How can we assess the total number of #SARSCoV2 infections that have occurred up to this point? As others have noted, it's not easy to extrapolate from confirmed cases to total infections. Some thoughts here, but no real answers. 1/9
The gold-standard approach here is to conduct serological surveys of the population. These serological assays measure the presence of antibodies in people's blood that can bind to the virus. These responses persist for months rather than ~2-3 weeks for viral shedding. 2/9
This sort of "seroprevalence" study showed that swine flu infected ~35% of the world after 1 year (journals.plos.org/plosone/articl…) and that Zika infected ~60% of individuals in Northeastern Brazil during its epidemic (mbio.asm.org/content/8/6/e0…). 3/9
There are a number of seroprevalence studies underway for #COVID19, including from @Stanford (stanforddaily.com/2020/04/04/sta…) and @PHE_uk (theguardian.com/world/2020/mar…). 4/9
Seroprevalence estimates will help to understand the total number of infections and should be more comparable across regions, as a cross-sectional serological survey should be closer to a random sample as opposed to testing which has varied in capacity. 5/9
However, at this point, I'm not aware of any seroprevalence results and so we're left modeling total number of infections based on confirmed cases and deaths (perhaps supplemented by viral genetics). 6/9
This report by @MRC_Outbreak (imperial.ac.uk/mrc-global-inf…), estimates total number of infections in different European countries based on observed cases and deaths, generally estimating between 1% and 15% of the population infected. 7/9 
These numbers fall well short of "herd-immunity", which would require between 50% (for R0 of 2) and 66% (for R0 of 3) of the population recovered to prevent epidemic spread. 8/9
The arrival of seroprevalence estimates should help to dial in a lot of our numbers. By learning total number of infections we can better estimate population-level severity. Hopefully data from these studies will start to emerge soon. 9/9
