key: cord-0687933-8eb8ixns
authors: Ulyte, A.; Radtke, T.; Abela, I. A.; Haile, S. R.; Berger, C.; Huber, M.; Schanz, M.; Schwarzmueller, M.; Trkola, A.; Fehr, J.; Puhan, M. A.; Kriemler, S.
title: Clustering and longitudinal change in SARS-CoV-2 seroprevalence in school-children: prospective cohort study of 55 schools in Switzerland
date: 2020-12-22
journal: nan
DOI: 10.1101/2020.12.19.20248513
sha: f0d2f5b8f5bcbbac310771b0f0ef110ce2da54df
doc_id: 687933
cord_uid: 8eb8ixns

Background and aims: The facilitating role of the school setting in SARS-CoV-2 infection spread is still debated and the potential of school closures to mitigate transmission unclear. In autumn 2020, Switzerland experienced one of the highest second waves of the SARS-CoV-2 pandemic in Europe while keeping schools open, thus offering a high-exposure environment to study SARS-CoV-2 infections in schools. The aim of this study was to examine longitudinal change in SARS-CoV-2 seroprevalence in children and the evolution of clustering within classes and schools from June to November, 2020, in a prospective cohort study of school children in the canton of Zurich, Switzerland. Methods: Children from randomly selected schools and classes, stratified by district, were invited to participate in serological testing of SARS-CoV-2 in June-July and October-November 2020. Parents of children filled questionnaires on sociodemographic and health-related questions. 55 schools and 275 classes within them were enrolled, with 2603 children participating in the first, and 2552 in the second testing (age range 6-16 years). We evaluated longitudinal changes of seroprevalence in districts and investigated clustering of seropositive cases within classes and schools. Results: Overall SARS-CoV-2 seroprevalence was 2.4% (95% CrI 1.4%-3.6%) in summer and 4.5% (95% CrI 3.2%-6.0%) in not previously seropositive children in late autumn, leading to estimated 7.8% (95% CrI 6.2%-9.5%) of ever seropositive children, without significant differences among lower, middle and upper school levels. Among the 2223 children with serology tested twice, 28 (40%) of previously positive were negative, and 109 (5%) previously negative became seropositive. Seroprevalence was not different between school levels or sexes, but varied across districts (1.7% to 15.0%). Between June-July and October-November 2020, the ratio of diagnosed to newly seropositive children was 1 to 8. At least one newly seropositive child was detected in 47 of 55 schools and 125 of 275 classes. Among the classes with at least 50% and [≥]5 children tested, 0, 1-2 or >=3 seropositive children were present in in 90 (58%), 57 (37%) and 7 (5%) out of 154 classes, respectively. Class level explained slightly more variation of individual serological results (standard deviation (SD) 0.97) than school level (SD 0.61) in the multilevel logistic regression models. Symptoms were reported for 22% of seronegative and 29% of newly seropositive children since summer. Conclusions: Under a regimen of open schools with some preventive measures in place since August, clustering of seropositive cases occurred in very few classes and not across entire schools despite a clear increase in seropositive children during a period of high transmission of SARS-CoV-2.

school-based studies of SARS-CoV-2 seroprevalence, and takes place in a country with one 108 of the highest SARS-CoV-2 incidences worldwide, offering unique insights into the change in 109 clustering of seropositive cases within classes and schools, and the association with self-110 reported symptoms. The aims of the study were to estimate the longitudinal change of 111 seroprevalence, clustering within schools and classes, and to assess the association with 112 reported symptoms. 113 114 . CC-BY-NC 4.0 International license It is made available under a perpetuity.

is the author/funder, who has granted medRxiv a license to display the preprint in (which was not certified by peer review) preprint

The copyright holder for this this version posted December 22, 2020. ; https://doi.org/10.1101/2020.12.19.20248513 doi: medRxiv preprint nationally coordinated research network in Switzerland, Corona Immunitas [21] . The canton 119 of Zurich, in which the study was based, comprises 1.5 million linguistically and ethnically 120 diverse residents, approximately 18% of Swiss population, residing both in urban and rural 121 settings. In 2020, physical attendance of schools was interrupted only between March 16 122 and May 10. Implemented preventive measures (e.g., masks for school personnel and 123 children in secondary schools, reduction of some large group activities) have fluctuated 124 since then. However, the schools have been in continuous operation since the start of the 125 school year on August 17 to the end of 2020. Once an index case was identified in a school, 126 children and school personnel were quarantined based on contact-tracing of close contacts. 127

