that had been infected, a figure that can be used to calculate an infection fatality rate of .33 percent — a figure almost identical to the infection fatality rate calculated following antibody tests in a hard-hit German town.

The authors of the Los Angeles study did not calculate a "true fatality rate," but when using the same 19-day infection-to-death lag the Stanford researchers used in their study, one can calculate an infection fatality rate of .31 percent in Los Angeles County when using the 1,148 total deaths recorded on April 29. That figure is almost perfectly in-line with the unweighted Santa Clara figures as well as the German town estimations. After large-scale antibody testing in New York, Governor Andrew Cuomo hypothesized the true infection fatality rate in his state was around .5 percent.

https://www.sfgate.com/news/editorspicks/article/Los-Angeles-antibody-study-coronavirus-mortality-15283660.php

CFR and IFR are two different things.  CFR is a way to gauge how serious an epidemic is.   IFR is more about making decisions moving forward. That number can also be very regional in it's application.  Keep in mind that Santa Clara County was especially hard hit, so likely the number of infections is higher there than other counties.

Also, keep in mind that the IFR that they have put forward is fairly inconsistent with the IFR from a cruise ship or another part of the US.  Not sure that these two studies alone can be extrapolated in a meaningful way yet.  There needs to be a lot more of them before we start to see a clearer picture.

So, if we split the difference and assume a .25% IFR and ultimately 60% of the US population is infected, we get something like 450,000 deaths?  Is my math correct?  Something is not adding up to me. We already have more than 90 000 and the seroprevalence is nowhere near 25%.  Maybe I am too tired to math tonight.

Edit:  Or, it is too soon to get anything valuable from these figures.

Maybe because the tests just aren't all that accurate yet. Some of them aren't specific enough, so they are getting a lot of false positives.  Many of the test results are coming back positive, but are possibly reacting to antibodies to other viruses.  We need better tests.  Garbage in/Garbage out.