For adults, two well designed randomized trials have shown that face masks have only a marginal or no effect at reducing Covid-19 transmission. The Bangladesh study that randomized villages showed a reduction somewhere between 0 and 18 percent [1], while the DANMASK-19 study that randomized individuals showed an effect size between a 46 percent reduction and a 23 percent increase [2]. This is similar to other respiratory diseases [3,4].

There is much less data available for children, with no randomized mask studies for Covid disease, so it is important to conduct observational studies. That is what Sood et al have done in this well described study, using a natural quasi-experiment from two neighboring school districts in North Dakota [5].

For children, one of the best observational mask studies used another natural quasi-experiment in Catalonia, where Coma et al. compared 5-year-old children that were not masked with 6-year-old children that were [6]. As with adults, they found no major benefit on Covid infections, with a marginal increased infection risk for the masked 6-year-olds of between 8 and 22 percent. That is in the same ballpark as the current study by Sood et al., who report an effect size somewhere between an 8 percent decrease and a 7 percent increase. 

A key issue in these four studies is the effective sample size. Since the Danes randomized individuals, the sample size was the number of individuals, and the analysis was done at the level of individuals. As a contrast, since they randomized villages in Bangladesh, that analysis was appropriately performed at the village level. For a variety of reasons and just by chance, some villages will have more Covid infections than other villages during the study period, so individuals in the same village are not independent observations. This means that the effective sample size is the number of villages.

In the Catalonian study, the 5 and 6-year-olds are mixed geographically, living in the same neighborhoods. If the disease is more widely present in one neighborhood than another during the study period, that should affect the 5^th and 6^th graders equally. Hence, the effective sample size is the number of children, and the analysis should be conducted at the level of individuals. With higher Covid incidence for older children, age is a confounder, but the difference between the 5 and 6-years olds was about the same as the differences found between the 4 and 5-years-olds, who did not wear masks, and between the 6 and 7-year-olds, who all wore masks.

What about the current study in North Dakota? The two adjacent school districts are in the western versus central part of the Fargo-Moorhead Metropolitan Area. Hence, the children are not geographically mixed as they were in the Catalonian study. When an infectious disease spreads through a city, it will typically affect one area at first, after which it moves on to other neighborhoods. One could therefore argue that the effective sample size is the number of school districts rather than the number of individuals, and that the analysis should be conducted at the level of school district rather than individuals.

Having adjacent school districts is better than using two school districts far apart, but the ideal would have been to use multiple pairs of neighboring school districts and to analyze the data at the level of school districts. With a sample size of only two school districts, such an analysis would have been too small to make any precise conclusions either way.

For an analysis at the individual level, geography is a plausible confounder. The null result of no effect could either be because masks have no or limited effect, as shown by other studies. Alternatively, it could be because mask wearing Fargo schools had more Covid circulating during the study period and that masks were effective at bringing the incidence down among school children, with the two effects canceling each other.

What we can conclude from this study is that wearing face masks at school does not provide 100% protection against Covid infections in children. If that was the case, there would have been very few Covid infections in the Fargo school district during the mask-mandate period. For more precise conclusions, we must rely on other observational and randomized studies.

1. Abaluck J, Kwong LH, Styczynski A, Haque A, Kabir MA, Bates-Jefferys E, Crawford E, Benjamin-Chung J, Raihan S, Rahman S, Benhachmi S. Impact of community masking on COVID-19: a cluster-randomized trial in Bangladesh. Science, 375, 2022.

2. Bundgaard H, Bundgaard JS, Raaschou-Pedersen DET, von Buchwald C, Todsen T, Norsk JB, Pries-Heje MM, Vissing CR, Nielsen PB, Winsløw UC, Fogh K, Hasselbalch R, Kristensen JH, Ringgaard A, Porsborg Andersen M, Goecke NB, Trebbien R, Skovgaard K, Benfield T, Ullum H, Torp-Pedersen C, Iversen K. Effectiveness of Adding a Mask Recommendation to Other Public Health Measures to Prevent SARS-CoV-2 Infection in Danish Mask Wearers : A Randomized Controlled Trial. Ann Intern Med.,174:335-343, 2021.

3. Jefferson T, Dooley L, Ferroni E, Al-Ansary LA, van Driel ML, Bawazeer GA, Jones MA, Hoffmann TC, Clark J, Beller EM, Glasziou PP, Conly JM. Physical interventions to interrupt or reduce the spread of respiratory viruses. Cochrane Database Systematic Review, Jan 30, 2023.

4. Liu IT, Prasad V, Darrow JJ. Evidence for Community Cloth Face Masking to Limit the Spread of SARS-CoV-2: A Critical Review, Cato Institute, Working Paper #64, 2021.