Full classes were quarantined only when two or more infected students were identified 128 within a class. All schools were obliged to develop a plan and implement specific preventive 129 measures by August (e.g., masks for teachers and children >12-years-old, distancing rules in 130 class-and teachers' rooms, tapering of school breaks, no mixing of classes, ban of group 131 gatherings such as excursions and camps beyond class units, no parents on school grounds) 132 to mitigate transmission but they varied from school to school. Common to all schools was 133 the requirement to keep children at home if they are sick beyond very mild symptoms such 134 as runny nose or mild cough, masking for adults in the school from 19 October and 135 additionally for children of secondary schools (above 12-years-old) from 2 November. 136 137 Population 138

Primary schools were selected randomly from the list of all schools in the canton of Zurich, 139 stratified by region, and the geographically closest secondary school (often, in the same 140 school building) matched. From 156 invited, 55 schools agreed to participate. Classes within 141 participating schools were selected randomly, stratified by school level: grades 1-2 in lower 142 level (attended by 6 to 9-year-old-children), grades 4-5 in middle level (attended by 9 to 13-143

year-old children) and grades 7-8 in upper school level (attended by 12 to 16-year-old-144 children). Invited grades were selected to ensure that the same cohort of children will 145 remain in the classes until April 2021 (children in grades 3, 6 and 9 often change the class 146 and school in the next year). We aimed to enroll at least three classes and at least 40 147 children in each school level at the invited schools. Major exclusion criterion was suspected 148 . CC-BY-NC 4.0 International license It is made available under a perpetuity.

is the author/funder, who has granted medRxiv a license to display the preprint in (which was not certified by peer review) preprint is the author/funder, who has granted medRxiv a license to display the preprint in (which was not certified by peer review) preprint Seropositive children in October-November, excluding those who also tested seropositive in June-July All tested children in October-November, excluding those who also tested seropositive in June-July T1+T2 seroprevalence (ever seropositive in summer and/or autumn) Proportion of children who had been infected with SARS-CoV-2 by October-November 2020, as reflected by ever being seropositive Children tested seropositive at least once in June-July or October-November (ever tested seropositive)

All tested children in October-November and seropositive children tested in June-July who did not participate in the testing in October-November * also including 18 children tested (serological results available for 12) in late August-early September is the author/funder, who has granted medRxiv a license to display the preprint in (which was not certified by peer review) preprint

The copyright holder for this this version posted December 22, 2020. ; https://doi.org/10.1101/2020.12.19.20248513 doi: medRxiv preprint denominator for the proportion of classes with clusters. Clustering of seropositive children 208 within classes was further examined by comparing our study results with a simulation of a 209 hypothetical study with the population (classes and numbers of children tested at T2) 210 identical to this study. In the simulation, independent chance of seropositive results (i.e., no 211 association of seropositive cases within a class) was assumed, equal to the observed 212 proportion of T2 seropositive among all T2 tested children. By comparing the number of 213 classes with clusters actually observed in our study and that in the simulation, we could 214 estimate if such number of clusters would be likely to be observed by chance. 215

Semi-structured interviews with the principals of schools with classes with observed 216 clusters of T2 seropositive children were performed after T2 testing to further investigate 217 the detected clusters. Interview questions covered numbers of diagnosed and quarantined 218 teachers and children in the affected classes, potential temporal sequence of infections and 219 other related circumstances. 220

To determine whether schools or specific classes explained more of the variance in 221 seropositivity, individual level serology results were modeled in a multilevel logistic 222 regression, with sex and school level (as a proxy for age) as fixed effects. Three models were 223 compared: with random effects for school level, with random effects for class level, and 224 with both random effects. is the author/funder, who has granted medRxiv a license to display the preprint in (which was not certified by peer review) preprint

The copyright holder for this this version posted December 22, 2020. 

At T2 testing, serological results were available for 731 children from lower school 244 level (median age 8, age range 6-10 years), 863 children from middle school level (median 245 age 11, age range 8-13 years), and 909 children from upper school level (median age 14, age 246 range 11-16 years). 1287 children were female, 1211 male and 5 reported other gender. 247

Median participation rate at T2 within a class was 47% (interquartile range 30%-62%). 248

Serological test results were positive for 74 children in T1 testing. For these children, 249 in T2 testing serology result was positive for 42 (60%, median age 10 years, age range 7-14 250 years, symptoms reported in 31/41 (76%) in January-July) and negative for 28 (40%, median 251 age 10 years, age range 7-14 years, symptoms reported in 17/24 (71%) in January-July), is the author/funder, who has granted medRxiv a license to display the preprint in (which was not certified by peer review) preprint

