Transmission of SARS-CoV-2 by Children

Abstract

Background

Six months into the COVID-19 pandemic, children appear largely spared from the direct effects of disease, suggesting age as an important predictor of infection and severity. They remain, however, impacted by far-reaching public health interventions. One crucial question often posed is whether children generally transmit SARS-CoV-2 effectively.

Methods

We assessed the components of transmission and the different study designs and considerations necessary for valid assessment of transmission dynamics. We searched for published evidence about transmission of SARS-CoV-2 by children employing a narrative review methodology through 25 June, 2020.

Results

Transmission dynamics must be studied in representative pediatric populations with a combination of study designs including rigorous epidemiological studies (e.g. in households, schools, daycares, clinical settings) and laboratory studies while taking into account the social and socio-economic contexts. Viral load (VL) estimates from representative pediatric samples of infected children are missing so far. Currently available evidence suggests that the secondary attack rate stratified by age of the infector is lower for children, however this age pattern needs to be better quantified and understood.

Conclusion

A generalizable pediatric evidence base is urgently needed to inform policy making now, later when facing potential subsequent waves, and extending through a future in which endemicity alongside vaccination may become the enduring reality.

Age has been one of the most important factors in the prognosis of those affected by COVID-19 (e1). Differential SARS-CoV-2 infection, transmission and clinical manifestation by age have important implications for social policy decisions, such as closure of schools. By April 2020, interventions to limit SARS-CoV-2 transmission left over 90% of children confined at home worldwide (e2). Such closures reduce contacts not only between students, but also prevent parents from working (1) and have proven effective during influenza epidemics (2).

It quickly became clear, however, that children were relatively spared by SARS-CoV-2. They account for a small proportion of confirmed cases, with symptoms absent or mild and fleeting (e3) and have half the odds of infection compared to adults (3). We review evidence about transmission of SARS-CoV-2 by children, and how relevant parameters may be ascertained through epidemiologic and laboratory techniques. It is essential to know if children need to adapt their individual contacts with close family members at risk for severe COVID-19. This evidence is also required for rational policies that balance harms and benefits of education and recreation, as well prioritization of future vaccination efforts. Most acutely, continued interruption of formal education carries a huge social cost to health and development of children (4), and to social and professional functioning of parents (e4). Evidence to guide balancing of those risks against potential epidemic spread is therefore urgently needed.

Methods

We assessed the components of transmission and the different study designs and requirements necessary for assessment of transmission dynamics. We searched for the evidence base of transmission of SARS-CoV-2 by children using narrative review methodology (e5, e6) until 25 June 2020, including a PubMed search using the terms “child”, “SARS-CoV-2” and “transmission” (ebox).

eBOX. Narrative review methodology.

• We used search terms in Pubmed for the domains “child” (eg. “child” OR childhood* OR “children” OR schoolchild* OR adolescen* OR juvenil*) and “COVID-19” (eg. Coronovirus*[ALL] OR “CoV”[ALL] OR “CoV2”[ALL] OR “COVID”[ALL] OR “COVID19”[ALL] OR “HCoV-19”[ALL] OR SARS-CoV-2” [nm] OR “spike glycoprotein, SARS-CoV” [nm]), developed with the aid of an experienced librarian. We had followed up the literature since the start of the epidemic with a first systematic search on 3 May 2020.

• Starting from the articles found in this search, we used a snowball search strategy, scanning useful references and similar articles and retrieving those that were considered relevant. We re-ran the search on 3 June to detect missed articles, adding a third, however initially limited developed, search domain “transmission” (transmi* OR dynamic*).

• We searched in Google Scholar using the following search terms: allintitle:child|children|infants|pediatric|paediatric|school COVID OR COVID19 OR CoV OR nCoV OR “Corona Virus” OR Coronavirus OR CoV2 OR“SARS 2” transmission, and a second using dynamic instead of transmission. Last search was performed 25 June 2020.

• We searched medXriv and arXiv, focused on the period between the last search date of the published systematic review of Ludvigson et al (24) from 11 May until 26 June 2020 within the COVID-19 databases developed (with search on 8 June and 26 June).

• We searched in registries, given their results not being published except on National Public Health websites, as we became aware of publication of pediatric specific data, through following up communications in the media and medical literature.

