Impact of school closures and re-openings on COVID-19 transmission

Background: Globally, the education of students at primary and secondary schools has been severely disrupted by the implementation of school closures to reduce the spread of coronavirus disease 2019 (COVID-19). The effectiveness of school closures in reducing transmission of COVID-19 and the impact of re-opening schools are unclear. Methods: Research criteria for this rapid review included empirical studies, published or pre-published worldwide before January 25, 2021, that assessed the effectiveness of school closures in reducing the spread of COVID-19 and the impact of school re-openings on COVID-19 transmission. Results: Twenty-four studies on the impact of school closures and re-openings on COVID-19 transmission were identified through the seven databases that were searched. Overall the evidence from these studies was mixed and varied due to several factors such as the time of implementation of public health measures, research design of included studies and variability among the levels of schooling examined. Conclusion: Preliminary findings suggest that school closures have limited impact on reducing COVID-19 transmission, with other non-pharmaceutical interventions considered much more effective. However, due to the limitations of the studies, further research is needed to support the use of this public health measure in response to the COVID-19 pandemic. This work is licensed under a Creative Commons Attribution 4.0 International License.


Introduction
As of March 11, 2020, the World Health Organization has declared the coronavirus disease 2019 (COVID-19) outbreak a pandemic (1). Globally, jurisdictions started to implement a variety of non-pharmaceutical interventions (NPIs) to limit the spread and the impact of COVID-19 disease caused by severe acute respiratory syndrome coronavirus 2 (SARS-CoV-2). Closing schools was one of the NPIs implemented; however, these closures not only disrupted the education and daily routines of students, but also the lives of teachers and parents.
While school closures have been implemented to combat the spread of COVID-19, they were also associated with negative effects on student's mental health and academic progress and lead to increased stress in parents and teachers (2). With a lack of school-based peer interactions and daily routines, it has been reported that students experience increased distress, loneliness, anxiety and depressive symptoms (2,3). School routines are crucial for maintaining the well-being of students, especially those with mental health or special education needs (4). In addition, school closures have been associated with reduced academic achievement due to delayed educational progress (3,5,6). It is uncertain whether virtual learning is equally effective and many students from low-income households lack access to, and accommodations with, online materials (6).
Given the negative impacts of school closures, it is important to consider whether they are significantly effective in reducing the impact of COVID-19. Initially, it was assumed that school closures would be effective in mitigating the spread of COVID-19 based on the evidence from both seasonal and epidemic influenza (7,8).
In contrast, modelling studies conducted in Ontario and across Canada during the first and second waves found that school closures had limited impact on reducing the transmission of COVID-19 compared with other NPIs (9-11). Other modelling studies reported modest effects of school closures in delaying peak case numbers early in the pandemic (12,13), while some CCDR • December 2021 • Vol. 47 No. 12 studies showed a smaller magnitude of effect when compared with other NPIs (14,15). Early modelling studies relied on the underlying assumption that there is a low transmission risk in children. Although modelling studies are excellent for making informed predictions, their accuracy is dependent on the assumptions and the quality of data used. Overall, there was a need to assess the potential impacts of school closures in reducing the spread of COVID-19.
This review summarizes empirical studies on the effectiveness of school closures and the impact of re-opening schools in reducing community transmission of COVID-19 and decreasing the incidence of COVID-19 in primary and secondary schools. The principal focus of this article was the impact of primary and secondary school closures, although if studies also included data from other types of schools this was included as well.

Methods
Our research criteria included empirical studies that assessed the impact of school closures and/or re-openings on COVID-19 that were published before January 25, 2021. Predictive modelling studies were excluded. Searches to retrieve relevant articles were conducted in PubMed, Scopus, BioRxiv, MedRxiv, ArXiv, SSRN and Research Square, by the Emerging Science Group of the Public Health Agency of Canada. Search terms included the following: school AND closure OR re-opening within a database of COVID-19 literature that is updated daily. References were also used to search for additional relevant studies. Included literature was confined to English and French languages. Articles (n=966) were then screened for relevance. A total of five observational studies and nineteen ecological studies were found to be relevant (see Appendix Table A1 and Table A2).

