@docslotnick, I saw that study. The problem with that study is that if the incidence of COVID-19 is fairly low (say 0.5%), then a 98% specificity (as this antibody test, which has just been approved, has in the best case, according to their own data) means that the vast majority of positives you have are false positive. WE actually don’t know if the test’s specificity is that good all the time; it’s not FDA approved, which may not be a huge issue, but to validate the test, they conducted a few hundred samples .Not sure that’s enough to feel confident about the results.
Example: Imagine 1,000 people. 5 truly have COVID. But 20 will test positive for COVID.
A 0.12% mortality rate is not consistent with data from NYC, in which already more than 0.1% of the population has died (city population = 8.2 million, deaths = ,8843, presumed COVID deaths = 12,000ish). And the few antibody tests done in NYC suggest around 15% prevalence there. Now, NYC’s health is poorer than Santa Clara’s, on average, but I’m having a really hard time squaring the assumed seroprevalence they’re finding with the actual, on the ground mortality rates we’re seeing in hard-hit areas.
There’s a bigger problem i my mind. This sample was drawn from a population of people solicited via Facebook ads. I can tell you, as someone who rushed to sign up for an NIH antibody study, that I did so because I had some extremely suspicious symptoms in February (fever of 102 with dry cough for 5 days, followed by what I thought was improvement, followed by a week or more of worsening cough, followed by 3 weeks of chest pain). Five days earlier I had flown on a flight, and it was the last day flights from China were being allowed into SFO. People from China were on my flight. But because I didn’t meet testing criteria at the time, no one even considered the possibility that it was COVID. Anyways, the point is, my guess is that the 3,300 people they sampled were so eager to go through drive-through testing, tugging their kids in tow, because they suspected they had the virus already. Which makes it not representative of overall prevalence.
Believe Wuhan numbers or not, but they found seroprevalence amongst healthcare workers of about 2-3%. In a German town with a large outbreak, they found seroprevalence of 14%, and in pregnant women in NYC with no covid symptoms, they also found about that seroprevalence. Extrapolating from Santa Clara’s number of symptomatic cases/true infection rate would mean that ALL of NEw York had the disease, and that just doesn’t match data.
Anyways, I think it’s interesting and we need more testing like that. But I don’t think it can tell us that much quantitatively about fatality rates yet. My personal guess is that the true IFR will wind up being somewhere between 0.3 and 1%, depending on the population. But if it’s on the lower end, it’s much more transmissible, which means it’s going to hit almost everyone in the population, and total deaths could be almost as high as if you had a higher fatality rate but a lower attack rate.
Either way you slice the data, a lot of people are going to die.