STUDY POPULATION

This study considered the 79 public, noncharter school districts in the greater Boston area, defined according to the U.S. Census Bureau as the New England city and town area of Boston–Cambridge–Newton (Fig. S1 in the Supplementary Appendix, available with the full text of this article at NEJM.org). Of these 79 school districts, 7 with unreliable or missing Covid-19 data were excluded (Table S1). The final sample included 72 school districts, which comprised 294,084 students and 46,530 staff during the study period. The study period was defined as the 40 calendar weeks of the 2021–2022 school year, which ended on June 15, 2022 (the end of the last full reporting week in all districts).

INTERVENTION AND PRIMARY OUTCOME

The primary exposure variable was whether a school district lifted or sustained its masking requirement in each reporting week. For all school districts, masking requirements were in place from the start of the study period through February 28, 2022, when the statewide masking policy was rescinded. A school district was considered to have lifted its masking requirement in a given reporting week if the requirement had been lifted before the first day of the reporting week (reporting weeks start on Thursday). The primary outcome was the incidence of Covid-19 among students and staff, considered together and separately.

DATA SOURCES

For each school district, data regarding weekly Covid-19 cases, student enrollment, and staffing during the 2021–2022 school year were publicly available from the Massachusetts DESE.^33,34 Throughout the study period, DESE required standardized weekly reporting of all positive tests for Covid-19 among students and staff, regardless of symptoms, testing type or program (e.g., testing of symptomatic persons or pooled polymerase-chain-reaction testing), and testing location (community setting or school setting). Details regarding DESE reporting requirements are provided in the Supplementary Appendix. DESE strongly encouraged, and provided full funding for, school districts to opt in to standardized Covid-19 testing programs; 2311 Massachusetts schools (approximately 95%) participated in at least one such program. From 1 month before the statewide masking policy was rescinded through the end of the school year, statewide testing recommendations did not differ according to masking or vaccination status (Table S2).³⁵

The dates during which masking requirements were in place for each school district were obtained from school-district websites or local news sources. For sensitivity analyses, data to be used for covariate adjustments, including data regarding Covid-19 indicators (i.e., measures of Covid-19 burden) according to city and town and Covid-19 vaccination coverage according to age, were publicly available from the Massachusetts Department of Public Health. For descriptive analyses, data regarding the distribution of students and staff according to sociodemographic characteristics and the distribution of students in populations selected and defined by DESE (low-income students, students with disabilities, and English-language learner [ELL] students) during the 2021–2022 school year were obtained from DESE.³⁴ In addition, data regarding building conditions and learning environment were obtained from the Massachusetts School Building Authority 2016–2017 school survey (most recent data).³⁶

CONTRIBUTIONS

The first, second, and last authors wrote the first draft of the manuscript and vouch for the accuracy and completeness of the data and the fidelity of the study to the protocol, available at NEJM.org. All the authors reviewed and edited the draft and made the decision to submit the manuscript for publication. No external funding was received for this study.

STATISTICAL ANALYSIS

Trends in the observed incidence of Covid-19 (weekly Covid-19 cases per 1000 population) before the statewide masking policy was rescinded were compared with trends after the policy was rescinded according to the length of time that school districts sustained masking requirements. A difference-in-differences analysis for staggered policy implementation was used to compare the incidence of Covid-19 in school districts that lifted masking requirements with the incidence in districts that sustained masking requirements in a given reporting week (i.e., school districts that had not yet lifted masking requirements) in order to estimate the effect of lifting masking requirements.^37,38

Difference-in-differences methods allow for the estimation of causal effects of policy changes enacted at the group level by comparing the change in the outcome over time in the intervention group with the change in the control group, under an assumption of parallel trends (i.e., in the absence of the intervention, outcomes in the intervention group and the control group would have remained parallel over time).^37,39 Unlike some observational methods, difference-in-differences methods are not biased by unmeasured time-invariant confounders or time-varying confounders with consistent trends across the intervention and control groups; the absence of such bias strengthens causal inferences. In this analysis, the weekly and cumulative effects of lifting masking requirements during the 15 weeks after the statewide masking policy was rescinded were estimated with respect to the incidence of Covid-19 cases in school districts that lifted masking requirements (i.e., the average [mean] treatment effect among the treated). Details regarding the difference-in-differences analysis are provided in the Supplementary Appendix.