Results
Twenty-four articles published prior to January 25, 2021 on the impact of school closures and/or re-openings on the spread of COVID-19, were identified. These included a cross-sectional study (16), two cohort studies (17,18), two cluster and outbreak investigations (19,20) and 19 ecological studies. Eleven of these studies are preprints or studies that have not yet been peer-reviewed. All studies identified in this review pre-date the identification of variants of concern.
Most observational studies assessing the impact of school closures/re-openings on the spread of COVID-19 in schools reported no significant effects (see Appendix Table A1). Four studies found no difference in incidence of cases both before and after closing schools for the holidays, following children who stayed at home vs those who went to school with strict surveillance, or following school re-opening (16)(17)(18)(19)). An outbreak investigation study reported a large outbreak from a high school in Israel, but this was confounded by the fact that the mask mandate was lifted just as there was a heatwave, which may have affected compliance with other recommended public health measures (20). Furthermore, it was noted that there was overcrowding in the high school that limited physical distancing, and extracurricular activities were not banned.
Of the ecological studies assessing community transmission (see Appendix Table A2), ten were conducted across multiple countries, five in the United States, two in Asia and two in Europe. Five studies reported that school closures and re-openings were not significantly associated with reduction in the transmission and incidence of COVID-19 and were much less effective in reducing transmission when compared with other NPIs (21-25). Four studies reported a reduction in the incidence of COVID-19 in the community ranging from 8% to 62% following school closures (26-29). Other studies reported a significant reduction in the effective reproduction number (R t ) (30-32). Three studies attributed significant reductions in mortality to school closures (29,33,34) and one study reported increased mortality with delayed school closures (35).

Discussion
Overall, the evidence from these studies was mixed and varied due to several factors. Based on the findings of the observational studies assessing the incidence of COVID-19 in schools, school closures and re-openings did not significantly contribute to COVID-19 transmission when infection prevention and control measures (IPAC) were implemented in schools. The IPAC measures implemented by the schools were similar across most of the observational studies and included masks, physical distancing, frequent cleaning, reduced class sizes and improved hand hygiene. The implementation of these measures in schools have been reported to act as a mediating variable because of the reduced transmission and risk of infection with IPAC measures (36).
The findings from the ecological studies assessing community transmission were inconsistent, with some studies reporting that school closures/re-openings were not significantly associated with reduction in transmission (21-25), and other studies reporting a significant reduction in R t (30-32) and mortality (29,33,34). In several of these ecological studies, it was reported that other NPIs such as lockdowns, gathering bans, mask mandates, non-essential business closure and travel restrictions were more effective than school closures in reducing the transmission of COVID-19. Ecological studies are considered a low level of evidence due to the research design, the multiple confounding factors and the high degree of variability in the results. All of the ecological studies included in this review analyzed data on school closures/re-openings early in the pandemic, between January-August 2020, when multiple NPIs were implemented simultaneously. Therefore, it was not possible to isolate the impact of school closures/re-openings on the number of cases of COVID-19 in the community. Additionally, only one of the ecological studies described if there was adherence to IPAC measures in the schools (25). These factors likely contribute to the heterogeneity between studies.
An important limitation of this review was the inconsistencies in the levels of schooling that were included in each study, which may have increased the variability in measures of how effective school closures were across studies. Most studies did not provide information on what schools were included when determining the impact of school closures on the spread of COVID-19. Some studies measured primary and secondary school closures alone and some measured them in combination with post-secondary schools. The risk of transmission may have varied significantly between students in primary and secondary schools because of potential differences in their behaviours and adherence to IPAC measures with resultant difference in reduction of viral transmission. Transmission was found to be lower in primary schools relative to secondary schools, based on the results of one study in a review that assessed this (19). Although not specifically stated in the previous study, their results were in line with what we know about the use of IPAC to limit transmission in these settings (37). Additionally, the relative impact of school closures and re-openings have been shown to vary according to the time of implementation, level of community transmission, and the structure of populations from different countries.
Based on the empirical evidence summarized in this article, school closures had a small effect on limiting the spread of COVID-19 in schools and the community and appeared to be much less effective than other NPIs. These findings are also consistent with modelling studies conducted across Canada (9-15). The implementation of school closures is currently based on when the transmission of COVID-19 in the community is high-as dictated by local health jurisdictions; however, the role of school closures and re-opening in areas with low community transmission is less clear and should be studied further.
School closures may be associated with negative effects on student's mental health and academic progress (2); thus, public health decision makers should consider if the apparent low efficacy of school closures in reducing transmission outweighs the many negative consequences on students' well-being. Overall, the confidence in this evidence is low given that the studies in this review vary by several factors and were conducted at different times and in a number of countries. Finally, the study period of this review is also a limitation, as there are marked differences in the 3 rd and 4 th waves compared with the 1 st and 2 nd waves of COVID-19 with the introduction of more transmissible variants of concern. How the present evidence will compare with that obtained during periods of the spread of more transmissible variants of concern is not known at this time and will require further study.

