This information has come out today about ways to stop the spread
The general public isn't going to wear eye protection? And would wearing glasses not do the same thing?
Had a gander at the paper. It's a review paper, so they've collected up evidence from 172 other papers and calculated the probability of infection based on all of those studies. There's more than 25000 patients in the sample so actually not a small study compared to others I've seen. I think it's more to do with knowing which preventative measure works best for say, hospitals since a lot of datapoints come from health care settings. I think a small problem with this review is that includes data from SARS (the first one) and MERS where transmission rate might be different.
Note: Certainty of evidence - Low. In other words, just more scientists trying to justify their salaries feeling they have to publish something. Also, why is the without intervention statistic different each time? I'll give you a reliable statistic - 80% of the statistics around covid are made up bullshit.
We're not all like that, but yes, there are some who try to push out papers quickly in order to apply for grants, especially now with lots of money floating around for covid-19. A problem is that people will latch onto papers which are in pre-print which will probably get pulled or rejected and would usually put their own bias on the statistics presented.
As with this review, it looks like quite a lot of work went into it. It's like a summary of 172 other papers. They've graded them, collated the data and reprocessed it for their own calculation. The difference in "without intervention statistics" is because the samples the three categories differ. As an example, 29 studies went into looking at physical distancing measures and only 13 studies looked at wearing face masks. You then have studies which looked at combinations of measures. I think if the sample size was a lot bigger, the values would converge to 13%, seeing as the physical distancing measure has the largest number of data points.
Their estimate of the certainty is more to do with the value of reduction I think and not to do with whether these measures have a positive effect or not. It's essentially a qualitative error bar. Even so, "low" is like 2/4 on their scale (from their supplementary materials). There are two categories below that. So I guess moderate (3/4) is actually fairly good, especially for biology where there are so many variables to consider. The information doesn't control from a controlled environment either which adds to the problem.