5. Sood N, Heick S, Stevenson J, Høeg TB. School Mask Mandates and Student SARS-CoV-2 Infections: Evidence from a Natural Experiment of Neighboring K-12 Districts in North Dakota, USA. Jurnal of the Academy of Public Health, 2025.

6. Coma E, Català M, Méndez-Boo L, Alonso S, Hermosilla E, Alvarez-Lacalle E, Pino D, Medina M, Asso L, Gatell A, Bassat Q, Mas A, Soriano-Arandes A, Fina Avilés F, Prats C. Unravelling the role of the mandatory use of face covering masks for the control of SARS-CoV-2 in schools: a quasi-experimental study nested in a population-based cohort in Catalonia (Spain). Arch Dis Child. 108:131-136, 2023.

This paper presents and analyzes a natural experiment on face masking as an anti-covid measure. It is a welcome addition to the literature on covid. The authors’ conclusions are supported by the data they present.

Natural experiments have been a cornerstone of public health research since John Snow’s "On the mode of communication of cholera," which examined cholera mortality in Victorian London, by water source. The key to Snow’s study is that water sources changed in a way that was so haphazard that it was essentially random, and was blind to the unwitting participants. This paper follows in that long tradition.

The authors follow covid incidence in two adjacent school districts in North Dakota in the 2021-22 school year. Crucially, the two districts are very similar in social and demographic characteristics, and differed principally in mask policy. The Fargo Public School District did not have a masking policy, while the West Fargo Public School District did, until 17 January 2022, when it switched to masking optional.  There was no statistical difference in covid incidence between the school districts.  The authors conclude that there is no association between mask policy and covid case rates in this natural experiment.

First, some comments particular to this study, then I will add some context. The authors are circumspect in their conclusions and statements. They never claim that masking is not effective in any setting or for any purpose. The paper is about covid incidence in two school districts, and the authors stay close to the evidence. As I have already noted, I believe the authors’ conclusions agree with their data. To put it in plain English: the mask mandate in the Fargo School District didn’t buy its pupils any protection against covid compared to the neighboring West Fargo Public School District.

One question I had was that in table 1, the race/ethnic composition percentages total 96% and 95%; I assume this is some combination of rounding by category and the fact that Native American is not enumerated. More importantly, I wonder if there might be differences in case detection across the two districts. The study usefully points out that both districts used the same covid tests, which removes one possible mechanism of difference. However, the crux of the matter here is that natural experiments are not drop-in replacements for randomized experiments. The error term may be correlated with the exposure, namely the two districts self-selected into quite different masking policies, and may have had different testing avidity. They also had different return-to-school policies (cf. authors’ table 1). This is one area in which this study differs from Snow’s essentially random study, as mentioned above. Also, maskers are never blind to their own masking.  Frankly, I don’t think such phenomena drives these results, but it’s worth noting that it may have contributed. I think school mask mandates in the real world, with actual masks as used by real people, just didn’t work as well as it had been hoped when the mandates were planned.

Big picture, what should we extrapolate from these findings? The authors don’t comment much, but I am at liberty to do so. The next pandemic of respiratory infectious disease (may it never arrive!) could have different transmission parameters than covid did in academic year 2021-22 in North Dakota. As such, saying that masks never work, based on this paper, would be throwing the baby out with the bathwater. To be clear, the authors don’t say this; I’m setting up a deliberate strawman. Nonetheless, I think the public health community should keep in mind this and other studies as it draws up plans for dealing with future pandemics. Different pandemics are, well, different, and perhaps school masking can help stem the tide in the next pandemic. However, and this is key, such a decision will have to be made in light of a prior that it did not work in covid in 2021-22. Thus, the stakes will have to be even higher, or the certainty greater that something is different.

Do masks work, at all?  For example, will wearing masks when attending an indoor sporting event protect the wearer, on a one-off basis? The present study doesn’t put that to the test. These are a suite of a questions which will likely require a suite of studies to answer, with each study being tailored to the applications. Natural experiments, as this one, may be harder to find for every application.

Covid is still with us; it is endemic now; people can and do argue about what endemicity means but I think it’s noncontroversial to say that covid is endemic. Should we mandate mask in schools now? Well, no; per above. Moreover, masks were meant as temporary crisis management during the pandemic phase. One thing we can do is improve indoor air quality; this should be subject to experimentation as well, and is amenable to randomized experiment (schools get retrofitted for air improvements, or not, at random). In the wake of John Snow’s work, drinking water supplies were cleaned up; civil engineering projects at enormous monetary cost. It remains to be seen if the covid pandemic will make any lasting changes to indoor air quality, but asking schoolchildren to don masks at the sign of the next pandemic should not be counted on to save the day.