Introduction
In-person learning in primary and secondary schools is at the center of the educational system internationally, and is broadly considered the optimal environment for intellectual, personal and social development of children and teenagers [
1]. However, prolonged contact between large numbers of school children and teachers facilitates the spread of infectious diseases by airborne, droplet and contact transmission [
2,
3]. Since the start of the COVID-19 pandemic, SARS-CoV-2 infection and transmission in the school environment have been under large scrutiny. Efforts to better characterize COVID-19 transmission dynamics within pediatric populations have suffered from a lack of data and potential biases due to selective collection of data [
4], as with testing regimens that focus on symptomatic individuals only [
5]; this is particularly true for research which aims to explore transmission dynamics among school-age children receiving in-person education. In spite of these challenges, systematic surveillance of cases in the school setting remains a useful method of describing and quantifying changes in transmission dynamics, by age and over time [
6].
The COVID-19 pandemic is a public health emergency affecting individuals of different ages differently. Pediatric cases are generally mild and not captured through hospitalization or health care visit data [
7,
8]. Analyses of data collected from admitted children produces biased estimates of infection rates in the pediatric population. Several outbreaks have been documented in school settings, with infection and transmission occurring among pupils and staff alike [
9,
10]. The role of in-school transmission and the interaction with community and household transmission requires additional scrutiny using quantitative epidemiological tools and data collection within the school environment, while linking with community data [
11]. A solid surveillance system that can be set up and used during epidemics can be a resourceful tool to measure the impact of interventions and to timely inform decision making.
In Belgium, schools were gradually and partially re-opened in the spring of 2020 as part of the exit strategy following the first wave and corresponding lockdown. All primary and secondary schools, both from the Flemish and French language school system, resumed full-time in-person education on September 1st, 2020. Flemish schools are under the responsibility of the Ministry of Education of the Flemish region and all public and private schools that are approved, financed, or otherwise subsidised by the Flemish government are connected to the school public health system directly. The network is primarily organized via Student Guidance Centers known as Centra voor Leerlingenbegeleiding (CLBs). Within this structure and with the help of the Belgian national public health agency, Sciensano, and the regional public health agency Vlaams Agentschap Zorg en Gezondheid (VAZG), a COVID-19 school surveillance system was set up in September 2020 with occasional adjustments during the school year. Parallel, in the French region, a similar system was developed with separate data collection and supervision, with final aggregation of all collected data by Sciensano.
The aims of this manuscript are to (i) describe the development of a surveillance and testing-and-tracing system for COVID-19 cases and potential school transmission in Flemish pre-, primary and secondary-schools and to (ii) describe the frequency and epidemic curve of SARS-CoV-2 confirmed school cases by age group from October 2020 to June 2021 in conjunction with non-pharmaceutical interventions (NPI) implemented to control transmission in schools, the community background incidence and national age-specific test data, using data obtained from the school surveillance database.
Discussion
We describe the evolution of the COVID-19 epidemic in a large Belgian region where schools have been reopened and remained open almost the entire school year 2020–2021, using data from a newly developed school network-based surveillance program. We show that it is feasible to set up a surveillance network within the present structures of the school and school health network when a digital platform is available. We provide a descriptive analysis of the Flemish region school COVID-19 surveillance program, with presentation of age-specific data.
Age is one of the most important predictors of morbidity and mortality among those with COVID-19 [
22‐
24]. The presence and importance of age-dependent susceptibility and transmission remain unclear, even more so with the change in dynamics with the spread of SARS-CoV-2 variants of concern, with increased transmission rates in all ages after the appearance of the alpha variant and with further increasing transmissibility among those infected with the omicron variant [
25], including in children [
26]. In addition, there is developing knowledge on the effect of sustained high and boosted vaccination coverage in the adult population [
27], while the pediatric population has either substantially lower vaccine coverage (5–11-year-old) or still lacks an authorized vaccine (under 5 years old). In Flanders, SARS-CoV-2 vaccination was introduced during the last trimester of the reporting period of this study in the Flemish adult population. Two-dose coverage reached 44% in the > 18-year-old population at the end of the study period, however this figure includes 79% in the prioritised 65 + population at a moment of low virus circulation, not suspected to have yet an impact on the transmission dynamics in school-aged children at any point of the study. Vaccination in 12–18-year-old children was initiated only at the end of the academic year, and authorization for the under 12 population dates from December 15
th, 2021, in Belgium [
28].
