Above is a figure from a recent study published by the American government’s CDC (Centers for Disease Control and Prevention)*. Essentially, Andreko and coworkers studied two populations in California: one of 600 people who had just tested positive for COVID-19, and one of 1,100 who had not. They asked both sets of people about mask wearing while in public settings indoors (i.e., shops, workplaces, etc). Then they got the risk reduction from the numbers of mask wearers in the two populations. The idea being that if masks work you would expect a higher fraction of the 1,100 who don’t have COVD-19 to wear masks, than the 600 people who do. The way to read the figure is that they estimate that wearing an N95/KN95 (similar to FFP2, KF94, …) while indoors reduces your risk of catching COVID-19 by 83%.
It is good to see this study, but it has advantages and disadvantages, as the authors know. The advantage is that it looks directly at what we are interested in: mask use in real life. As opposed to in a lab. One disadvantage is that basically what they are measuring is the reduction of risk of, for example, people who wear N95 masks, which is not quite the same as the reduction in risk due to wearing an N95 mask. For example, if people who wear a N95 mask indoors are also more careful about social distancing, than people who do not wear a mask, that could also be a factor in the reduction of risk estimated here, even though this would not be due to the mask. In this way, the study could overestimate the benefits of wearing a mask.
Another disadvantage is due to the relatively small number of participants. This translates into large uncertainties in the estimates. The image above does not show the error bars – which are large. For example, they quote the 95% confidence interval (CI) for the reduction in risk from wearing a cloth mask as 0.17 to 1.07. A 95% CI means they are 95% confident the true reduction in risk is somewhere in the region 0.17 to 1.07. This is a big range, and it includes 1.0 – which is no reduction at all. So the study has too few participants to strongly rule out cloth masks having little or no effect.
But the CI for the good masks (N95 etc) is a risk reduction of 0.05–0.64, i.e., data suggest a 95% chance of the risk when wearing a N95 mask being in the range 5 and 64% of the risk without a mask. Or in other words the risk is reduced by a factor between twenty and one third – again a large range. This is just an uncertainty due to the small-ish population size, it does not include any possible bias due to wearers of N95 etc mask social distancing more (or less) carefully.
It would be nice to have additional, bigger, studies, but it is good to see this study. We know that N95/KN94/FFP2 etc masks filter 90%+ of the size of particles that carry COVID-19 from lab studies, and we now have increasing amounts of direct evidence that these masks work in practice and reduce the risk of infection. The two types of study complement each other, and paint a pretty compelling picture that we can reduce COVID-19 transmission by wearing masks indoors.
* You may be curious as to why the Centers of Disease Control and Prevention have the initials CDC. I think it is because Prevention is a later addition to the title, and most people call them CDC so they stuck with that.
Chance & Necessity 