Update (March 16): A Morbidity and Mortality Weekly Report released by the Centers for Disease Control and Prevention on March 11 found that Arkansas school districts with universal mask requirements last fall had a 23 percent lower incidence of COVID-19 among both staff and students than districts without mask requirements.

As the Omicron wave subsides in much of the US, and the world prepares to mark the two-year anniversary of World Health Organization’s declaration of COVID-19 as a pandemic, schools are grappling with whether children can safely take off their masks. Some states, including New Jersey, Connecticut, Massachusetts, Delaware, New Mexico, New York, Maryland, and Oregon, have announced that they will lift mandates for students in the coming weeks, leaving masking decisions to individual districts or parents. Others are waiting. On Friday afternoon, the Centers for Disease Control and Prevention (CDC) updated its masking guidance, recommending masking indoors (including in schools) only in high-risk areas for most people. 

School mask requirements remain a contentious issue; late last month, for example, an advocacy group called the Urgency of Normal (UoN), which has on its roster multiple doctors and researchers, put together a toolkit for parents and schools advocating for keeping schools open regardless of case numbers, restoring pre-pandemic activities for children, and removing most COVID-19 precautions in schools, including mandatory masking. The toolkit claims that studies don’t clearly show that masks on kids reduce COVID-19 transmission in schools, stating, “Student masking has no scientifically established benefit in real-world use.” A spate of recent opinion pieces and articles also criticize mask mandates for kids. 

But many other experts strongly disagree that masks don’t reduce COVID-19 transmission in schools: they say studies are clear that, while imperfect, masks—particularly surgical masks or higher-filtration options such as KN95s—work very well and that schools have been relatively safe, in large part, because kids have been masked. “They’re trying to tell people that masks don’t matter,” but that’s not true, says Julie Swann, an industrial and systems engineer at North Carolina State University who models the implications of various COVID-19 policies. 

The ABC Collaborative, which is led by the Duke University School of Medicine and the University of North Carolina School of Medicine and receives funding from the National Institutes of Health, has worked with schools throughout the pandemic to monitor COVID-19 transmission, study what mitigation measures work well or are not necessary, and help schools make decisions based on data to keep students and staff safe. Its own toolkit, published this month, summarizes the research, including new data from a prospective, nationwide study on mask-mandatory versus mask-optional school districts which found that mask mandates are associated with significantly reduced COVID-19 transmission in schools (the study has been peer-reviewed and provisionally accepted for publication). [Editor’s note: this study has now been published in Pediatrics.] It also provides guidance for districts on determining the right time for their communities to remove masks. Yet another toolkit, put out by a group called Urgency of Equity, also presents evidence that universal masking works. 

As the debate continues, here’s a look at what studies have shown about masks’ effectiveness in curbing SARS-CoV-2 transmission in schools, what this means for when kids can take them off, and where points of contention remain.   

How to study masking in schools

Mask mandate removal advocates often suggest that randomized controlled trials are needed to determine whether universal masking reduces COVID-19’s spread in schools. But directing a group of children not to wear masks in order to serve as a control group isn’t realistic, experts say. “Ethically, you really can’t randomize. . . . You would be hard-pressed to get a masking trial through a reputable institutional review board, let alone get schools and parents to sign a consent for that,” says Danny Benjamin of Duke University, the principal investigator and chair of the National Institute of Child Health and Human Development’s Pediatric Trials Network and the ABC Collaborative. Similarly, “we don’t ask a group of physicians to randomize to whether they’re going to wash hands in between patients, to say that hand-washing prevents transmission,” says Benjamin, because “the plurality of evidence, even though there hasn’t been a perfect randomized trial, is that hand washing reduces transmission of infection.” 

You would be hard-pressed to get a masking trial through a reputable institutional review board, let alone get schools and parents to sign a consent for that.

