In the case of flu, young children are often the driving force of the epidemic. That is why long before the emergence of COVID-19, we were already conducting research on how school closures could potentially slow the spread of infectious diseases.

The role of children in the pandemic

From a large international study across eight European countries, we learned that the number of contacts is approximately 30% lower on weekends compared to weekdays, and about 9% lower during vacation periods in comparison to a regular school or work week. For most countries, this reduction in contacts is associated with a decrease in the basic reproduction rate for respiratory diseases, with a decline of around 21% during weekends and 17% during vacations, although there were some variations between countries.

Specifically for Belgium, we utilized a meta-population model (based on contacts and mobility) and data from the winter 2008-2009 flu season to assess the impact of temporary school closures. In this analysis, we found that the timing of school closures is crucial. The Christmas holidays play a significant role in containing the spread of the virus, though it somewhat depends on when the peak of the flu season occurs. At the same time, we also know from previous research that school closures have far-reaching consequences for the mental well-being, not only of the children themselves but also of their parents and, indeed, virtually all segments of the population. Closing schools can indeed be effective in curbing epidemics, but it comes at a significant cost. Therefore, data on virus spread and infections among children and adolescents are indispensable for policymakers to make decisions and implement measures that minimize collateral damage.

In the case of COVID-19, the role of children remained unclear for a long time. Children rarely appeared to become seriously ill, but as long as they can carry and spread the virus, they can play a significant role in the dynamics of the epidemic. Since children often remain asymptomatic after infection with the coronavirus, there was a real risk dof underestimating the virus circulation in this age group during the first few weeks and months of the pandemic.

Why keeping schools open should be a priority

Data from contact tracing in January and February 2020 in China revealed that the attack rate of the virus—the percentage of susceptible individuals who actually become ill—did not differ significantly among age groups, although symptoms in children were often less severe. Some studies suggested that children transmit COVID-19 less frequently than adults, but the rapidly changing measures made the results difficult to interpret. The initial studies on the presence of antibodies in children during the first wave in Spain (April 2020) and in the United States (May 2020) reported that between 0.7% and 5.8% of minors had developed immunity to the coronavirus. This was a significantly lower percentage compared to adults, where 1 in 10 had antibodies during the same period. A Swiss study compared the presence of antibodies in 5- to 9-year-olds and 10- to 19-year-olds, finding significantly fewer antibodies in the youngest children (only 0.8%) while the figure for teenagers was comparable to that of adults (9.6%). These early studies suggested that the role of children in the COVID-19 epidemic was at least different from a typical flu epidemic. However, relying solely on antibodies provides only a limited view of the situation.

How much virus circulates in our schools?

Meanwhile, we worked with colleagues from Leuven's University Vaccine Center to understand the local situation better. Since we knew that there were significant regional differences, we aimed to compare areas with high and low infection rates within our country. At the end of September and the beginning of October 2020, just before the second wave, blood samples were collected from over 350 children aged 6 to 15 in Limburg. The children came from Pelt in the northern part of Limburg, where the COVID-19 cases had remained relatively low in the preceding months, and from Alken, a municipality about 40 kilometers further south that had experienced one of the largest COVID-19 outbreaks in our country during the first months of the epidemic.

While only 4.4% of children from Pelt had antibodies against SARS-CoV-2, this percentage was more than three times higher in Alken, at 14.4%. None of the children had experienced severe symptoms. Only four of them had confirmed COVID-19 infections, indicated by a positive PCR test, all of which had occurred in March 2020. Based on this, we assumed that the majority of immunity in the children was the result of infections during the first wave, most likely before the schools closed. Interestingly, all children with antibodies in Pelt were over 12 years old, and no participating primary school students from Pelt had antibodies. In contrast, in Alken, there was no difference in the prevalence of antibodies between children under and over 12 years old.

This study seems to indicate that younger and older children are equally susceptible when virus circulation is high. This in turn suggests that if a significant portion of the population older than 12 years were vaccinated, thus keeping viral circulation low, younger children would also be protected. However, other factors need to be taken into account, such as how fast acquired immunity (after infection or vaccination) wanes and how new variants behave.

This study, conducted just after the summer of 2020, provides a first local insight into the circulation of the virus among children during the first wave. However, as with the very first data from abroad, blood analyses that detect the presence of antibodies naturally provide very little information about how infections occur within classes and schools or among households. To analyze the dynamics of outbreaks in schools, PCR tests and contact tracing are needed to map active cases of COVID-19, not only in schoolchildren but also in their parents and other family members, as well as the teachers in the classrooms.

This is exactly what we did in collaboration with colleagues from Liège during the fall of 2020. We closely monitored children, parents, and staff from the same primary school from September 21 to the end of December. They were tested for SARS-CoV-2 using a saliva sample every week. Those who tested positive were contacted and isolated. Family members of the infected person also went into quarantine. The guidelines regarding the duration of quarantine changed several times during the study period, from 14 days to 7 and ultimately 10 days.

A total of 181 participants took part: 63 children, 82 parents, 15 teachers, 17 other school staff, and 4 individuals who were both parents of one of the participating children and teachers at the same school. All children were between 5 and 13 years old, and all adults were between 30 and 59. They belonged to a total of 47 families and 13 classes. One-quarter of the participants, precisely 45, tested positive during the study period. This included 13 children and 32 adults. After taking clusters within classes and households into account, there was no difference in the infection rate between children and adults.

Since all data were available on when someone's symptoms began and the contacts between the participants (and the specific RNA sequence of the virus), we could reconstruct who had infected whom. Most infections occurred between teachers and staff and between children at school. In some cases, there was further transmission within households, namely children infecting their parents or teachers infecting their partners. In most cases, infections within a household originated at school, despite additional hygiene measures in place there.