• We ascertained reviews on the body of literature of COVID-19 in children released by Public Health Authorities (eg. France, Belgium, Canada [Quebec]).

• We have followed up reported literature on https://search.bvsalud.org/global-literature-on-novel-coronavirus-2019-ncov/ and the EPPI-centre, with focus on the domains of Transmission/risk/prevalence, with and without adding specific terms for child(ren). All searches were dated from December 2019 up to the dates given above.

• We screened the results on title and abstract for relevant information. No language restriction was applied. Articles in English, French, Dutch, Spanish and German were interpreted directly. Public Health pages were translated using Chrome Google Translate if other languages applied.

• No formal protocol of the narrative review was written out or registered, a full written out data collection form was not created at the beginning of the search. A search and terms log-book was kept.

• A narrative review comes with its own limitations. The process is less rigorous and less systematic than this of a systematic review and most importantly does not formally or by use of any instrument evaluate the quality of the evidence.

Components of Transmission

Disease transmission is a complex interaction between infectious and susceptible hosts, an agent and the environment (5). For a child to become a possible source of infection, they need to have been previously exposed to, replicate and effectively shed the infectious virus. Components of transmission dynamics, the spread of the infection over time, are biological, behavioral and contextual. Knowledge of all three components is required to define effective transmission and its evolution over time.

Children differ from adults in biological and behavioral characteristics related to transmission. For example, they differ in contact type, rate, duration and intensity and they primarily interact with other children (e7). It remains unknown which variables play critical roles in the transmission dynamics of SARS-CoV-2 (6). Current consensus is that respiratory droplet expulsion is key, especially during speech (7), but this has not been studied among children. There is presently insufficient evidence for feco-oral transmission. Context also matters, and most documented transmission occurs in adult settings like bars, conferences, meat-plants and ships (8). Like other infectious diseases, SARS-CoV-2 data suggest that ˜20% of the cases are responsible for ˜80% of local transmission (9). So far, these ˜20% appear to be predominantly adults, not children (8, e8).

Most work on transmission dynamics focuses on susceptible hosts and calculates risk of becoming infected. To know whether children transmit the SARS-CoV-2 virus as effectively as adults, one needs to know the secondary attack rate (i.e. number of new cases an initial case infects, per 100 exposed individuals [e9]) stratified by age of infector. This is the most direct measure of the infectiousness of a particular agent.

Epidemiological and Experimental Designs

The question of pediatric contagiousness is not a theoretical one, but a real-world phenomenon in the specific context of the current pandemic. A natural approach is via epidemiologic designs, observing populations directly and measuring chains of transmission to ascertain how many cases stem from an infected child.

The primary design for this purpose is the study of households, which are well-defined settings that offer the closest and most intense contact with and between children (e10, e11). These may be incidence or prevalence studies, using PCR or serology, or both. Pediatric transmission can also be studied in outbreaks at schools, daycares, or in clinical settings. In a subset of investigations, phylogenetic tools (e12) can also be used to enhance specificity. Further inference can be derived from intervention studies, using longitudinal data to evaluate the impact of prevention measures on reducing transmission. Finally, the epidemiologic profile is enhanced by population surveys, either of acute or past infection (10). These designs share the virtue of being direct observations of the series of events that culminate with transmission, and therefore provide direct answers to scientific questions about epidemic spread. The results of epidemiologic designs may be subsequently used in mathematical models, to expand inference to hypothetical and counterfactual scenarios, and for the purpose of forecasting (e4). In contrast to the artificial sterility of the laboratory, “epidemiology is the gold standard to measure transmission potential of patients” (11).

Because inference about transmission dynamics occurs at the population level, attention needs to be given to how and which study participants are recruited (Table 1) (12). Systematic differences in recruitment can lead to bias in population estimates, for example if only symptomatic children are followed, or only hospitalized cases are recruited (e13). This generates a skewed impression of the true frequencies of events and representativeness issues, and also leads to biases that arise from associations in the sample that are artifacts of the recruitment process (13). Analyses must also account for imperfect measurement of case status as a function of diagnostic test accuracy (14, e14). These potential biases may be exacerbated when they are age-dependent, for example if test accuracy varies by age, or pediatric cases are rare. Small sample sizes in addition result in imprecise estimates.

