Abstract
Importance The degree to which children and young people are infected by and transmit the SARS-CoV-2 virus is unclear. The role of children and young people in transmission of SARS-CoV-2 is dependent on susceptibility, symptoms, viral load, social contact patterns and behaviour.
Objective We undertook a rapid systematic review to address the question “What is the susceptibility to and transmission of SARS-CoV-2 by children and adolescents compared with adults?”
Data sources We searched PubMed and medRxiv up to 28 July 2020 and identified 13,926 studies, with additional studies identified through handsearching of cited references and professional contacts.
Study Selection We included studies which provided data on the prevalence of SARS-CoV-2 in children and young people (<20 years) compared with adults derived from contact-tracing or population-screening. We excluded single household studies.
Data extraction and Synthesis We followed PRISMA guidelines for abstracting data, independently by 2 reviewers. Quality was assessed using a critical appraisal checklist for prevalence studies. Random effects meta-analysis was undertaken.
Main Outcomes Secondary infection rate (contact-tracing studies) or prevalence or seroprevalence (population-screening studies) amongst children and young people compared with adults.
Results 32 studies met inclusion criteria; 18 contact-tracing and 14 population-screening. The pooled odds ratio of being an infected contact in children compared with adults was 0.56 (0.37, 0.85) with substantial heterogeneity (95%). Three school contact tracing studies found minimal transmission by child or teacher index cases. Findings from population-screening studies were heterogenous and were not suitable for meta-analysis. The majority of studies were consistent with lower seroprevalence in children compared with adults, although seroprevalence in adolescents appeared similar to adults.
Conclusions There is preliminary evidence that children and young people have lower susceptibility to SARS-CoV-2, with a 43% lower odds of being an infected contact. There is weak evidence that children and young people play a lesser role in transmission of SARS-CoV-2 at a population level. Our study provides no information on the infectivity of children.
Question What is the evidence on the susceptibility and transmission of children and young people to SARS-CoV-2 in comparison with adults?
Findings In this systematic review and meta-analysis, children and young people under 18-20 years had an 435 lower odds of secondary infection of with SARS-CoV-2 compared to adults 20 years plus, a significant difference. This finding was most marked in children under 12-14 years. Data were insufficient to conclude whether transmission of SARS-CoV-2 by children is lower than by adults.
Meaning We found preliminary evidence that children have a lower susceptibility for SARS-CoV-2 infection compared with adults, although data for adolescents is less clear. The role that children and young people play in transmission of this pandemic remains unclear.
Background
The degree to which children and young people under 20 years are infected by and transmit the SARS-CoV-2 virus is an unanswered question.1-3 These data are vital to inform national plans for relaxing social distancing measures including reopening schools.
Children and young people account for 1-3% of reported cases across countries4-8 and an even smaller proportion of severe cases and deaths.5,9 Children appear more likely to have asymptomatic infection than adults and analyses based upon symptom-based series underestimate infections in children.
The role that children and young people play in transmission of SARS-CoV-2 by is dependent upon their risk of exposure, their probability of being infected upon exposure (susceptibility), the extent to which they develop symptoms upon infection, the extent to which they develop a viral load sufficiently high to transmit and their propensity for making potentially infectious contact with others, dependent upon numbers of social contacts across age-groups and behaviour during those contacts.
Different study types may provide useful information on susceptibility and transmission in children compared with adults, yet each is open to bias. Contact-tracing studies with systematic follow-up of all contacts to estimate secondary attack rates (SAR) in children and adults can provide strong evidence on differential susceptibility. Findings from some contact tracing studies suggest that children have lower SARS-CoV-2 SAR than adults,10 although others have found no difference by age.11 One study from South Korea has suggested adolescents but not children may have higher SAR,12 although a separate analysis of child cases from the same population identified minimal transmission from these cases.13
Population-screening studies may identify infection through viral RNA detection or antibodies indicating prior infection. However the prevalence of SARS-CoV-2 in children in a population is not a direct indicator of susceptibility or transmission as the expected prevalence depends on exposure, susceptibility, proportions of children in the population, mixing rates among children and between adults and children and timing of social distancing interventions that disrupt mixing.