—Danny Benjamin, Duke University

More realistically, researchers could run a prospective cohort study, selecting two schools, school districts, or groups of schools, one with mask mandates and one without, and following them for several months. Ideally, the primary endpoint would be the number of students infected with COVID-19 while in school—so-called secondary transmission. This is key: many kids get COVID-19 outside of school, and these cases have nothing to do with the effectiveness of mask-wearing in school. A good study would control for vaccination rates in the schools and for testing policies between the groups—that is, if one school did more testing, or if the students or teachers had higher rates of vaccination, results could be skewed. Ideally, researchers would also know the percentage of students who chose to wear masks in mask-optional schools, and the percentage of students who wore masks properly in both kinds of schools. It turns out that no single study has yet met all of these criteria. Some, the UoN emphasizes, lack a control group. But according to Thomas Murray, a pediatric infectious disease physician and researcher at Yale New Haven Children’s Hospital, to assess the effectiveness of masking in schools, “what you have to do is you have to look at all the aggregate data” from the many studies of varying quality that have been conducted.  

A consistent trend

Several studies have looked at schools with universal masking and observed how often secondary transmission occurs. One study published in Pediatrics last December showed that 20 North Carolina school districts with masking had very low in-school transmission of the virus: There were only 64 secondary infections out of 2,431 close contacts, a secondary transmission rate of 2.6 percent. On the other hand, 808 kids and staff acquired COVID-19 from the community, meaning that students and staff were 12.4 times more likely to be infected in the community than at school. There was no unmasked control group in this study, and masking isn’t the only difference between the two environments, but it’s a big one. Scientists working in multiple other US states have reported similar results. Separate research groups “doing a combination of contact tracing, testing, and whole genome sequencing, consistently time and again show the same thing: You do universal masking your secondary [transmission] rate is below 5 percent,” says Benjamin. By comparison, a recent meta-analysis shows that household secondary transmission rate is about 21 percent (it ranges from 6.1–51.2 percent), and is 5.9 percent for social events with family and friends. 

Conversely, reported superspreading events suggest that SARS-CoV-2 can in some cases spread readily without masking in schools. An elementary school classroom in Marin County, California in May–June 2021 was well-ventilated, with windows opened, and students all masked and six feet apart. An MMWR report found that days before their teacher tested positive for the Delta variant, she had taken off her mask a few times to read the students a book. Half of the students ended up testing positive for COVID-19 soon afterward. Similarly, in a German school where masks were not required in the classroom, one sick teacher infected 31 students and two staff members. An Israeli school that allowed students to take off their masks for three days during a heat wave subsequently had a COVID-19 outbreak in which two independent cases resulted in 153 students and 25 staff members testing positive. 

At St. Louis University in Missouri, masking was required in classrooms and all public spaces, including labs, libraries, public areas in residence halls, and at all indoor events. Students were allowed to be unmasked in their own dorm room or apartment with their roommates. COVID-19 cases still emerged, as reported in an MMWR study that collected data between January–May 2021, and contact tracers investigated close contacts of each student who tested positive. The researchers found that 7.7 percent of close contacts where both people were masked during their contact tested positive, whereas 32.4 percent of close contacts in which one or both parties were unmasked tested positive; the adjusted odds ratio was 4.9, says lead author Terri Rebmann, an infectious disease epidemiologist at St. Louis University. 

Another study aimed at assessing the effects of masking was published in JAMA Network Open last month; its authors surveyed 6,654 childcare professionals across the country over the course of a year, asking about policies in their programs to reduce COVID-19 transmission, including staggered arrivals and departures and taking temperatures. They found that masking kids was associated with a statistically significant 13 percent lower risk of the program closing due to a COVID-19 case, and child masking “was the [mitigation] that had the most impact,” on this outcome, says study author Murray. 

Yet another study, this one conducted in schools in Germany in which masking policies changed over time and published last December in Frontiers in Public Health, used a multivariate linear regression model to assess the causes of changing secondary infection rates in those schools depending on various mitigation policies. It found that teacher/caregiver and student masking significantly reduced the number of secondary cases. 