Table 1. Requirements for further studies to improve the evidence base on the epidemiology of SARS-CoV-2 transmission by children.

Requirement	Elaboration
Diverse study designs	– Household studies – Seroprevalence and sero-incidence studies – Prospective cohort studies, e.g. in schools and daycares – (Cluster-) Randomized controlled trials – Case-control studies – Impact evaluation—time-series and quasi-experimental – Population surveillance studies including children – Clinical laboratory studies, including phylogenetic data – Basic laboratory and animal studies, experimental designs – Mathematical modeling studies
Avoidance and quantitative assessment of biases	– Selection bias in recruitment of study population or stratification – Information bias from missing data or measurement error – Confounding bias from unadjusted risk factors correlated with exposure – Survivor bias from loss to follow-up – Type II errors from significance testing with low power
Study population generalizability and representativeness	Pediatric sample must be drawn representatively from a population of interest to avoid biases and generalize inferences, and this population must be adequately described
Consideration of biological, behavioral and contextual factors	Collection of contextual variables, including social determinants of health such as neighborhood, cultural practices and political economic factors that impact risks and behavioral responses
Diagnostic—laboratory testing interpretation	– Understanding different diagnostic testing methods and their limitations – Distinguishing between presence of RNA and infectious virus – Recognizing the importance of the pre-analytical phase of sample collection – Statistical interpretation of test results, taking into account the test characteristics and the effect of prevalence – Correction for imperfect reference standard

Results from laboratories, animal models and controlled experiments on SARS-CoV-2 and other respiratory pathogens contribute to knowledge about transmission, but issues of generalization remain central (e15). For example, how does transmission between hamsters translate to transmission dynamics and intervention effects in a 4^th grade classroom (15)? Experiments offer the advantage of controlled environments but never approximate real-world conditions. These studies therefore provide important mechanistic insights, but do not directly inform policy decisions. For example, environmental studies provided crucial information about duration of SARS-CoV-2 on different types of surfaces (16). To what extent transmission and infection occur from such media remains unknown, with epidemiology suggesting it may be relatively infrequent (e16). A recent study estimates that a minute of loud speaking generates at least 1000 virion-containing droplet nuclei that can remain airborne for more than 8 minutes (7). This suggests a plausible mechanism behind super-spreading events, but actual transmission cannot be predicted without accounting for real-world contextual factors such as airflow patterns (e17).

Laboratory Studies

The presence of sufficient infectious virus is a necessary condition for transmission, and this minimal number of viable organisms required to infect a non-immune individual is expressed as the median tissue culture infective dose. This is an in vitro phenomenon and remains unknown for SARS-CoV-2. It is cumbersome to measure and therefore high levels of viral replication are commonly quantified using a more convenient proxy, the viral load (VL) expressed in copies/ml (e18).

The VL provides a quantitative estimate of the amount of target RNA obtained from clinical samples and can be indirectly measured by PCR testing. Finding RNA does not imply presence of viable and replicating virus, however. Virus quantification by PCR testing uses a backwards calculation of cycle threshold (Ct) values (e19). The Ct value is the cycle number during the amplification phase when fluorescence of a PCR product can be detected above the background signal. The lower the value, the larger the initial amount of viral genetic material present in the sample. The minimal VL necessary to infect a secondary case is not known for SARS-CoV-2 and likely varies across hosts (17), although presumably more virus translates into increased infectivity. Pediatric cases with high VL have been reported (e20), but transmission is never guaranteed even with a large number of viral copies detected in the respiratory tract (18).

To quantify the amount of virus present in a typical infected child, one needs to account for pre-analytical and analytical sample and assay characteristics (19, e21). The respiratory specimen type, such as nasopharyngeal versus nasal swab, determines the sensitivity and specificity of a PCR test and measured VL will therefore vary (19). The ideal specimen to diagnose SARS-COV-2 infection in children has not been determined and could potentially be different from adults.

It has been shown that variation in VL is dependent on the timing of specimen collection in relation to initial exposure, (14, 19), with high VL early in the disease course (20). This is confirmed in samples from pediatric case-studies and case series (e20). Infectious virus is no longer detected after 7 days since onset of symptoms (21). An estimated 56.4% (34.9–78.0%) (e22) of adult transmission takes place in the pre-symptomatic phase (22). It remains unclear if this is mainly because of high peak VL at that moment, or because infector and infectee are unaware of the infection and therefore taking no precautions, or both, and if this plays out similarly in children.