Several sensitivity analyses were performed: a formal test for parallel trends before masking requirements were lifted; adjustment for time-varying covariates, including Covid-19 indicators at the community level, vaccination coverage, and previous incidences of infection among students and staff; and an assessment of the potential effects of differences in testing definitions or programs across districts (Tables S3 and S4). In the main analysis, data for the weeks in which school districts did not report Covid-19 cases were corrected (these weeks were originally recorded as having zero cases), all school districts in the greater Boston area were included as comparison districts, and weighting according to school population size was performed to capture the effect of masking policies at the population level.

Finally, to provide insight into Covid-19 policy decisions in schools and their potential to exacerbate or mitigate inequities in Covid-19 incidence and educational outcomes, descriptive analyses were performed. The decision to sustain or lift masking requirements was assessed according to various school-district characteristics, including sociodemographic characteristics of the students and staff and physical characteristics of the learning environment.

PRIMARY ANALYSIS

Of the 72 school districts in the greater Boston area that were included in the study, only Boston Public Schools and Chelsea Public Schools sustained masking requirements throughout the study period (Fig. S2A). Of the remaining school districts, 46 districts (64%) lifted masking requirements in the first reporting week after the statewide masking policy was rescinded, 17 (24%) lifted masking in the second reporting week, and 7 (10%) lifted masking in the third reporting week (Fig. S2B). Cumulatively, 46 districts lifted masking requirements and 26 districts sustained masking requirements by the first reporting week after the policy was rescinded, 63 lifted and 9 sustained masking by the second reporting week, and 70 lifted and 2 sustained masking thereafter.

Before the statewide masking policy was rescinded, the trends in the incidence of Covid-19 observed in the Boston and Chelsea districts were similar to the trends in school districts that later lifted masking requirements. However, after the statewide masking policy was rescinded, the trends in the incidence of Covid-19 diverged, with a substantially higher incidence observed in school districts that lifted masking requirements than in school districts that sustained masking requirements. These trends were observed among students and staff overall (Figure 1A), as well as among students alone (Figure 1B) and among staff alone (Figure 1C).

The strength of the association between school masking policies and the incidence of Covid-19 in school districts varied according to the incidence of Covid-19 in surrounding communities; the strongest associations were observed during the weeks when the incidences in surrounding communities were highest (Figs. S3 and S4). The weekly effects that were observed among students and staff overall were similar to those observed among students alone and those observed among staff alone, with slightly greater effects observed among staff than among students. In addition, the weekly effects were consistent with the cumulative effects.

Overall, the lifting of masking requirements was associated with an additional 44.9 Covid-19 cases per 1000 students and staff (95% CI, 32.6 to 57.1) during the 15 weeks after the statewide masking policy was rescinded (Table 1). This estimate corresponded to an additional 11,901 Covid-19 cases (95% CI, 8651 to 15,151), which accounted for 33.4% of the cases (95% CI, 24.3 to 42.5) in school districts that lifted masking requirements and for 29.4% of the cases (95% CI, 21.4 to 37.5) in all school districts during that period. The effect was more pronounced among staff. The lifting of masking requirements was associated with an additional 81.7 Covid-19 cases per 1000 staff (95% CI, 59.3 to 104.1) during the 15-week period, with these cases accounting for 40.4% of the cases (95% CI, 29.4 to 51.5) among staff in school districts that lifted masking requirements. Because persons who had a positive test for Covid-19 were instructed to isolate for at least 5 days, the additional cases translated to a minimum of approximately 17,500 missed school days for students and 6500 missed school days for staff during the 15-week period (Table S5).

SENSITIVITY ANALYSES

The results were shown to be robust in a range of sensitivity analyses, including analyses that assessed potential differences in testing programs and analyses that adjusted for Covid-19 indicators at the community level and vaccination coverage according to age (Fig. S5). Results of sensitivity analyses are provided in the Supplementary Appendix.