The positive association between age and number of cases at the start of the school year in this study appears, at least temporally, to have disappeared, through a combination of age-specific NPIs, and NPI and age-dependent testing rules. A decrease in cases in grade 11–12 after the shift to 50% education in November 2020, and a later decrease in cases in grade 5–6 after the introduction of face masks in March 2021 were observed. The figures show the sequential overtaking of younger age groups with the largest number of cases per 100,000 population after the initiation of the grade specific interventions in the older groups, negating the age-dependent trends in case counts that were present at the start of the academic year and thereby closing the age-differences in susceptibility and infection. Mask wearing was previously introduced in all staff and pupils from grade 7 and up early in the 2020–2021 academic year. Studies investigating the effect of mask mandates [
29] estimate a lower incidence rate ratio in districts with mask mandates for all pupils 2-year-old and up and a decrease in incidence after mask wearing introduction. Formal comparison studies between regions with differing mask mandates can build on this evidence base. During the 2021–2022 academic year, in the fall of 2022, a large fourth wave hit Flanders and universal masking in schools was introduced in grade 1 and up and this NPC introduction can be the basis of further work using the data of the continued surveillance network. We find no statistically significant difference between cumulative cases from grade 5–6 through grade 9–10 after March 2020 for the 2020–2021 academic year, our study period.
We show that testing has been very heterogeneous over time during the epidemic in the pediatric population, which is different from testing dynamics within the adult population. Testing regimens influence the tests per population performed more in the pediatric than the adult population [
30]. Changing testing strategies, linked to contact risk stratification, in addition to decreased incentives to get testing for children while not at school, i.e., during school vacations or distance learning, provide missed opportunities for a more complete capturing and continuous follow up of the epidemic evolution in school aged children. Rules for assignment of high-risk versus low-risk contacts have implications for further testing, evaluation and quarantine. They directly affect case investigation and detection in the affected pupils. Adherence to these procedures therefore further affects and feeds the case finding and surveillance data, though without the possibility of having the ultimate complete data set and numerator of the truly infected. Our data thus far do not allow an unbiased estimate of the impact of testing regimens on the age-specific proportion of detected cases versus those undiagnosed [
31], nor can we quantify, in this sample, the impact of testing as a sole NPI. However, without the timely HRC identification, quarantine, school-initiated testing, diagnosis and isolation of positive cases, which is made possible through the integration of the school surveillance system in the public health framework, it is likely that additional cases would have remained undiagnosed, thus allowing continued spread of the virus. The high-quality tracing activities of the CBL staff, with their ability to prescribe additional testing, contributed substantially to the efficiency of the general contact tracing, in addition to feeding the surveillance system. We cannot, however, completely describe the administrative hurdles experienced in the set-up and maintenance of the surveillance system by the responsible agencies. The burden on the CLB clinical and general staff corps has been high. It needs to be recognised that time and energy investment in surveillance systems focused on one infectious disease divert time and effort from other core activities oriented towards supporting the broader health in the school-aged population.
A stark finding is that pupils from schools with the lowest SES carried a quantified and important higher burden of COVID-19 over the 2020–2021 school year in our study. Flanders has no large network of private schools and the NPIs equally apply in all schools, however, the differences in cumulative incidence show how social characteristics of pupils as part of social arrangements in society also in the Flemish region differentially shape infectious disease dynamics and result in unequal disease burden. Data on disparities in the burden of COVID-19 in the general population have been published [
32‐
34], with ethnic/racial minorities or historically racialized groups carrying a higher infection, disease and mortality burden and interactions of race/ethnicity with low SES through pathways of occupation and living conditions, including inability to isolate, resulting in overrepresentation in the COVID-19 case counts. Schools and pupils mirror our society and investigating the risk factors for infection and disease should include the assessment of the role of social determinants of health [
33,
35], independent and specifically for the pediatric population [
36,
37]. The pandemic clearly illustrates how this infectious disease pandemic is equally a social pandemic and implies that we need to provide additional support to the most vulnerable children and families.
