More on the Stanford Antibody Study

Readers know that for some time I have been frustrated by the lack of effort put into antibody testing, believing that it was more important than infection testing in terms of understanding the true scale of the epidemic and its clinical consequences for the population.    (Medrxiv Paper)   I am repeating some of the summary and comments made yesterday because the significance of these results cannot be overstated, even with appropriate cautions for generalizability.  The authors in the paper, who are all highly respected researchers from Stanford or other prestigious institutions, first lay out some information on the early course of the epidemic and some of the hysterical modeling results that were published and drove policy-making.  Next the researchers go over the usual issues in regard to determining an accurate number of infections early in an epidemic.  Santa Clara County had some of the earliest cases in California and the researchers on April 3 and 4 conducted an antibody testing study in the County.  At this point there were around 1000 cases in the County at this time.  3,300 adults and children were tested.  They used Facebook ads for recruitment and attempted to balance where participants came from to achieve proper representation of the entire population in the County.  They also re-weighted results from actual participants to represent the actual geographic, gender and ethnic distribution in the county.  They also applied estimates regarding the likely accuracy of the testing method.

50 people tested positive for antibodies in the raw numbers, a prevalence of 1.5%.  After the adjustments were applied, the rate was between 2.49% and 4.16%.  Based on the population of the county, this implies that between 48,000 and 81,000 people had been infected on April 1, compared with the 956 persons on that date found by actual infection testing, or at least 50 times lower.  Based on their estimates of deaths in the County, using conservative assumptions, the actual infection fatality rate was between .12% and .2%.  And other factors, as explored in the future, would likely lower that rate.  It is important to note that all of these individuals identified as having been infected must have been asymptomatic or had mild illness, so the ratios of people obtaining serious illness need to adjusted dramatically downward in models as well.  The most important implication for policy-makers is that current social distancing policies may not be the best approach, especially given the huge economic costs, but also in terms of the health of the population as a whole.  A better approach may be to relax social distancing enough to ensure a controlled exposure to the virus among non-vulnerable populations, which would lead to very few serious illnesses and would dramatically slow transmissibility of the virus.