Conclusion
The findings of this review may have implications for public health decision making and future research on mitigation strategies for schools. The preliminary evidence provided in this review suggests that school closures and re-openings may have only a limited impact on the transmission of COVID-19 within a community. However, there is still a high degree of uncertainty due to the high variation in the methodology and results across the various studies. Additional research is needed to further explore more systematically the impacts of school closures and to determine how and when they may be used most effectively in controlling the epidemic.
Important knowledge gaps to consider are how 1) the presence of new variants of concern and 2) the rollout of COVID-19 vaccinations will impact the transmission of COVID-19 within the schools and communities. The evidence presented in this article pre-dates the introduction of variants of concern; therefore, additional research is needed to understand how the emergence and spread of these variants will impact the effectiveness of school closures or what the impact of school re-openings will have on the spread of COVID-19.

Authors' statement
MEJ -Methodology, investigation, writing-original draft RE -Conceptualization, writing-review and editing, supervision LW -Writing-review and editing PH -Writing-review and editing NA -Writing-review and editing MS -Writing-review and editing SB-A -Writing-review and editing

Competing interests
None. This study analyzed the association between school re-opening dates and COVID-19 cases across twenty-one Italian regions by using a database on positive cases in elementary, middle and high schools and SARS-CoV-2 incidence in the general population. IPAC measures included temperature control, hand hygiene, mask mandate for students/staff, physical distancing, ban on sports and music and reduced duration of school.
Several COVID-19 outcomes were measured during school re-openings: growth of incidence, R t , and secondary infections.
There was no evidence that the second SARS-CoV-2 wave was driven by school re-openings across the regions.
SARS-CoV-2 incidence among students was lower than the general population of all but two Italian regions.
The increase in R t was not associated with the different school opening dates.
School closures implemented in two regions did not affect the decline of R t .
Fontanet (2020) (18) Retrospective cohort study France Feb-Apr 2020 This retrospective cohort study included primary school pupils, teachers, non-teaching staff, parents and relatives exposed to SARS-CoV-2 in February and March from six schools. IPAC measures were not described.
A questionnaire covering sociodemographic information and history of recent symptoms was completed by participants. Blood samples were also tested for the presence of anti-SARS-CoV-2 antibodies using a flow-cytometry-based assay. Three introductions of SARS-CoV-2 occurred prior to school closures. Spread within schools vs families was investigated in this sero-epidemiological study. IAR was compared between school contacts and family contacts to understand the potential impact of the school closure.
No secondary infections from COVID-19 introductions in schools was detected among students and teachers.
Among pupils who were infected, their parents were significantly more likely to be infected (61.0% versus 6.9%; p<0.0001), The same was identified among relatives of infected pupils compared with non-infected pupils (44.4% versus 9.1%; p=0.002).
Transmission did not appear to be impacted by the closure of schools.

Cross-sectional study
Israel Mar-May 2020 During a national lockdown, an alternative school was used for healthcare workers' children to attend with strict symptom surveillance. Families with children who remained at home were compared with children at this alternative school. IPAC measures in the school included daily disinfecting, face mask use by staff and frequent hand washing.
This cross-sectional study included 70 children who attended the alternative primary school and 36 who stayed home, along with their 78 parents.
Data was collected through a short questionnaire; nasopharyngeal and oropharyngeal swabs were obtained and tested for SARS-CoV-2 by RT-PCR, and blood was collected for SARS-CoV-2 IgA and IgG titres.
Symptoms were reported in approximately 16% of children in both groups: those who attended the school (n=11/70) and those who did not (n=6/36).
Positive serology tests showing previous exposure was detected in less than 2% of each group and they were not significantly different from each other.
There was no evidence of increased infection in those at school compared with those at home.
COVID-19 rates were higher in students in grades 7-9 than in grades 10-12. School closures were not found to have significant effects on population-adjusted infections in the long-term (90 th to 120 th day).
Banholzer (2020) (27) Ecological study 20 countries Apr 2020 In this study, the impact of NPIs on the relative reduction of new COVID-19 cases using a Bayesian hierarchical model with a time-delayed effect for each NPI. IPAC measures were not described.
Compared with other NPIs examined, school closures appeared to be one of the least effective NPIs.
Banholzer (2021) (26) Ecological study 20 countries Feb-May 2020 Using a semi-mechanistic Bayesian hierarchical model, this study aimed to measure the effectiveness of seven NPIs in reducing the number of new infections. IPAC measures were not described.
This reduction was lower than two other NPIs (event bans and venue closures).

countries
Jan-May 2020 This study estimated the effectiveness of NPIs in 41 countries using a Bayesian hierarchical model by linking intervention implementation dates to national case and death counts.
Intervention effect sizes were categorized by the median reductions in the R t of less than 17.5% (small), between 17.5 and 35% (moderate) and at least 35% (large). NPIs examined included: limiting gatherings to fewer than 1,000 or fewer than 100 or fewer than 10, closing some businesses, closing most businesses, closing schools and universities, and stay at home orders. IPAC measures were not described.
The percentage reduction in R t associated with closing both schools and universities in conjunction was 38% (95% CI: 16%-54%), which was categorized as a large effect size.
The individual effects of school closures was not measured.