The copyright holder for this this version posted December 22, 2020. ; https://doi.org/10.1101/2020.12.19.20248513 doi: medRxiv preprint 3.2%-6.0%) (Figure 2 ). The proportion of children having had SARS-CoV-2 infection (ever 258 seropositive, T1+T2) by autumn was 7.8% (95% CrI 6.2%-9.5%). The seroprevalence at T1 259 and T2 does not add up to the proportion of ever seropositive children (T1+2) as the 260 populations included in the numerator and denominator of these three outcomes are not 261 exactly the same (see Table 1 and explanation in Appendix2). The range of newly 262 seropositive children in the districts of the canton of Zurich was 1.7%-15.0%, and the range 263 of proportion of ever seropositive children 3.5%-21.2% (Figure 2 ). T2 seroprevalence in 264 lower, middle and upper school levels was 4.4% (95% CrI 2.7-6.7%), 5.0% (95% CrI 3.0-7.4%) 265 and 3.9% (95% CrI 2.1-6.2%), respectively, and T1+T2 seroprevalence in lower, middle, and 266 upper school level was 8.5% (95% CrI 6.1-11.4%), 8.0% (95% CrI 5.7-10.7%) and 6.4% (95% 267 is the author/funder, who has granted medRxiv a license to display the preprint in (which was not certified by peer review) preprint

The copyright holder for this this version posted December 22, 2020. ; https://doi.org/10.1101/2020.12.19.20248513 doi: medRxiv preprint children was 1 to 13. 280

The number of newly T2 (ever T1+T2) seropositive children within a school-level 281 ranged from 0 to 12 (0 to 14), and within a class from 0 to 10 (0 to 11). At least one newly Distribution of newly and ever seropositive children within tested classes with more than 5 288 children and more than 50% of the class tested is shown in Figure 3 . is the author/funder, who has granted medRxiv a license to display the preprint in (which was not certified by peer review) preprint

The copyright holder for this this version posted December 22, 2020. ; https://doi.org/10.1101/2020.12.19.20248513 doi: medRxiv preprint Table 2 . Assuming a uniform 5.4% seropositivity rate across all 302 tested children, and numbers of children tested within classes as observed in this study, a 303 simulation study showed that 7 or more clusters would be expected by chance in 14% of 304 repetitions, with median expected number of classes with such clusters 4 (95% CrI 1 -9). 305

Thus, even if infections within classes were not associated, in a population with the classes 306 structure and total number of seropositive children as in this study, we would expect to see 307 4 clusters of three of more seropositive children in a class. 308 is the author/funder, who has granted medRxiv a license to display the preprint in (which was not certified by peer review) preprint

The copyright holder for this this version posted December 22, 2020. ; https://doi.org/10.1101/2020.12.19.20248513 doi: medRxiv preprint July and November 2020 310 311 Information was obtained through semi-structured interviews with school principals, including the 312 information about the probable index case. Class from School 2 had a participation rate of 47% and 313 therefore is not shown in Figure 3 .

In the multilevel logistic regression models of individual serology results of T2 316 serology, school level (as a proxy for age) and sex were not significant predictors. Estimated 317 standard deviation (SD) of random effects for the school decreased once the random effect 318 for class was added (from 0.606 (95% CI 0.400-0.864) to 0.467 (95% CI 0-0.786)). Estimated 319 SD of random effects for class remained relatively stable and was bigger than the effect of 320 school once the random effect of school was added (from 0.969 (95% CI 0.692-1.290) to 321 0.767 (95% CI 0.371-1.170)). 322

Symptoms between the summer break and November 2020 were reported in 21.8% 323 seronegative and in 28.7% newly seropositive children (T2). The distribution of individual 324 symptoms is depicted in Figure 4 . Although reported rarely in general, only loss of smell or 325 taste was more frequent in seropositive than in seronegative children, (3/101 (3.0%) vs 326 4/1923 (0.2%)). The most frequently reported symptoms in seropositive children were 327 headache (13.9%), runny or congested nose (11.9%), sore throat (11.9%), and fatigue 328 . CC-BY-NC 4.0 International license It is made available under a perpetuity.

is the author/funder, who has granted medRxiv a license to display the preprint in (which was not certified by peer review) preprint

The copyright holder for this this version posted December 22, 2020. ; https://doi.org/10.1101/2020.12.19.20248513 doi: medRxiv preprint is the author/funder, who has granted medRxiv a license to display the preprint in (which was not certified by peer review) preprint

The copyright holder for this this version posted December 22, 2020. is the author/funder, who has granted medRxiv a license to display the preprint in (which was not certified by peer review) preprint

The copyright holder for this this version posted December 22, 2020. ; https://doi.org/10.1101/2020.12.19.20248513 doi: medRxiv preprint decreased substantially since summer, from 1 to 89 [1] to 1 in 8 cases, meaning that 368 diagnosis of the infected children had dramatically improved. 369