A number of authors have concluded that children and young people may be less susceptible to SARS-CoV-2,2,14 although there are multiple sources of bias in each study type which can complicate straightforward analysis. In contact-tracing studies, testing of only symptomatic contacts will introduce significant bias, as will seroprevalence studies drawn from clinical contact studies (e.g. primary care) or residual laboratory sera. Many studies undertaken quickly during the pandemic are under-powered to identify age-differences.
We undertook a systematic review and meta-analysis of published and unpublished literature to assess the susceptibility to SARS-CoV-2 in children and adolescents compared with adults. We limited this review to contact-tracing studies and population-based studies as these are likely to be most informative and least open to bias.
Methods
Our review question was “What is the susceptibility to SARS-CoV-2 by children and adolescents compared with adults?”
We undertook a rapid systematic review and included contact tracing studies or prevalence studies in published or preprint form and including data from a national public health website reporting government statistics and studies. Studies were required to provide data on proven SARS-CoV-2 infection (PCR or serology) and report either rate of secondary infections in children and young people compared with adult or infection prevalence or seroprevalence in children and adolescents separately to adults.
We excluded reports of single household/institution outbreaks; studies of hospitalised patients, clinical studies and cohorts defined by symptoms; studies of unconfirmed cases i.e. cases based on self-report or symptoms, including contact-tracing studies where only symptomatic contacts were traced; modelling studies or reviews unless these reported new data; and prevalence studies with ascertainment based upon clinical contact and seroprevalence studies of residual sera, as these are likely to under-represent children
Where studies were drawn from populations that overlapped, we excluded studies where the time periods overlapped but included studies where time-periods did not overlap. We did not include in this review seroprevalence studies only in children as these did not allow comparison with adults.
We searched two electronic databases, PubMed and the medical preprint server medRxiv on 16 May 2020 and updated this on 28 July 2020. We used the following search terms in PubMed: (“COVID-19”[tw] OR “2019-nCoV”[tw] OR “SARS-CoV-2”[tw]) AND ((child* OR infant*) OR (“transmission”[tw] OR “transmission” [mh]) OR (“Disease Susceptibility”[tw] OR “susceptibility”(mh)) OR (“epidemiology”[tw] OR “epidemiology” [mh]) OR (“contact tracing”[tw] or “communicable disease contact tracing”[mh])). In medRxiv we undertook separate searches for ‘child and covid-19’, ‘covid-19 and epidemiology’, ‘covid-19 and susceptibility’ and ‘covid-19 transmission’ as more complex Boolean searches are not available.
Figure 1 shows the PRISMA flow diagram.
PRISMA flow diagram for search
One researcher (RV) screened studies on title and abstract to identify potentially eligible studies for full-text review. Full text studies were then reviewed by two researchers for eligibility and data were extracted independently by two researchers (RV and OM or CW). We hand-searched cited references in all potentially eligible studies for additional studies and identified additional studies through authors’ professional networks.
Data were extracted on country, study type, study context with regards social distancing measures and school closures at the time of the study, case definition, testing method, sampling method, and infection rates in adults and children.
Methodological quality of included studies was assessed independently by 3 authors (OM, CW, RV) based on a critical appraisal checklist for prevalence studies.15 We assessed risk of bias using two additional criteria: whether symptomatic contacts (in contact-tracing studies) or individuals (population-screening studies) were more likely to participate than asymptomatic ones; and whether the obtained sample was >75% of the intended sample. Studies were categorised as high quality if they met all quality criteria and had low risk of bias on both criteria; medium if they had low risk of bias on 1 or more criteria and met ≥5 of 7 quality criteria; low if they had met <5 quality criteria; or Uncertain if multiple domains could not be scored.