Getting a firmer grasp on masking’s effects

“The fundamental challenge that those of us doing high quality research in K–12 education have faced in the effort to compare masked to unmasked within-school transmission rates, is that prior to Fall 2021, schools that did not require masks also did not voluntarily record within-school transmission and share those data,” writes Benjamin in an email to The Scientist

To gather large-scale, prospective data, the ABC Collaborative approached every school district in the country, more than 13,800, asking them to participate in a study on masking; 143 expressed interest, and 85 completed an initial survey. These districts were followed, with 61 submitting infection data on primary and secondary cases. Of these schools, 46 universally masked, 9 had masking policies that changed throughout the year, and 6 were mask-optional. The vaccination rate was similar for masked versus mask-optional districts (it was actually a bit higher in mask-optional schools), and the team adjusted for community rates. Altogether, the study included 1,112,899 in-person students and 158,451 in-person staff. Preliminary results showed a 70–80 percent reduction in secondary transmission in universally-masked compared to mask-optional districts, and the protective effect of masking was stronger after adjusting for community rates, district size and vaccination. In universally-masked districts, 100 cases of COVID-19 acquired outside of school resulted in five cases transmitted in school. In voluntarily-masked districts, 100 cases acquired outside of school resulted in 43 cases transmitted in school. A paper on the study has been reviewed and provisionally accepted at the journal Pediatrics; a summary of the paper is publicly available on the ABC website. 

Many imperfect studies 

The UoN’s toolkit lists five studies it says had proper control groups and were large and occurred during times of higher levels of COVID-19 in the community; the group says these studies don’t show that masks clearly reduce COVID-19 transmission. UoN then lists three additional studies that found that masks were associated with less COVID-19 in schools, but writes that these studies don’t have appropriate control groups, so their findings “can’t tell us anything about whether student masking is effective.” 

Several of these studies, though, don’t home in on secondary transmission within schools. Instead, they look only at sheer numbers of COVID-19 cases detected among students and school staff. Researchers conducted the first of these studies from November 16–December 11, 2020, in 169 K–5 schools in Georgia and reported their findings in the CDC’s Morbidity and Mortality Weekly Report (MMWR). After adjusting for county-level COVID-19 levels, the investigators found that COVID-19 incidence was 37 percent lower in schools that required mask use by adults, and a non–statistically significant 21 percent lower in schools that required masks for students, compared to schools in which mask use was optional. According to the UoN’s toolkit, this lack of statistical significance means that the researchers “couldn’t be sure the benefit was real.” Benjamin says this study is viewed as one of many that support masks’ utility in interrupting transmission, but, since the researchers didn’t contact trace to see whether fewer kids in the schools with mask mandates got COVID-19 while at school, they can’t know the true effect of the masks. He says, “if you want to assess within-school transmission, you absolutely must measure within-school transmission.” 

The UoN also cites an unpublished, preliminary analysis from the UK conducted over several weeks in October 2021, in which the researchers compared student absences due to COVID-19 in schools that mandated face masks to those that did not. Absence rates in both groups decreased over this time period, with absence rates in the masked group falling slightly faster, though the difference was nonsignificant. Another study, reported in a May 2021 preprint and not yet peer-reviewed and published, did not find a correlation between mask mandates and COVID-19 rates among students in Florida during the 2020–2021 school year. But again, both studies’ primary weakness is that they examine total numbers of students who got COVID-19 or were absent because of COVID-19, not secondary cases contracted within school. 