Global
Mar 2020 This study aimed to assess the effect of school closures on COVID-19 cases globally by measuring correlation between the incidence rate on the day of school closure and the incidence rate in the following days. IPAC measures and level of schooling included in the study were not described.
The results indicate that there was a strong correlation between the day of educational facilities closure and the incidence rate in the following days (16 th (p=0.004), 30 th (p=0.002) and 60 th (p=0.031) days since the 100 th confirmed case in each country).
Early closure of schools is statistically significantly correlated with lower incidence rates further on during the different phases of the epidemic.

Papadopoulos (2020) (39)
Ecological study Global Jan-Apr 2020 The impact of lockdown measures was assessed globally using publicly available data. The timing and association of early NPIs with log 10 national deaths (Log D ) and log 10 national cases (Log C ) was compared between nations. IPAC measures and level of schooling included in the study were not described.

countries
May 2020 The aim of this study was to determine the associations between NPIs and COVID-19 outcomes.
Associations with NPIs were assessed with their respective stringency index on several outcomes that form the epidemic curve: mean mortality rate, time to peak, peak deaths per 100,000 population, cumulative deaths after peak per 100,000 population and ratio of the mean slope of the descending curve to the mean slope of the ascending curve. IPAC measures and level of schooling included in the study were not described.
School closures were effective against all outcomes, except time to reaching the peak of the epidemic curve.
In non-European countries, school closures were most effective against mean mortality rate (rs=-0.757, p=0.049).

Esra (2020) (30)
Ecological study Global Jan-May 2020 This study used globally reported data on SARS-CoV-2 cases to fit a Bayesian model framework to estimate the association with NPIs and transmission.
NPIs examined include stay home mandates, gathering limits, school closures (primary, secondary and tertiary educational institutions) and mask policies. IPAC measures were not described.
There was an estimated mean reduction in R t of 12% (95% CI: 5%-19%) with school closures (primary, secondary and tertiary educational institutions).
A weighted random effects regression was used to determine the association between epidemic growth RRR and climate measures and public health interventions such as school closures, restrictions of mass gatherings and measures of social distancing during an exposure period 14 days previously. IPAC measures and level of schooling included in the study were not described.
This association was more pronounced in areas that implemented two or three NPIs compared with one NPI.

Global
Jun 2020 This study examined the variation of NPIs in 130 countries in two periods: 1) prior to first COVID-19 death and 2) 14-days-post first COVID-19 death.
This study examined associations with daily COVID-19 deaths per million and each 24 day period (time between virus transmission and mortality). IPAC measures and level of schooling included in the study were not described. This study aimed to determine if school closures were associated with a decrease in the cumulative incidence of COVID-19 and mortality.
The impact of primary and secondary school closures was assessed using publicly available data from all 50 states. IPAC measures were not described.
These associations were stronger in states with a low cumulative incidence of COVID-19 at the time of the school closure.

US
Apr 2020 This study aimed to measure the impact of NPIs on the effective R t of COVID-19 in US states.
The average R t was measured during the weeks after each state reached 500 cases. R t was measured at the week immediately following 500 th case (days +1 to +7) and at a one-week delay from 500 th case (days +8 to +14).
NPIs examined included stay at home order, educational facilities closure and non-essential business closure. IPAC measures and level of schooling included in the study were not described.

US
May 2020 This study aimed to examine the effects of NPIs on the cumulative incidence rates of COVID-19 in the US on a state-level in the 25 most populated cities, while adjusting for socio-demographic risk factors.
A negative binomial regression was used to calculate adjusted rate ratios by comparing two levels of a binary variable: "above median value," and "median value and below" for days to implementing an NPI.
NPIs assessed in this study included: days to closing of nonessential businesses, days to stay home orders, days to restrictions on gatherings, days to restaurant closings and days to schools closing. IPAC measures and level of schooling included in the study were not described.
Delays in closing schools was positively associated with cumulative incidence at the state level.

Liu (2020) (22)
Ecological study US Feb-Apr 2020 This study estimated the impact of nine different NPIs on reduction of the effective R t by using the daily number of reported new cases and inferred infections in 50 states. IPAC measures and level of schooling included in the study were not described.
This reduction was smaller than six other NPIs assessed (stay-at-home order, face masks, gathering ban, non-essential business closure, declaration of state of emergency and interstate travel restriction).

Yehya (2020) (35)
Ecological study US Jan-Apr 2020 In this study, a state-level analysis was conducted to determine association between later implemented NPIs with higher mortality rates.
Using a multivariable negative binomial regression, the association was tested between timing of emergency declarations and school closures with 28-day mortality. Day 1 for each state was set to when they recorded 10 or more deaths. IPAC measures and level of schooling included in the study were not described.