Analysis
Contact tracing and population prevalence studies were considered separately. Random effects meta-analysis with restricted maximum likelihood estimation was undertaken using the meta commands in Stata 16 (St ataCorp; College Station, TX). Odds ratios were used as the primary metric for contact tracing studies. Prevalence ratios were used as the primary metric in population-based studies. We planned subgroup analyses using restricted maximum likelihood based upon quality of study and age of children / adolescents.
We followed the PRISMA guidelines in reporting findings.
No funding was obtained.
Findings
The PubMed search resulted in 3465 and the medRxiv search in 10,461 studies, of which 113 and 90 respectively were examined in full text and 16 studies included (Figure 1). We identified a further 6 studies through reference-checking and 10 studies through professional networks. In total 32 studies were included (Table 1) with quality/bias assessments shown in Appendix Table 1. Eighteen were contact-tracing studies (CTS) (3 were school CTS), and 14 were population-screening studies. Two were high quality, 22 medium, 7 low and one uncertain.
Characteristics of included studies
Contact tracing studies
Six were from mainland China, two from the USA and one each from Taiwan, Japan, South Korea, Israel, the Netherlands, Brunei and India, with school CTS from Australia, the Ireland and Singapore. Lower secondary attack rates (SAR) in children and young people compared with adults were reported by 11 studies; 5 from provinces of China, including Hunan10,16 Hubei,17,18 and Beijing;19 and 6 studies from other countries, including Taiwan,20 Japan,21 the USA,22,23 Israel24 and the Netherlands,7,25 although confidence intervals were wide in some studies.
No significant differences in SAR by age were reported in four studies: from Guangdong province, China,26 Brunei27 and the states of Tamil Nadu and Andhra Pradesh in India28 with one study from South Korea reporting high SAR in <19 year-olds.12 In three of these, SAR in younger children were low compared with adults but those amongst teenagers were as high as or higher than adults. 12,27,28
We undertook a random effects meta-analysis of SAR in children and young people compared with adult, with data able to be included from 14 studies. We combined data on children and young people <20 years and adult age-groups >20 years, thus odds ratios (OR) and prevalence rates for adults may differ from those reported in studies. The pooled OR estimate for all contact-tracing studies of being a child with secondary infection compared with adults was 0.56 (0.37, 0.85) with high heterogeneity (95%) (Figure 2).
Pooled estimate of odds of being an infected contact among children compared with adults for all contact tracing studies
In meta-analysis of 8 CTS grouped by age of child (Figure 3), the pooled OR for children <12-14 years was 0.52 (0.33, 0.82), significantly lower than adults, whereas for adolescents this was non-significant (OR=1.23 (0.64, 2.36). Chi-square test suggested this group difference was significant (chi-2 =4.54, p=0.033).
Pooled estimate of odds of being an infected contact among children and among adolescents compared with adults for contact tracing studies
When only the 8 medium/high-quality (low risk of bias) studies were examined, this finding was no longer significant (OR 0.68 (0.41, 1.11), however the difference in estimates between low and medium/high quality studies was not significant (p=0.202). (see Appendix Figure 1).
We hypothesised that CTS including only household contacts might provide a clearer indication of the relative susceptibility to infection of children versus adults because all contacts within households might be assumed to receive a similar exposure to infection from index cases. A post-hoc analysis by type of contacts (Appendix Figure 2) showed studies of household contacts gave a lower pooled odds ratio (0.41 (0.22, 0.76)) than did studies of all contacts (0.91 (0.69, 1.21)) (between group variance; df=l, chi2= 5.31, p=0.021).
Three studies undertook contact-tracing in schools. A state-wide population-based CTS in educational settings in Australia before and during school closures29 found that 27 primary cases (56% staff) across 25 schools or early-years nurseries resulted in 18 secondary cases in 4 settings, including an outbreak of 13 in one early-years setting initiated by a staff member with no evidence of child to adult transmission. The SAR was 1.2% (18/1448) overall, 5/1411=0.4% excluding the early-years outbreak and 2.8% (18/633) in those tested. Other national CTS undertaken in schools in the Republic of Ireland30 and Singapore31 before schools closed identified very few secondary cases in schools.