The found differences in cumulative incidence by province reflect the heterogeneous provincial distribution of community cases. Even in the densely populated and well-connected region of Flanders, SARS-CoV-2 infection rates show a local distribution [
38], dependent on heterogeneous mixing patterns and subject to different policies [
39,
40]. Research shows that transmission in schools depends on the level of community transmission [
41,
42]. School cases show a high correlation with the evolution of community cases in the general population in our study (i.e., Fig.
4).
There are multiple lessons learned from the implementation of the surveillance system and the analysis of its collected data. To improve data quality and to minimize the missingness of important variables, the introduction and inclusion of surveillance within an existing digital platform has proven to be crucial given collection of data by use of excel documents is insufficient, mainly to capture meta-data correctly. Variable definition can lead to unintended loss of information, for example, while the clinicians from the CLB’s gathered data on the presence of symptoms in tested individuals, this was not reflected in the data collection where only a single reason for testing (symptoms or HRC) could be entered. Hence, the reason of testing could not directly be used to assess the impact symptoms play in this pediatric population on testing and case detection. Nor was it possible to, with minimal risk for misclassification, calculate the secondary attack rate (SAR) in schools, an estimate of interest [
43]. Further adaptations have already been implemented, including the use of linked laboratory surveillance system collected data, will allow analyses of outbreak size, SAR and reporting of proportion of symptomatic cases for following surveillance periods. We address underascertainment of cases through the automatization of the system, which resulted in a detectable increase in reporting efficiency. We cannot, however, sufficiently assess reporting delays in our analysis and formally quantify timeliness, nor its change over time, and assess the facilitating or delaying determinants. Weekend days and vacation periods suspectedly are suspected to lead to reporting delays, which we observed (Fig.
7), however such delays have decreased in duration following linkage of the surveillance system in January 2021.
The school environment of course does not only contain school children. The close interaction with the adult teaching and support staff has been the main concern for keeping the schools open, once the relatively low direct impact on morbidity [
44,
45] and extremely low mortality [
46] in the pediatric population became evident. The joint collection of cases in staff and pupils undoubtedly would add beneficial data, however this was shown complex and mainly burdened by administrative and other hurdles in this sample.
Availability of surveillance data serves multiple purposes [
47]. The surveillance data described in this study are presented biweekly to the Ministry of Education and also included in the Sciensano weekly updates [
20,
21]. Surveillance programs can detect and follow the evolution of registered cases in the school environment and provide baseline and follow-up data that can be compared to community surveillance data.
Suggestions to improve the data collection tool are the following: Addition of data of school absenteeism [
48,
49], inclusion of total number of children tested (including data on test negatives), separate collection of the variable capturing the presence or absence of symptoms and the opportunity to report additional instituted interventions during a class or school outbreak for later evaluation of its impact on the size and time to containment of the outbreak. The surveillance network can be used for follow-up and to perform impact evaluations [
50] of changes in school level interventions, like e.g. shifts in in-person education, changes in mask mandates, test modalities (antigen versus PCR) used for outbreak investigations, and others. With the availability of high-quality data, predictive models can be developed as alarm systems informing on the epidemic evolution and including calculation of the (changes in) SAR. Estimation of the attributable effect of interventions on the change of cases can additionally provide evidence relevant for policy decisions. In the future, one can also foresee that information on vaccination coverage and other details on circulating SARS-CoV-2 variants would be valuable for further study and should also be included in subsequent analyses.
Publisher’s Note
Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.