To estimate the distribution of VLs in children attending schools, one needs to follow basic epidemiologic principles of generalizability (Table 1) (e23), using a study sample with representative demography, including age, sex, and social class. For valid population inference, selection into the study should not be based on presence of symptoms, severity of disease or presence of relevant co-morbidities. Diligent use of laboratory data in conjunction with epidemiological investigation illustrates the real potential to learn about transmission probabilities and effective transmission in relation to the skewed curve of VL (23).

Results

Knowledge Synthesis from Existing Studies

Published household transmission studies show that children are rarely the index case and investigations of cases and clusters suggest that children with SARS-CoV-2 seldom cause secondary cases (24). A review of household cluster studies compared household transmission during H5N1 influenza epidemics where 30/56 of the index cases were children, to the COVID-19 pandemic, where 3/31 clusters had a child who potentially infected a secondary case (25).

Most of the household cases published to date are from countries affected early in the pandemic and include confined children with consequently restricted exposure. In a Swiss study (26), 3/39 of households had a child who developed symptoms prior to other household contacts, but without evidence of transmission from the child. A preprint from Israel (e24) including 3353 people in 637 households, estimated children up to age 20 to be 85% as infective as adults (that is, relative 15% less infective). Published data from the first 13 families show an adult index case in all but one instance (27).

In settings where schools remained open or using data prior to closures, there is little evidence of outbreaks or major transmission into the community. In Australian schools (28) two children contracted COVID-19 after exposure to 9 infected students and 9 infected staff among 735 students and 128 staff. An outbreak in a French high-school is described among 15–17 year olds, with limited cases among sibling contacts (29). An Irish study (30) describes exposure to an infected child in primary school, two in high-school and 3 infected adults, and yet follow-up and testing detected no cases. In Sweden, where schools for children up to 15 years remained open, only hospitalization data are published (e25), during a period in which cross-sectional PCR surveillance showed overall positivity <3% in both children and adults (e26). Data free from selection bias on the impact on broader transmission and a useful comparison remain lacking.

Elsewhere, prevalence data using serology from published sero-surveys of population samples and their households (31), from residual blood sample data (e27) and national public health agency initiatives (32), reveal lower sero-prevalence in children than adults. A pre-print from the Paris region reports a high sero-prevalence of 10.7% among 605 children ages 0–15, sampled at visits to ambulatory pediatricians. This is a highly selected sample in which 17.1% and 32.3% of the included children had confirmed or suspected COVID-19 positive household contacts, respectively (33). This is substantially higher than the Paris average, and shows the sample to be unrepresentative of the overall population. Many more reports are posted by public health agencies (3) but do not adequately describe sampling and methods to infer pediatric rates.

Although VL estimates from representative pediatric samples remain unavailable, symptomatic children and adults showed similar VL in one descriptive study without formal statistical investigation (34). VL above a putative infectiousness threshold were present in 29.0% of 38 patients ages 0–6 years versus 51.4% among 3153 adults in a German study (35), but with insufficient information about sampling to suggest generalizability.

Modeling studies have shown effects confining children among a suite of interventions (36), however, even assuming equal infectiousness in children compared to adults, they show limited impact at the peak and the need for prolonged closure to control transmission (37). More recent models include lower susceptibility and infectivity and decreased impact (38).

Despite limited evidence, the general pattern emerges that transmission from children occurs, but contributes much less to evolution of the epidemic than do contacts between adults, and school-re-openings have not lead to transmission spikes in low transmission countries (e28). Possible changes in transmission among older students, as suggested using German data, warrant further assessment. Unlike other respiratory diseases such as influenza, it seems quite clear that the secondary attack rate from pediatric cases is substantially lower than for adults, and that mechanisms underlying this difference require elucidation.

Conclusion

Rather than simply a question of viral load, policy decisions such as school reopening are complex considerations that require balancing competing risks and benefits in the broader context of fear and uncertainty (e29). School closures have negative impacts on mental, educational, nutritional and social development, and disrupt relationships between children, peers and families (e30). They most adversely compromise children with special needs and those from marginalized households, exacerbating inequalities (39).

