In talking through Thanksgiving I have said a number of times that the choices are (1) anxious (if you choose to see people) or (2) sad (if you do not). We pulled the trigger on sad. So, I’m sad. But finally deciding did have the benefit of making me less anxious. I hope whatever you have landed on to do for the holiday you’re feeling okay about it.

Confession: I have never made a traditional Thanksgiving dinner. The first time we missed Thanksgiving with my family, Jesse and I went to the Wisconsin Dells with our best friends and ate Thanksgiving dinner at Cracker Barrel. No joke. My kids are insisting on a typical Thanksgiving, though, so now I have to figure out how to cook a turkey (I was going to try this, but I’m open to suggestions).

I wanted to write today about masks; in particular about that new Danish study which many people seem to be hearing about from their mask-skeptical relatives.

I do not, here, want to go through all of the detailed evidence we have on how masks work. COVID-Explained has a long explainer. There is a lot of interesting scientific evidence around the ways masks prevent spit spray, the varying efficacy across materials, and so on. The bottom line there is that masks prevent spread of aerosols.

(There are other reasons masks might work, notably preventing face and mouth touching. Remember before COVID-19 when you’d sometimes lick your finger to be able to open those tear-off vegetable bags at the grocery store? Admit it. And think about it: no one does this anymore.)

What I want to focus on today is the issue of mask effectiveness in the real world. Theoretical and realized effectiveness are closely linked but not exactly identical, since mask effectiveness in the real world also depends on behavior.

We have a lot of evidence that people wearing masks is correlated with less COVID. For example, there is this paper showing slower COVID-19 growth rates early in the pandemic in US counties which adopted mask mandates. Or this one, showing a similar effect of indoor mask mandates in Canada. Or this more specific analysis showing lower COVID growth in states which introduced mask mandates before opening indoor dining, versus only after. These population level links could reflect masks protecting the user, or protecting others, but the bottom line is that more mask-wearing is correlated with less COVID-19.

The mask-skeptic counterargument should be obvious to those who think a lot about patterns of correlation and causation. How do we know it is masks that matter, and not other behavior differences? Places where masks are mandated are also more likely to limit indoor dining and gyms, for example, and likely to have residents who are more careful in other areas (the NYT, this weekend, had a stark description of variations across the US in approaches to the virus.)

However: this skepticism must be tempered by the basic science of masks. I spend a lot of this newsletter complaining about correlation versus causation. Behind many of my complaints, though, is the recognition that there is no reason to think there would be a causal link. Eating chia seeds is correlated with being in better health, but there is really no underlying biological reason to think that link is causal. So I am very, very skeptical.

The same is not true of masks! There is a lot of reason to think they prevent the spread of COVID-19. The combination of the observational evidence with the basic biological plausibility makes this much more compelling than, say, chia seeds.

But: of course we would always prefer randomized evidence. Which is why people were so excited about the idea of this randomized trial in Denmark, designed to evaluate mask-wearing. The paper, published in the Annals of Internal Medicine, recruited a sample of 6000 people, of whom about 5000 ended up completing the trial. Half of the sample were assigned to the treatment group, and encouraged to wear masks when out of their home and around other people. The other half were control, and didn’t receive this explicit encouragement.

The authors followed the participants over April and May 2020 and observed COVID infections in the two groups. The group who was encouraged to wear masks had an infection rate of 1.8%; the control group was 2.1%. This is a lower rate in the mask group, but not “statistically significant” — basically, the authors cannot reject the hypothesis that there is no difference.

Obviously, this has been taken by some to suggest that masks are not helpful. However, this is a significant over-reach of these results for a number of reasons.

  • Most obviously, the study doesn’t talk about risks to others. If we think about masking policy it could be a good idea if masks protect the wearer or other people. That isn’t captured at all in this study, so the best evidence is still what we have from the policies on masking (which suggests they work).
  • But almost as important: this study is very “under-powered” in statistical speak. Basically, the sample size isn’t large enough to draw any strong conclusions given the relatively low rates of COVID-19 in this population. As the authors note in the abstract, their results are consistent with masks reducing infections up to 46% (or increasing them up to 23%). If masks did lower infection rates by 46% — which, again, would be consistent with these data — that would be incredibly good. Yes, it’s not as good as the vaccine, but it’s also just a piece of paper or cloth you put on your face.
  • A final issue is adherence. In a sense, the personal question you may want the answer to is whether strict adherence to mask-wearing would lower your infection risk. Only 46% of the treatment group in this study reported strict adherence (another 47% reported “predominant” adherence). This low adherence adds to the noise in the study; if we try to estimate the impact of strict mask wearing, we’ll get even wider bounds — even less confidence in any particular estimate.

Where does this leave us? It isn’t that we learn nothing from this study. If you expected mask use to completely eliminate the possibility of infection, that is rejected by the data here. But within the range of what most people would have expected, this represents fairly limited new information.

The fact is, the vast majority of the evidence either explicitly indicates that masks work or is consistent the view that they do. Wearing a mask may also have benefits in terms of reminding us to be careful in other ways.

Can we go too far with mask-shaming? Yes. There is no need for dirty looks at the solo, carefully distancing, runner out at 5:30 in the morning in 25 degree weather for not having a mask over their face (yes, this happens, trust me). And I think we may do a disservice to the mask cause by taking it too far and causing people to dismiss the call for masks as crazy.

But all of our evidence suggests that, especially inside, masks are key. Stores, schools, restaurants (if you must), meetings, seeing family. The vaccine is coming. But it isn’t here yet. For now, put some cloth on your face.

And Happy Thanksgiving.