Observed clustering of cases within a class does not necessarily signal an outbreak 370 (internal infection spread) has occurred in the school. In the seven classes with observed 371 potential clusters, seropositive children were likely not part of the same infection 372 transmission chain in at least two of the classes. In six classes, at least some of the 373 seropositive children were previously diagnosed or quarantined. The results of the 374 simulation showed that even if seropositive status was assigned to children of the study 375 population completely randomly, clusters of seropositive children would be observed in 376

seven or more of the tested classes with 14% probability. Thus, even if the seropositive 377 children were not associated within classes (i.e., infections completely independent of each 378 other), it would be not unreasonable to expect to see as many clusters as observed in this 379 study just by chance. 380

When clustering did exist, it seemed to be related to the class rather than to the 381 school, as suggested by the multilevel models. This could mean that, as could be expected, 382

infection is more likely to spread within a class rather than school (if at all). Potentially, the 383 random effect of the school would become even smaller once the incidence in the 384 community (district) is controlled for. This is another reason why focused class-based 385 quarantine measures may be more efficient than penalizing whole schools or even the 386 entire school system for localized clustering. is the author/funder, who has granted medRxiv a license to display the preprint in (which was not certified by peer review) preprint

The copyright holder for this this version posted December 22, 2020. ; https://doi.org/10.1101/2020.12.19.20248513 doi: medRxiv preprint prospective population level view, corresponding to school structure thanks to sampling on 403 the school and class levels. In addition, having measured the baseline seroprevalence in 404 June-July 2020, we were able to study the incidence of newly seropositive cases and their 405 clustering in classes and schools in autumn. The study had a very high retention rate, with 406 89% of enrolled children retested in autumn. Together with the newly enrolled children 407 from the same classes joining in autumn, the study had a high overall participation rate, 408 especially given that it included venous blood sampling in children. High participation rate 409 within a large proportion of classes allowed to study clustering on class level, which has not 410 been possible in other (rare) seroprevalence studies in children [33] . 411

The study has a few limitations. is the author/funder, who has granted medRxiv a license to display the preprint in (which was not certified by peer review) preprint is the author/funder, who has granted medRxiv a license to display the preprint in (which was not certified by peer review) preprint is the author/funder, who has granted medRxiv a license to display the preprint in (which was not certified by peer review) preprint is the author/funder, who has granted medRxiv a license to display the preprint in (which was not certified by peer review) preprint 

-negative result at the testing, + positive result at the testing, NA -results not available.

Cells in yellow are used for the analysis of seroprevalence at the corresponding time points. 563 * previously seropositive results that are seronegative at T2 are not counted, as immunity might be still 564 persisting, thus leading to a different further susceptibility and rates of infection than in those previously 565 seronegative. 566 ** previously seropositive results that are still seropositive at T2 are not counted in T2, as they are more 567 likely related to infections in spring (before T1) rather than in summer and autumn (between T1 and T2).

Raw proportion of seropositive results does not correspond to the seroprevalence estimated with 569 Bayesian hierarchical models (see Methods and Results), which additionally corrects for test accuracy and 570 population structure parameters. Estimated seroprevalence T1+T2 is higher than the sum of T1 and T2 is the author/funder, who has granted medRxiv a license to display the preprint in (which was not certified by peer review) preprint

The copyright holder for this this version posted December 22, 2020. ; https://doi.org/10.1101/2020.12.19.20248513 doi: medRxiv preprint

COVID-19 in 7780 pediatric patients: A systematic review

COVID-19 in children

Europe: a multinational, multicentre cohort study

COVID-19, school closures, and child poverty: a social crisis in the making

The Long-Term Health Consequences of Child Physical Abuse

Emotional Abuse, and Neglect: A Systematic Review and Meta-Analysis

Estimation of US Children's Educational Attainment and 491 Years of Life Lost Associated With Primary School Closures During the Coronavirus Disease

Susceptibility to SARS-CoV-2 Infection Among Children and 494

Adolescents Compared With Adults

Children are not COVID-19 super spreaders: time to go back to school

Role of children in household transmission of COVID-19

A large COVID-19 outbreak in a high school 10 days after 501 schools' reopening, Israel

SARS-CoV-2 Transmission and Infection Among Attendees 504 of an Overnight Camp -Georgia

Secondary transmission of COVID-19 in preschool and school 507 settings in northern Italy after their reopening in September 2020: a population-based study

Surveillance of COVID-19 school outbreaks

SARS-CoV-2 infection and transmission in educational settings: a 513 prospective, cross-sectional analysis of infection clusters and outbreaks in England

SARS-CoV-2 infections in Italian schools: preliminary findings 516 after one month of school opening during the second wave of the pandemic

Seroprevalence and immunity of SARS-CoV-2 infection in children and

Appendix 1 Comparison of individual serological results in study participants in summer (T1)