DESCRIPTIVE ANALYSES

School districts that chose to sustain masking requirements longer had higher percentages of low-income students, students with disabilities, and ELL students (Figure 3A), as well as higher percentages of Black and Latinx students (Figure 3B) and Black and Latinx staff (Figure 3C), than school districts that chose to lift masking requirements earlier. In addition, school districts that chose to sustain masking requirements longer tended to have school buildings that were older and in worse physical condition (e.g., with outdated or absent ventilation or filtration systems) and to have more students per classroom (Figure 3D).

Schools are an important yet politically contested space in the Covid-19 response, which makes analyses such as this one particularly relevant to decision makers. We estimated that the lifting of masking requirements in school districts in the greater Boston area during March 2022 contributed an additional 45 Covid-19 cases per 1000 students and staff during the following 15-week period. Overall, this estimate corresponded to nearly 12,000 additional Covid-19 cases among students and staff, which accounted for one third of the cases in school districts that lifted masking requirements during that time and most likely translated to substantial loss of in-person school days.

We observed that the effect of school masking policies was greatest during the weeks when the background incidences of Covid-19 in surrounding cities and towns were highest, a finding that suggests that universal masking policies may be most effective when they are implemented before and throughout periods of high SARS-CoV-2 transmission. Under the CDC guidance at that time, as well as the updated guidance issued in August 2022, universal masking would not have been recommended until the incidences of Covid-19 in schools and surrounding communities were already nearing their peak (May 2022); by this time, a substantial proportion of the effects of masking polices that we observed had already accrued. As such, relying on lagging metrics such as CDC Covid-19 Community Levels and Covid-19 hospitalizations to inform school masking policies is most likely insufficient to prevent Covid-19 cases and loss of in-person school days, and policymakers might instead consider measures of community transmission (e.g., SARS-CoV-2 wastewater concentration or Covid-19 incidence) to inform such policies.

Understanding Covid-19 policy decisions requires attention to power and existing historical and sociopolitical contexts.^10,40 Structural racism and racial capitalism operate through multiple pathways, including higher levels of household crowding and employment in essential industries and lower levels of access to testing, vaccines, and treatment; these structural forces differentially concentrate the risk of both SARS-CoV-2 exposure and severe Covid-19 in low-income and Black, Latinx, and Indigenous communities.^9-11,18 In our study, school districts that chose to sustain masking requirements longer tended to have school buildings in worse physical condition and more students per classroom, and these districts had higher percentages of students and staff already made vulnerable by historical and contemporary systems of oppression (e.g., racism, capitalism, xenophobia, and ableism). In Boston and Chelsea, more than 80% of the students are Black, Latinx, or people of color, and these cities were among the Massachusetts cities and towns that were hit hardest by Covid-19. Students and families in these school districts have strongly advocated and organized for governmental action to increase Covid-19 protections in schools, emphasizing their role as essential workers, the risk to vulnerable family members, and the unequal consequences of missed work and school.^41,42 The decision in some school districts to sustain school masking policies longer may therefore reflect an understanding among parents and elected officials that structural racism is embedded in public policies and that policy decisions have the potential to rectify or reproduce health inequities.^10,14,16,40

A growing body of work suggests that knowledge of differential conditions and inequitable effects may decrease support for Covid-19 protections among systematically advantaged groups, whose relative position largely insulates them from Covid-19 harms, while simultaneously increasing support among groups that are directly affected by systems of oppression.^43-45 For example, in a randomized trial in which White persons were assigned to receive information about structural causes of persistent Covid-19 inequities across racial or ethnic groups or to not receive such information, those who received the information were less likely to support Covid-19 prevention policies and were less likely to report individual concern about Covid-19 and empathy for the groups that were most affected.⁴⁵ In several studies and polls, Black and Latinx parents were more likely than White parents to support school masking requirements and less likely to have confidence that schools could operate safely without additional protections.^43,44,46 Failure to consider unequal baseline conditions and ongoing inequitable effects of Covid-19 policies risks further exacerbating inequities in Covid-19 incidence and educational outcomes.