Transmission dynamics inevitably change over time (e31) and are modified by other interventions. The contribution of children to the spread of COVID-19 is therefore a highly contingent question. Many children still have limited exposure to infection, and become infected less frequently. When infected, they are generally less sick than adults. Six months into the pandemic, children have not shown any evidence of being a significant factor in its propagation (40). As societies worldwide relax or reintroduce restrictions, pragmatic studies to measure changes in transmission in various groups, particularly children, are needed more than ever.

Table 2. Summary of evidence base of transmission by children, by study design.

Study design, references	Setting	Main findings
Household cluster studies (25– 27, e24)	Proportion of pediatric index cases and secondary infections versus adult or H5N1 index cases	– Child was rarely index case and rarely caused secondary cases – Different from influenza – Children had lower relative infectivity
School outbreak investigations (28– 30)	High schools and primary schools; staff/student cases; PCR or seroprevalence in close contacts	– Primary school student and staff cases associated with few secondary cases – Older student and staff cases associated with no secondary cases – Sero-positivity in older students: 38.3%, teachers: 43.4%, staff: 59.3%; but much lower in household contacts – Irish cases each exposed 125–475 children and 25–28 adults at school
Sero-prevalence (31– 33, e27)	Population samples;residual clinical samples	– Geneva: In 455 children ages 5–19: 6% sero-positivity vs 8.5% in adults – Spain: 2.9% seropositive among 5–9 yo children vs 5.2% overall; lower in younger children
Clinical laboratory, viral load studies (34, 35)	RT-PCR and cell culture from symptomatic children	Cultivable SARS-CoV-2 in naso-pharyngeal specimens from 52% of 23 symptomatic children
Time-series (e28)	Comparative analysis of school-closures/re-openings; effect on growth rates daily hospitalizations or confirmed cases	– No increased transmission from school-reopening amidst low community transmission, nor from partial return of younger groups amidst higher transmission – No increased staff cases, but increased student cases on return, especially older students (Germany)
Modeling study (36– 38)	Epidemiological modeling school closures/re-opening; assumptions on susceptibility, contact and infectiousness	– Intervention consistently decreases number of cases and delays epidemic – Limited impact of school closures, – Impact lower compared to influenza

Abbreviation: yo = “years old”

Table 3. Relevant evidence base on transmission of SARS-CoV-2 by children.