A nationwide MMWR report looked at US counties with or without school mask mandates from July 1–September 4, 2021. It found that counties without the requirements had significantly larger increases in child COVID-19 cases (from the Delta variant) after school started than those with required masking. The UoN writes that this study didn’t look at vaccination rates, which tend to be higher in counties that mask. However, the researchers did adjust for pediatric vaccination rates and community transmission levels, though they could not control for teacher vaccination rates and school testing data. Again, this did not examine within-school transmission, only total cases. “This study is similar to most of the studies that the anti-mask group has quoted,” says Benjamin. “Put simply, it is weakly supportive of masking helping, but it is an ecologic study and does not have within-school transmission as an endpoint.” 

UoN cofounder Scott Balsitis, a viral immunologist at California-based biopharmaceutical company Gilead Sciences (which is not connected to his advocacy work), also points to an as-yet unpublished study that Tracy Beth Høeg is running currently in North Dakota. Høeg, a physician epidemiologist and UoN co-founder, is looking at two similarly sized school districts in the same municipality, one that started the school year with a mask mandate and one that didn’t (the mask-mandatory school switched to optional masking on January 17, allowing a crossover design). Balsitis writes in an email that he thinks it’s a well-designed study because it's prospective, it focuses on one community to minimize the effect of other variables, and it has a crossover design to expose effects of the masks over confounding variables. The data hasn’t yet been peer-reviewed or published, but Høeg says preliminary results show that student case rates did not seem to be higher in the mask-optional district. It’s unclear how much of a role confounding factors may play (the team is still gathering data on them in the two districts), and most importantly, the study did not look at within-school transmission. 

Homing in on within-school transmission

Only three of the eight studies UoN listed looked at within-school transmission of COVID-19: One study from Spain wasn’t designed to analyze masking, but rather how children’s ages might affect secondary transmission. Researchers looked at data on more than 1 million students inside bubble groups in schools in Catalonia, Spain, from September to December 2020. The schools created bubble groups of 20–30 students and teachers who stayed together all school day long, away from other school members; if one of them tested positive for COVID-19, the entire group tested and then quarantined at home for 10 days. The UoN toolkit points out that kids aged 3–5 did not have to wear masks in schools, whereas kids aged 6 or older did, and they write, “In-school COVID transmission was the same in 4–5 year olds where masking was not used and in 6–7 year olds where masking was required.” Yet it is not known how many kids ages 3–5 masked voluntarily, or the effects of any other confounding factors that may have been different among the various age groups. 

Another was a preprint examining eight masked school districts in Massachusetts during 2020–2021, which the group says did not have an appropriate control group because all students had to abide by the state’s mask mandate. It found that 29 of 1,327 close contacts tested positive (2.2 percent of close contacts became infected). It also reported that the secondary transmission rate was significantly higher—11.7 percent—during the rare exposures in which both parties were unmasked, with 28 infected people who had 67 unmasked close contacts causing seven within-school transmissions. The UoN cites a line from the paper that states, “all reported classroom exposures were masked, so these results do not directly inform the impact of masking within classrooms.” However, students come together in school in places other than the classroom; and the paper indicates that any unmasked time at school is associated with increased spread. 

A study that has drawn much criticism from UoN and others who oppose mandatory school masking was published in MMWR last September and looked at schools with and without mandatory masking policies in two Arizona counties in July and August 2021, when the Delta variant predominated. It found that schools without a mask mandate were 3.5 times more likely to have a COVID-19 outbreak than schools with one. Contact tracing was used to determine whether cases were linked to in-school transmission, and an outbreak was defined as two or more laboratory-confirmed COVID-19 cases with onsets within a two-week period that were epidemiologically linked and not connected to each other outside of school.   

See “Shrinking Quarantine” 

Critics note that the researchers behind that study didn’t control for vaccination rates in children and teachers. Coauthors J. Mac McCullough, a public health economist at Arizona State University, and Megan Jehn, also at ASU, say they weren’t able to determine vaccination rates in the schools because that data is private in Arizona, but they did control for community transmission levels, which McCullough says are driven in part by the community vaccination rate. He adds that in elementary schools, where no child was old enough to be vaccinated at the time, “we still found an association between lack of mask requirements and more outbreaks.” These data weren’t in the original paper because of space considerations, says McCullough. “The reality is, the data showed that there were more outbreaks in schools that did not require students and faculty to be masked.” Some critics question whether schools in the two groups were open for the same amounts of time. The researchers say the median difference, two days, is too small to matter. A difference in testing policies between masked and mask-optional schools, if one existed, could also have influenced the results. 