Because universal masking policies in schools have been contentious, we anticipate several critiques. One such critique is that the benefits of universal masking in schools are outstripped by potential disruptions to teaching, learning, and social development. These effects warrant further rigorous evaluation; however, to date, there is no clear existing evidence that masking inhibits learning or harms development.^47,48 In addition, such effects might be considered alongside the spectrum of benefits of universal masking, including fewer missed school days and staffing shortages, reduced risk of illness for students and their families, and reduced economic hardship for caregivers, who might miss work if their child is sick or if they become ill themselves. For example, in Lexington, MA, a comparison district approximately 10 miles from Boston, mean student and staff absences due to Covid-19 during weeks when masking was optional were 50% higher than absences during previous weeks, when masking was required (see the Supplementary Appendix).

In addition, severe Covid-19 and post-Covid conditions remain substantial risks among school-age children. Like much of the United States, the greater Boston area has low Covid-19 vaccination coverage among children (only 53% of children 5 to 11 years of age had been fully vaccinated in Boston and Chelsea through October 2022, as compared with 67% in comparison districts), with substantial inequities according to race or ethnic group and socioeconomic status. Furthermore, we observed greater benefits of sustaining masking among staff, a finding that emphasizes that universal masking is an important component of comprehensive workplace protections for staff, who may be at a higher risk for severe Covid-19 than students. In addition, staff absences may be especially consequential for students who need additional educational supports and services, including ELL students and students with disabilities.

A second common critique is that there are alternative approaches to reducing transmission and severe disease, such as improved ventilation and increased vaccination coverage. Our findings show that the better ventilation and higher vaccination coverage in school districts that lifted masking requirements than in districts that sustained masking requirements were insufficient to prevent all Covid-19 cases in these schools. Therefore, although we cannot weigh the full spectrum of individual and societal implications of masking policies, our study highlights the important role of interim universal school masking policies in mitigating the effects of Covid-19 while longer-term, more sustainable policies are developed to increase vaccination uptake and improve learning environments.

A key strength of this study is our use of difference-in-differences methods with staggered dates of the lifting of masking requirements. Although there are some factors related to SARS-CoV-2 exposure that differed across school districts, difference-in-differences methods yield robust analyses in the context of sources of confounding that do not change over time (e.g., sociodemographic characteristics or building conditions) or do not coincide with the policy change of interest. In sensitivity analyses, the benefits of masking requirements persisted after we controlled for Covid-19 indicators at the community level, vaccination coverage, and previous incidence of infection. Furthermore, we found that school districts that lifted masking requirements were districts that would have been expected to have lower incidences of Covid-19 (on average, they had buildings in better physical condition and had higher vaccination coverage), which suggests that any residual confounding by Covid-19 risk would have led to underestimation of the harms of lifting masking requirements overall.

A limitation of this study is that we did not have data regarding Covid-19 testing in individual school districts. However, DESE ended the practice of required testing of only unmasked close contacts in January 2022, and data from that “test-and-stay” program show that far too few schools continued with the program for it to explain our results. Under the most extreme assumptions, additional testing of unmasked close contacts could explain less than 7% of the estimated excess cases. Overall, our findings should be interpreted as the effect of universal masking policies and not as the effect of masking per se, since masks were still encouraged in most school settings. Despite this consideration, the effect of lifting masking requirements was substantial.

The winter wave of the B.1.1.529 (omicron) variant during the 2021–2022 school year will not be the final Covid-19 surge to affect students and staff, and ongoing efforts to address inequitable environmental risks and effects of Covid-19 in school settings are urgently needed. Our results support that universal masking with high-quality masks or respirators during periods of high community transmission is an important strategy for minimizing SARS-CoV-2 spread and loss of in-person school days. Our results also suggest that universal masking may be an important tool for mitigating the effects of structural racism in schools, including the differential risk of severe Covid-19, educational disruptions, and health and economic effects of secondary transmission to household members. School districts could use these findings to develop equitable mitigation plans in anticipation of a potential winter Covid-19 wave during the 2022–2023 school year, as well as clear decision thresholds for removing masks as the wave abates.