Population screening studies
Data from prevalence studies for children and young people compared with adults is shown in Figure 4. We did not undertake a meta-analysis of population-screening studies, given the important differences in the populations, epidemic time-points and methodologies involved.
Ratios of the prevalence of SARS-CoV-2 infection in children and young people compared with adults in population-screening studies
Four studies reported virus prevalence. National prevalence studies from Iceland32 and Sweden33 undertaken while primary schools were open, showed lower prevalence amongst children and young people than adults, as did a municipal study from Italy34 undertaken just before lockdown while schools were open. However a nationally-representative survey from England covering lockdown and the subsequent month identified no significant differences by age.35
10 studies reported seroprevalence, 3 being nationally representative. A lower seroprevalence was identified in children and in some instances adolescents compared with adults in a number of studies, including a nationally representative study in Spain (ENE-COVID-19),36 a Dutch nationally-representative study (Pienter Corona study),7,37 and city or regional studies from Iran,38 the USA,39 Switzerland40 and Japan41 although no difference by age was found in a survey in 133 sentinel cities in 26 Brazilian states.42 Two community-based studies following localised outbreaks found lower seroprevalence amongst children and young people than adults in Lombardy, Italy43 and Thuringia, Germany,44 with a second German post-outbreak study finding no overall association with age.45
Examination of seroprevalence findings in children separately to adolescents (Appendix Figure 3) suggested that seroprevalence lower than adults amongst younger children (<10 years) but not in adolescents, although this was not formally tested.
Discussion
We identified 37 studies from 23 countries that met our eligibility criteria and provided information on susceptibility to and transmission of SARS-CoV-2 in children and young people compared with adults. We excluded studies and study types open to very significant bias, yet studies were predominantly of medium and low quality, with only two high quality studies. The majority of studies were from middle and high-income countries in East Asia and Europe.
We found preliminary evidence from 15 contact-tracing studies that children and young people have lower susceptibility for SARS-CoV-2 infection than adults, with a pooled odds ratio of 0.57 (0.39, 0.83). This estimate was little changed when only medium or high quality studies were examined, although power was reduced and the confidence interval included one. Only one study13 found a higher odds of infection in 0-19 year olds than adults, although this finding was confined to 10-19 year olds. When studies were categorised by age of the children, lower susceptibility appeared to be confined to younger children (less than 14 years), who had a 48% lower odds of infection compared with adults aged ≥20 years. The age bands of the studies were not aligned making direct comparisons challenging.
Data from population-screening studies were heterogenous and were not suitable for meta-analysis. Findings consistent with lower seroprevalence in 0-19 year olds compared with adults were reported by two national studies, one regional study and all of the municipal post-outbreak studies, although confidence intervals were wide in some cases. Two virus prevalence studies similarly reported lower infection rates in ≤20 year-olds. In contrast, other studies reported no age-related differences. No studies reported higher prevalence in children and adolescents. Examination of seroprevalence findings in children separately to adolescents showed that the majority of studies were consistent with lower seroprevalence in children compared with adults, although seroprevalence in adolescents appeared similar to adults in all studies.
The findings from the CTS and prevalence studies are largely consistent in suggesting that children below approximately 12-14 years are less susceptible to SARS-CoV-2 infection, resulting in lower prevalence and seroprevalence than adults. Data specifically on adolescents are sparse although consistent with susceptibility and prevalence more similar to adults. Our findings on susceptibility are similar to a modelling analysis by Davies et al.,46 who estimated that those under 20 years were approximately half as susceptible to SARS-CoV-2 as adults.