Study (First author, reference number, country and time of data collection)	Study design, Study Population Sample size	Main findings*
Zhu Y. (25), China, Singapore, South Korea, Japan and Iran; December 2019 / March 2020 [house-style]	Summary of household cluster studies comparing proportion pediatric index cases and secondary infections during SARS-CoV-2 epidemic versus H5N1 epidemic; 87 households	In 3/31 (9.7%) SARS-CoV-2 household transmission clusters the index case was a child versus in 30/56 (54%) H5N1 clusters
Posfay-Barbe K. (26), Switzerland; March 2020 / April 2020	Household cluster study of first pediatric (<16 yo) sars-cov-2 pcr+ cases; 40 positive children and 111 household contacts	In 31/39 (71%) households adults assigned as index case versus 3/39 (8%) households with child symptoms prior to other household members
Dattner I. (e24), Israel, start detection of cases—May 2020; (Somekh E.(27) publication of initial families)	Household cluster study of households with at least 1 PCR+ case, all family members tested; 637 households, n=3353; children <21 yo n=1544; initial 13 families: 36 adults (>18 yo); 58 children	– 58 households with probability=1 that index case is adult; probability=0 in 34 households – Initial 13 families: one 14 yo infector. – Relative infectivity of children versus adults (modeled) was 0.85 (95% 0.65–1.10)
NCIRS (28), Australia: 5 March 2020–3 April 2020	School Outbreak investigation in 10 high schools and 5 primary schools of close contacts defined as “face to face contact for ≥15 minutes or in the same room for two hours with a case while infectious”; contacts: students n=735, staff n=128 exposure to 9 infected students and 9 infected staff	– 10 high-schools: 8 PCR+ students, 4 PCR+ staff, 598 students, 97 staff contacts; 235 contacts PCR test: all neg; 75 serology tested: 1 positive student – 5 primary high schools, 1 PCR+ student, 5 PCR+ staff, close contacts 137 students, 31 staff. PCR test 53: 1 PCR+; 21 serology tested: same student sero-positive
Fontanet (29), France: 13 January 2020–27 March 2020	School Outbreak investigation; serology testing (IgG); infection attack rate; 15–17 yo students (n=240), teachers (n=53), and students’ parents (n=211) and siblings (n=127) – total: n=661	Sero-positivity: 38.3% 15–17 yo pupils; 2.7% of siblings <14 yo, 10.2% siblings (all ages) and 11.4% parents, teachers 43.4%, 59.3% school staff
Haevey (30), Ireland; prior to 12 March 2020	School Outbreak investigation; n=1155 contacts (includes school contacts: children n=924, adults 101) of 6 PCR+ cases; contacts clinical follow up, symptomatic contacts PCR tested	– SARS-CoV-2 PCR+ cases: 1 primary school (10–15 yo), 2 high-school (10–15 yo) PCR+ cases; 3 infected adults; – Follow-up testing did not show school-related secondary cases – 2 secondary cases through adult-to-adult contact outside school environment
Hildenwall (e25), Sweden; 13 March 2020–14 May 2020	Cases during school opening, children 0–18 yo n=63 cases	– 63 admitted cases 0–17 yo, 30 non-incidental cases; hospitalization rate for COVID-19 cases of 9/100 000; – proportion of all SARS-CoV-2-positive admitted children 16–18 yo (10/63, 16%), for whom schools have been operating on distance, children 1–5 yo (11/63, 17%)
Stringhini (31), Switzerland (Geneva); 6 April 2020–9 May 2020	Sero-prevalence study population samples (SEROCoV-POP), ELISA adjusted for test performance, weekly during 5 weeks, all ages (n=2766, 1339 households), children (n=455)	– In 5–19 yo: 6% sero-positivity vs in 20–49 yo: 8.5%; – relative risk (95%CI) sero-positivity compared to 20–49 yo: 0.32 (0.11–0.63) 5–9 yo, 0.86 (0.57–1.22) in 10–19 yo
Cohen (33), France; 14 April 2020–12 May 2020	Cross-sectional (population-enrolled by ambulatory pediatricians) survey using PCR and sero-prevalence, during confinement, 0–15 yo (n=605)	PCR+ 11 (1.8%), sero-positive 65 (10.7%); positivity independent of number of children in the household
Estudio ENE second round (32), Spain;18 May 2020–1 June 2020	Sero-prevalence study, population wide, IgG rapid test positives; 0–19 yo n=11 730 (<1yo: 263, 1–4 yo: 1679, 5–9 yo: 2896, 10–14 yo: 3549, 15–19 yo: 3343)	5–9 yo: 2.9% (2.2–4.0) sero-positivity (95%CI) vs 5.2% (4.9–5.5) overall; (<1 yo: 2.2% [0.7–6.8], 1–4 yo:2.4% [1.5–3.8], 5–9 yo:2.9% [2.2–4.0], 10–14 yo: 3.8% [3.0–4.8], 15–19 yo: 3.8% [3.0–4.8])
Havers (e27), USA (Connecticut, South Florida, Missouri, NYC, Utah, Washington state); 23 March 2020–3 May 2020	Sero-prevalence study, convenience sample of residual clinical blood samples commercial laboratories; ELISA (Spike protein, screening plus confirmatory essay) adjusted for sensitivity and specificity; children 0–18 yo n= 1001, limited number <5yo, all ages n=11 933,	Sero-positivity (95%CI) for 0–18 yo vs all-ages, age-standardized: WA 0.66 (0.0, 2.52) vs 1.13 (0.70, 1.94) NYC 2.74 (0.9, 5.03) vs 6.93 (5.02, 8.92) FL 2.41 (0.00, 7.79) vs 1.85 (1.00, 3.23) MI 1.36 (0.00, 4.14) vs 2.65 (1.65, 3.86) UT (insufficient pediatric data) CT 0.81 (0.00, 2.89) vs 4.94 (3.61, 6.52)
L‘Huillier (34), Switzerland; early pandemic	Clinical laboratory, viral load study, RT-PCR and cell culture, n=23 (0–16 yo) symptomatic children	– Cultivable SARS-CoV-2 can be present in URT naso-pharyngeal specimens from symptomatic children, isolation success 12/23 (52%) children; – Median viral load at time of diagnosis was 3.0x106 copies/ml (mean 4.4x108, IQR 6.9x103–4.4x108).
Jones (35), Germany; 26 January 2020–May 2020	Clinical laboratory, viral load study (VL 250 000 copies threshold for isolation, infectious virus in cell culture at 5% probability=infectiousness), total n=3303 PCR+, ages 1–19 yo n=154	Infectiousness: 29.0% of 38 patients 0–6 yo, 37.3% of 0–19 yo (n=150) versus 51.4% among 3153 adults
Stage (e28), data from Germany, Denmark, Norway, Sweden; March 2020–June 2020	Time-series, comparative analysis of school-closures and school-reopening using case studies; effect on growth rates in daily hospitalizations or confirmed cases, accounting for – different age cohorts of students – timing of closure and reopening – background incidence	– No increased transmission in low community transmission settings after return of all students – No increased transmission in higher community transmission settings after partial return of younger year students – Increased transmission in students but not in staff on return more older students – School closures partially but non-differentially responsible for reduction growth rate
Ferguson (37) June 2020	Mathematical modeling study; epidemiological modeling of the impact of non-pharmaceutical interventions, including school closure – Assumption of children attack rates: equal as in influenza; children transmit as much as adults	– School closure is predicted to be insufficient to mitigate an epidemic in isolation – Combining social distancing of the entire population, case isolation, household quarantine and school and university closure predicted largest impact – Limited impact of school closures at the peak, prolonged closure necessary to maintain controlled transmission
Matrajt (36), modelling population Seattle metropolitan area; June 2020	Mathematical modeling study; Assumptions: children and adults equally infectious, children group with highest number of contacts; children <19yo	– Targeting to all age groups, delayed the epidemic the longest, >50 days – Targeting adults >60 yo and children resulted in 10 500 (45%) fewer cases than baseline at the epidemic peak; targeting adults-only: 11 000 (47%) fewer cases for 25% reduction in contacts in adults <60 yo and 21 000 (91%) fewer cases for 75% adult contact reduction
Davies (38), Data from: Canada, China, Italy, Japan, Singapore – model for Milan, Birmingham, Bulawayo; June 2020	Mathematical modeling study; population-wide estimates; assumptions: relative susceptibility toinfection in children 0–9 yo: 0.4; infectivity in children based on clinical fraction 0.29 in 0–9 yo. Infectivity dependent on clinical fraction and infectiousness 0 to 100% of clinical cases	– Interventions aimed at children had/showed relatively small impact on reducing transmission, particularly if the transmissibility of subclinical infections is low – Simulation 3 months of school closures: school closure decreased peak incidence 10–19% vs 17–35% for influenza

