(Especially @anon29537550, have a look at this.)
@all:
That’s a meta-analysis. They included 66 studies in their analysis in regard to asymptomatic cases, 20 studies in the analysis for the Secondary Attack Rate, 18 in the analysis of the Serial Interval. Results look robust to me on a first assessment.
Bottom line takeaway: This basically confirms most of the stuff we already know.
One important detail takeaway: this meta-analysis estimates, across studies from very different settings across the globe, an overall percentage of 8.4% of asymptomatic cases. That, sadly, fits the results we so far got from larger antibody studies (to the best of my knowledge).
To translate into understandable numbers (and I know this is much more complicated, so please don’t anyone start arguing about my take, ok?): if we know that about 90 people in our town have been sick with Covid-19 during the last months, we could very roughly estimate that about 100 actually had it. Not much more. Very rough hot take: even if all of them survived, there is a long, long way to herd immunity.
Another finding: from the moment someone gets symptoms, we have about 2.8 days to put all of their contacts into quarantine. Full quarantine, that is. Otherwise, we will have secondary infections, and the chain of infections goes on. (Just FTR, that also old news.) Takeaway: If we want to get rid of local clusters, we will have to do massive contact tracing efforts.
Data on children isn’t good, they do report findings but I would be very cautious with that, hence I don’t repeat those here.
Looking at the methods, I don’t find any strong flaws. Disclaimer: while I took part in meta-analysis research, this is not my specialty, and I just browsed this paper. I have some grievances, some of which can be overcome while others are inherent (i.e., I can’t really claim I see no publication bias, so I expect publication bias - but then, this part is generally difficult, and we are talking about so vastly different circumstances in the statistical and human populations here that I feel you can’t do perfect here). I expect this study to be published research, maybe even fast-tracked. As usual, I have the feeling some editing for getting rid of any potential ambiguity of language would be great.