Background
The global spread of the novel coronavirus (SARS-CoV-2) has triggered the implementation of community-based public health measures around the world [
1,
2], including school closures (SC) [
3], to mitigate the impact of the COVID-19 pandemic. The evidence that SC reduces disease spread in the community comes mostly from experience with influenza virus [
4,
5]. Children are thought to play an important role in influenza transmission through close contact with classmates, friends, teachers, caregivers, and family. Furthermore, they have a reduced ability to implement hygiene and may have higher viral shedding for a longer period than adults [
6]. However, evidence supporting SC to reduce infectious disease transmission in the wider community is uneven [
7‐
10] and the epidemiology of influenza virus and SARS-CoV-2 may differ in important ways. For example, the incubation period of COVID-19 is estimated to have an average of 5.2 days [
11,
12], compared to 1.5 days estimated for influenza A viruses [
13]. Furthermore, the reproduction number of COVID-19 exceeds 2 in most settings, considerably higher than estimates for influenza in the range 1.2–1.8 [
14]. These factors may significantly influence the effect of SC during COVID-19 pandemic.
At the time of writing, the province of Ontario has instituted indefinite SC [
15] as well as other social distancing measures. SC is disruptive to families and society with measured and unmeasured direct and indirect costs [
16‐
19]. Determining the optimal duration of SC may depend on emerging local epidemiologic data and learning from the experience of other countries that have implemented closures earlier in this pandemic [
3,
8,
20]. The effect of SC on COVID-19 epidemiology is likely to be affected by other measures such as self-isolation (SI). It is therefore imperative to evaluate the potential outcomes of this public health strategy, along with other social distancing measures, to inform decisions on the continuation of SC during current outbreaks or future waves of COVID-19.
In this study, we developed and parameterized a computational model with demographics of Ontario, Canada [
21], and the latest estimates of COVID-19 characteristics [
11,
22], to simulate disease spread. We varied the length of SC under a range of scenarios for voluntary SI of symptomatic infected individuals who reduce their contacts within the community by staying at home. Compared to SI, our results indicate that SC may have limited impact on the overall attack rate (i.e., the proportion of the population infected throughout the outbreak) and the subsequent need for critical care of COVID-19 patients. We explicate that the longer incubation [
11,
22] and pre-symptomatic [
23] periods and larger reproduction numbers of COVID-19 [
11,
24,
25], compared to influenza, are key epidemiologic attributes determining low effects of SC alone during COVID-19 outbreaks.
Discussion
In the absence of therapeutic antivirals or a vaccine to prevent the spread of SARS-CoV-2, public health and government officials require evidence on appropriate measures to manage population interactions that prevent and slow transmission so that health care systems are not overwhelmed. Research supporting SC for disease outbreaks among other forms of social and physical distancing has been mixed to date, often because of differences in study method and local contexts [
5,
43]. Studies found differing outcomes for infection transmission [
20,
44,
45] and attack rates [
20] depending on the type of study, local conditions, and the nature (duration, scope) of SC. A review of studies in 2014 [
43], for example, concluded that “in the absence of evidence to guide practice, public health decision-makers may determine the need to close schools on a case-by-case basis,” taking circumstances of each epidemic, characteristics of the community affected by the epidemic, and other available strategies into account [
43]. In contrast to studies deriving data from evidence of decreased transmission during holidays, a recent study measured interactions during reactive SC, demonstrating more than 50% reduction in contact between students [
4].
Significantly, several published studies have found that closing schools did not alter contact patterns between school children, family, and older relatives [
29,
46,
47]. In our study, we took into account restrictions of social contacts among school children from 60 to 80% during SC [
4], without affecting contact patterns between and among other age groups. We found that in the absence of measures to interrupt disease transmission by symptomatic cases, SC has relatively limited impact on outcomes measured here, including attack rates, total number of intensive care unit beds required (Additional file
1: Figs. S2-S4), as well as delay in the peak of outbreaks. However, when SI was implemented, a more substantial reduction of disease burden was achieved, depending on the proportion of cases practicing SI, and the effect of SC was more pronounced in flattening the outbreak and delaying the peak time. This suggests that in the context of COVID-19 infection, rapid case identification and SI by infected individuals, including mildly symptomatic individuals, is of critical importance and a high-impact strategy that can determine the trajectory of outbreaks. This is particularly important in the context of recent findings of high viral loads in infected persons at 0.7 days before symptom onset [
23], with potential for significant contribution of silent transmission during the pre-symptomatic stage [
48].
While SC has been shown to slow the spread of seasonal influenza in both observational [
49,
50] and modeling studies [
20,
51], the underlying epidemiologic determinants of these observations are not well understood. The incubation and pre-symptomatic periods, and the reproduction number of SARS-CoV-2 are considerably different from both seasonal and pandemic influenza infections [
13,
14]. The incubation period for COVID-19 is estimated at 5.2 days (95% CI: 4.1 to 7.0) [
33], which is at least three times longer than the short incubation period of 1.4 days (95% CI 1.3–1.5) for influenza A [
13]. Furthermore, the lower bound of estimates for the reproduction number of COVID-19 [
11] is comparable to or higher than the upper bound for most estimates of reproduction number of influenza epidemics [
14]. While not within the scope of this study, evaluation of the relation between epidemiologic characteristics of infectious diseases and the impact of social distancing measures, and particularly SC, during outbreaks merits further investigation in future studies. The social and economic disruption caused by SC to families and society must of course also be weighed in taking a decision to close schools. Although SC has limited effect on the overall attack rate, it can still reduce attack rates among school children. Thus, alternative measures to reduce contacts among children post-lockdown, such as attendance in shifts, can be considered. Early in the pandemic, it appeared that symptomatic and serious COVID-19 illness in children was uncommon compared to adults [
52,
53]. However, recently increasing cases of a serious and sometimes life-threatening syndrome, pediatric COVID-associated multi-inflammatory syndrome (PMIS) have been reported [
53]. The etiology of this temporally associated illness is not yet clear. If PMIS is causally related to COVID-19, reduction of pediatric illness by SC may be warranted.
Our findings should be considered in the context of model assumptions and parameters that are based on early estimates and may be subject to uncertainty. As new information and data become available, a better quantification and parameterization of our model could provide more accurate projections on the effectiveness of SC in the presence and absence of SI. Furthermore, the effect of other social distancing measures that have been implemented during the COVID-19 pandemic, such as canceling large public gatherings, having many people work from home, and university closures, is not addressed in the model. Quantifying the effects of these measures in future studies would also provide further insights into control of emerging diseases.
Conclusion
Our study demonstrates that while SC will mitigate disease transmission during the COVID-19 pandemic when combined by other social distancing measures, it may have markedly lower effectiveness in reducing attack rates and hospitalizations compared to SI. As an important measure of social distancing and in an effort to protect children, SC has been implemented in many countries affected by the COVID-19. However, its effectiveness on curbing the outbreaks has not been investigated. Our findings highlight the importance of epidemiologic parameters of particular infectious diseases on the effectiveness of SC. Importantly, our results show that, in the context of COVID-19 outbreaks, public health measures can be expected to have different effectiveness. For the greatest impact, social distancing measures should be directed at those interventions which are most likely to interrupt the chain of transmission during both pre-symptomatic and symptomatic illness. Interventions such as SI, working at home, social distancing, and mask wearing when moving in the community are expected to reduce the spread of COVID-19 significantly, which would allow the healthcare systems to manage critical care capacity for treatment of severely ill patients within finite resources available.
Publisher’s Note
Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.