ELISA, Enzyme-Linked Immunosorbent Assay; CT = “Connecticut”; FL = “Florida”; IQR, interquartile range; MI = “Missouri”; NYC = “New York City”; PCR = “polymerase chain reaction for Sars-CoV-2”; URT = “upper respiratory tract”; UT =”Utah”; WA = “Washington State”; yo = “years old”; 95%CI = “95% Confidence Interval”

*A detailed critical review of the available evidence is beyond the scope of this table. Every study needs to be assessed for selection bias, information bias, confounding, survivor bias, interpretation of diagnosis assessment, the study population and context (timing in the epidemic, social context etc) where this was performed and the statistical soundness of the analysis. The main findings are to be interpreted by the reader taking at minimum these aspects into account.

Key Messages.

• Quantifying the transmission dynamics of SARS-Cov-2 in children is necessary for evidenced-based policy formation

• Transmission dynamics change over time and are dependent on characteristics of the contact, the susceptible and infected host, the virus and the environment; ultimately transmission is a complex function of biological, behavioral and contextual factors, including the social context

• Viral measurements form only one part of the answer of how infectious children are and need to be interpreted in light of the sample collection and processing techniques

• Basic epidemiological concepts such as representativeness and generalizability are necessary to go from data to policies

• We need both laboratory and epidemiological studies to answer essential scientific questions about the COVID-19 pandemic and to inform decision making about the optimal interventions to improve the wellbeing of children and adults