We found few data that were informative on the onward transmission of SARS-CoV-2 from children to others. Data from the large Australian school contact-tracing study suggest that, at a population level, children and young people might play only a limited role in transmission of this pandemic. This is consistent with the data on susceptibility noted above, i.e. suggesting that lower rates of secondary infection mean that children and young people have less opportunity for onward transmission. There is evidence of transmission from children to others in households and in schools, and there have been reported outbreaks in schools.47,48 Other very small studies in Ireland30 and Singapore31 have found low numbers of secondary cases resulting from infected children attending school. This is consistent with a national South Korean study, which found the SAR from children to household members was extremely low.13 The available studies suggest children and young people play a lesser role in transmission of SARS-CoV-2, in marked contrast to pandemic influenza.49
Limitations
Our study is subject to a number of limitations. We remain early in the pandemic and data continue to evolve. It is possible that unknown factors related to age, e.g. transience of infection or waning of immunity, bias findings in ways we don’t yet understand. Some studies were low quality and nearly all included studies were open to bias. The secondary infection rate in some CTS was low and this may represent an underestimate of the unmitigated household attack rate of SARS-CoV-2 as transmission chains were cut short because of strict control measures.50 Most of the CTS were undertaken when strict social distancing measures had been introduced, e.g. closures of schools and workplaces, restriction of travel. This would have reduced contacts outside the home, especially contacts between children, but it may have increased contacts between children and adults by increasing the household contact rate. The number of contacts nominated and traced for 0-19 year olds was low compared with adults in some studies,12,28 which may have introduced bias. We identified 3 CTS from Guangdong province11,51,52 which were excluded as they overlapped with Liu et al.,26 however findings were unchanged if these studies were included. We included two recent large CTS from India28 and South Korea12 however numbers of children and data quality appeared low, making firm conclusions difficult.
For population screening studies, the numbers of children tested was small in most of the studies, and was frequently less than the 15-25% of the population that are < 18 years in most countries. This likely reflects lower recruitment of children and may be a source of bias, although the direction of this bias is unclear. Age-differentials in sensitivity of swab or antibody tests may also confound findings. Interpreting the observed prevalence and seroprevalence studies requires thorough quantification of social mixing and transmission between age groups and how that changed during lockdowns and social distancing interventions.
Summary and implications
There is preliminary evidence that children under 12-14 years have lower susceptibility to SARS-CoV-2 infection than adults, with adolescents appearing to have similar susceptibility to adults. There is some weak evidence that children and young people play a limited role in transmission of SARS-CoV-2 however this is not directly addressed by our study.
We remain early in our knowledge of SARS-CoV-2 and further data are urgently needed, particularly from low-income settings. These include further large, high quality contact-tracing studies with repeated swabbing and high-quality virus-detection and seroprevalence studies. Studies which investigate secondary infections from child or adolescent index cases in comparison to secondary infections from adult index cases are particularly needed in order to assess transmission. Monitoring of infection rates and contact-tracing studies within child-care and school settings will also be important. A range of serological studies are planned in many countries and these need to be sufficiently powered to assess differences in seroprevalence across different age groups and include repeated sampling at different time periods as social distancing restrictions are lifted. We will continue to update this review, including further data as available and updating preliminary data from some included studies.
Data Availability
All data available from included studies, all available as open access
Declarations
Ethics
No ethical approvals were required for these secondary analyses of existing datasets.
Consent for publication
Not applicable
Funding
No funding was received for this review.
Access to data
Russell Viner had full access to all the data in the study and takes responsibility for the integrity of the data and the accuracy of the data analysis.
Availability of data and materials
All included articles are on public access – see Appendix Table 1 for hyperlinks.
Competing interests
All authors declare they have no competing interests.
Funding
No funding obtained for these analyses.
Author’s contributions
RV and RME conceptualised the review. RV developed the search terms with the assistance of JT. RV undertook the initial searches. Data extraction was undertaken by RV, RE and OM. Quality assessment was undertaken by OM, CW and RV. Meta-analyses were done by RV with input from GM-T and JT. FvdK supplied additional data from one study. RV, CB and RME led the writing of the paper. All authors contributed to editing the paper and approved the final manuscript.
Footnotes
The review was updated to 28 July. Our review now contains 32 articles.
References