Based on the studies the UoN cites and the in-progress North Dakota study, Balsitis says, “If there is an effect of masking students, it’s too small to confidently quantify. For that reason I feel good about taking the masks off of kids.” He says that researchers “for the most part have not done good studies to understand truly what the effects of two years of near-constant masking of children in some communities really is. So I think the burden of proof properly belongs with those who want to not only use, but make mandatory, an intervention.” In a follow-up email, he writes that vulnerable people should check with their doctor about how to properly use N95-type respirators, which have been shown to effectively filter the virus both in droplets and in tiny aerosols (see “The physical science of masking,” below). UoN argues that having vulnerable children wear N95-type masks is more effective than universal masking with lower-quality masks. 

Still, the studies UoN cite represent a fraction of the work that has been published on masking in schools. Benjamin and others argue that while control groups are certainly useful for understanding the cause of reduced transmission in schools, overall trends in research findings convincingly point to masks’ effectiveness even without a control group. “The antimask crowd neglects to mention the consistency of the findings across time and across research groups, using different research methods to show how much masking in school helps relative to what is happening in the community,” he says. 

When to unmask?

Even though many experts who spoke with The Scientist say that masks reduce transmission in schools, they all agree that masks do need to come off at some point. There’s less agreement on when that should be. Monica Gandhi, an infectious disease physician at the University of California, San Francisco, used to be part of the UoN, but parted with the group and has since written that we need specific metrics for lifting mask mandates in schools, such as a COVID-19 hospitalization rate of less than 10/100,000 in the region per day, or at least 15 percent remaining ICU capacity based on a tool the US Department of Health and Human Services has created. Others have made similar suggestions. “There’s enough population-level transmission studies and the biological plausibility that the masks work that I think having metrics to remove masks” is prudent, she says. 

Swann’s model, which has not yet undergone peer review, indicates that removing masks in schools could lead to a 38.1–47.0 percent increase in infections for ages 5–9, a 27.6–36.2 percent increase for ages 10–19, and a 15.9–19.7 percent increase for the total population. She uses laboratory-based estimates that masks reduce COVID-19 transmission by half, as well as a simulated social network based on real demographics of North Carolina residents. Communities shouldn’t take masks off, she concludes, until their health care systems can handle these potential increases. “The biggest risk in February 2022 is the hospitals that have too many people,” says Swann. “When that happens, mortality increases for all causes.” Swann says that communities would need high vaccination rates among both kids and adults to mitigate this increase (currently, just 25 percent of children ages 5–11 and 57 percent of children ages 12–17 are fully vaccinated nationwide). 

“In terms of the decision to normalize [lift mask mandates in schools], unfortunately there’s not a magic number,” says Murray. “It’s going to be a combination of different things, like community transmission levels, rates of vaccination in the population, ability to get access to rapid testing. If we had super high vaccination rates among our younger population, it would certainly facilitate getting back to normal faster, not only because of the protection that they would have and the reduced transmission, but also because it’s no longer a reservoir for the virus to replicate and select for new variants.” 

The characterization of COVID-19 as similar to the common cold or influenza in children—I just don’t think that’s entirely accurate.

—Thomas Murray, Yale New Haven Children’s Hospital

The ABC Collaborative’s toolkit, which is based on its North Carolina findings and its new nationwide study, includes a chart where one can plug in the number of primary cases entering the schools and see predicted secondary cases based on whether universal masking is in place. A study from Norway in August to November 2020, pre-Delta, which found low secondary transmission when there was no mask mandate in the schools, “does provide support that if cases in the community are low (much lower than what we’re observing [in the US] in Feb 2022), that optional masking can be considered” without serious consequences, writes Benjamin. The infectiousness of the variant and community caseload play a role. He adds, “Some time this spring, communities should make a decision that takes into account vaccination status, community transmission, hospital staffing, and what’s right for their community.”