In contrast to some other data, this study in Liège suggests that a significant amount of virus is indeed being transmitted in schools. There are several reasons why we detected more infections in this study than expected. First and foremost, the testing method is different from the conventional PCR test with a nasal swab and may be more sensitive in detecting infections in young children who do not always cooperate well with a painful test. Additionally, all participants were tested very regularly, even if they showed no symptoms. Finally, local virus circulation may play a role, as the region around Liège was particularly hard-hit during the second wave.

Nevertheless, based on this data, we can conclude that regular testing and contact tracing are important to keep schools open safely. The role of schools in the spread of the coronavirus should also be taken into account when implementing a vaccination strategy — in other words, in answering the question of whether children should be vaccinated and when.

How should we test?

We used individual-based models to simulate virus spread in elementary schools and determine the best testing strategy to slow it down while keeping schools open: symptomatic testing (testing everyone with symptoms and isolating those who test positive), reactive testing (the same as symptomatic testing but also testing and potentially isolating all classmates of a person who tests positive), or repetitive testing (testing everyone in school every week and isolating those who test positive). Each testing strategy involves closing a class after more than two cases and closing a school after more than 20 cases, each time for 10 days.

We considered both the total number of new infections and the total number of missed school days for those infected. The first parameter is important to account for further spread (at home and elsewhere) and any complications due to an infection, while the second parameter aims to assess potential learning loss.

Of all the scenarios, repetitive proactive testing proved to be the best method for minimizing the number of new infections as much as possible. Testing only when symptomatic or in response to an infection was not very efficient. Since, especially in children, many infections remain asymptomatic, the virus can continue to spread under the radar. To limit learning loss, repetitive testing could be combined with a control strategy in which only infected students need to isolate themselves, rather than the whole class or school.

In an ideal world, this repetitive test would be a PCR test, as it is the most accurate. To reduce costs, tests could be pooled by class. The PCR test can be performed using a saliva sample, rather than the more painful nasal swab.

The model takes into account many aspects but, as always, it has limitations. We assumed perfect adherence to the testing strategies, but we also evaluated the consequences if adherence is significantly lower. We used virus parameters as we know them for the Wuhan variant of the coronavirus, as well as for the delta variant. We considered the lower likelihood of detecting symptoms in children than in adults and also incorporated current immunity and vaccination data into the model. Teachers were assumed to have contact only with the students in their class. To account for possible interactions between teachers at school, an additional structure, namely a virtual teachers' room, would need to be added to the model.

We developed this model for the spread of COVID-19, but it could be quickly adapted to other viruses or pathogens, or to new variants, as we did for the delta variant. This will allow us to swiftly simulate the best approach for the next epidemic or pandemic, where children may be more or less contagious or experience more serious symptoms.

What about babies?

While we had many questions about virus transmission among school-age children, the situation for babies and toddlers was even murkier. To get a better idea of their role, we needed to gather data on infections in the context of childcare. Fortunately, we could do so through an already running study screening children between 6 months and 2.5 years old for the presence of antibodies against pneumococci, bacteria that can cause lung, ear, and brain membrane infections.

Thanks to two additional nose swab samples from each nostril, more than 1,000 children from 100 daycare centers were tested for COVID-19 between May 2020 and February 2022. Even in early March 2020, before the lockdown and daycare closures, 84 children from eight different daycare centers in six provinces were tested. At that time, no children tested positive for COVID-19. In May and June 2020, 95 children from 12 daycare centers were tested, but again, all results were negative. In a later phase between November 2020 and May 2021, 751 babies and toddlers from 95 daycare centers were tested, and they all still tested negative.

The following winter, between November 2021 and February 2022, 11 out of 299 children did test positive, with nine out of 42 daycare centers affected. Six cases were detected before New Year 2022. These children were infected with the delta variant. The five positive cases in 2022 were all confirmed as infections with the omicron variant. They included four girls and seven boys, all between 9 and 26 months old. Nine of them had mild symptoms—a runny nose—but this was also the case for about half of the children who tested negative. There were no geographical clusters, with three infections in Brussels, six in Flanders, and the remaining two in Wallonia.

Although this research involved a one-time test, these figures still strongly suggest that there is little risk of transmission among babies and toddlers in daycare centers. Even without strict measures (masks were never mandatory), we only found positive cases when there was a high level of virus circulating in the broader population.

What role does higher education play in the spread of the coronavirus?

Studies from the United States have shown that college campuses are hotspots for virus transmission among young adults. Unlike schoolchildren, college students in higher education have a more extensive mix of high-risk social contacts, large-group classes, parties, sports activities… Some students also have jobs alongside their studies, etc.

Add to this the rather unique phenomenon of students traveling from all corners of the country to their dorms on Sunday evening and returning home with their laundry on Friday evening, and you have all the ingredients for a potential virological disaster, especially in early October 2020.

A survey by the independent Leuven student magazine Veto, involving more than 700 students, provided a sobering outlook. More than half of the surveyed students either commute as frequently as before the crisis, and 40% report that they rarely or never follow the university's dormitory guidelines aimed at limiting viral transmission.

We investigated the role higher education students played in the autumn wave of 2020 and found a correlation between the number of COVID-19 cases in a municipality and the proportion of students in higher education during certain periods. Flemish municipalities with a larger share of students had more COVID-19 cases during the summer months, especially among those aged 70 and older, but this effect decreased later. In Wallonia and Brussels, this effect increased toward September, affecting all age groups but was most pronounced in the 18 to 39 age group. The most significant increase in COVID-19 cases occurred in October, particularly in Wallonia. This data suggests that the reopening of higher education may have been an important factor.