The CDC’s updated guidance bases its definition of a high-risk area on a confluence of factors, including new cases and the number of new COVID-19 hospital admissions relative to the total population, as well as the percentage of inpatients with COVID-19. If a school is in a community that isn’t considered high-risk, masks need not be mandatory, the agency says. 

William Schaffner, an infectious disease physician and public health specialist at Vanderbilt University, says, “I would pay attention to pediatric hospitalizations. They are nudging down across the country, but we really want to see them sustained” over a few weeks or months. 

“What are people willing to tolerate?” asks Murray, whose hospital saw increases in babies admitted for croup as a result of the Omicron variant. “The characterization of COVID-19 as similar to the common cold or influenza in children—I just don’t think that’s entirely accurate. While most children do have mild disease and do fine, there are post–COVID-19 complications even in children with mild disease that are extremely significant and not seen with other respiratory viruses.” 

Swann points out that even with mitigation measures like masking, distancing, and online or hybrid schools over the past two years, about 1,000 children in the US still died of COVID-19 since the beginning of the pandemic, and many others developed long COVID, about which not much is known in kids. More than 12.5 million children have tested positive for COVID-19 since the beginning of the pandemic. Almost 7,000 children have developed multisystem inflammatory syndrome in children (MIS-C). By comparison, influenza typically kills between 150–200 kids per year, though that may be an undercount. For the 2020–2021 winter flu season only one pediatric influenza death was reported; this was at a time when COVID-19 mitigation efforts were in effect, which likely played a role in the low number of flu cases and deaths, says Swann. 

And children need not become seriously ill themselves to suffer the virus’s effects: a new study from Imperial College London estimates that more than 5 million children worldwide had lost a caregiver to COVID-19 by the end of October 2021. 

See “SARS-CoV-2 Antigens Leaking from Gut to Blood Might Trigger MIS-C

“It looks as though the Omicron-fueled pandemic is now starting to abate—emphasis on starting—but we aren’t there yet,” says Schaffner. “My point of view? Better to wear the mask a month too long rather than take the masks off a month too soon. So hang in there.”  

The physical science behind masking

SARS-CoV-2 is carried within large droplets that fall to the ground quickly after someone sneezes or coughs, and in aerosols, which are smaller and can remain suspended in the air for several hours. When a person wearing a mask inhales, air curves around the mask fibers, but aerosols can’t swerve around the woven mesh, and crash into or stick to the fibers. Masks with more, or more tightly-woven, layers are more likely to catch the virus, and well-fitting masks leave no gaps for the virus to sneak around. Well-fitting N95 masks successfully filter aerosols containing SARS-CoV-2, as do KN95 and KF94 masks that hug the face without gaps; surgical masks, though less effective than these high-filtration options, tend to be more effective than cloth masks. 

Lab studies of two masked mannequin heads facing each other, with one mannequin connected to a customized compressor nebulizer that exhaled a mist of SARS-CoV-2 from its mouth, and the other mannequin connected to an artificial ventilator and viral collection unit, found that surgical masks are 60–70 percent effective at preventing the virus from being emitted and are 50 percent effective at protecting the other mannequin from being exposed to the droplets and aerosols. N95s are even better, reducing the second mannequin’s exposure by 80–90 percent, while cloth masks are much worse, at 20–40 percent effective.

“You can say it stands to reason that a mask, appropriately applied, if you don’t expect perfection, likely will attenuate the spread of the virus,” says William Schaffner, an infectious disease physician and public health specialist at Vanderbilt University. “We have been saying things akin to that in the health care setting for years.”