The role of children in household transmission of COVID-19: a systematic review and meta-analysis

Objectives To explore household transmissibility of SARS-CoV-2 in children in new-variants dominating periods. Methods Through retrieval in PubMed and Embase, studies were included in two parts: meta-analysis of the household secondary attack rate (SAR) and case analysis of household pediatric infections. Results A total of 95 articles were included: 48 for meta-analysis and 47 for case analysis. Pediatric COVID-19 only comprised a minority of the household transmission. The total pooled household SAR of child index cases and contacts were 0.20 (95% confidence interval [CI]: 0.15–0.26) and 0.24 (95% CI: 0.18–0.30). Lower household transmissibility was reported in both child index cases and contacts than in adults (relative risk [RR] = 0.64, 95% CI: 0.50–0.81; RR = 0.74, 95% CI: 0.64–0.85). Younger children were as susceptible as the older children (RR = 0.89, 95% CI: 0.72–1.10). Through subgroup analyses of different variants and periods, increased household SAR was observed in children (Wild: 0.20; Alpha: 0.42; Delta: 0.35; Omicron: 0.56), and no significant difference was found in household SAR between children and adults when new variants dominated. Conclusion Although children were found not to be dominant in the household transmission, their transmissibility of SARS-CoV-2 appeared to be on the rise as new variants emerged.


Introduction
As of 29 April 2022, there have been 510.2 million confirmed COVID-19 cases and 6.2 million confirmed deaths globally, and people Therefore, understanding of the role of children in the household transmission of SARS-CoV-2 is still evolving, a further analysis is necessary. This study aims to (1) assess the prevalence of pediatric COVID-19 in family clusters; (2) estimate the household secondary attack rate of children in different periods and variants; (3) compare the transmissibility of SARS-CoV-2 in different age groups and explore its potential determinants.

Methods
This systematic review and meta-analysis were conducted following the Preferred Reporting Items for Systematic Reviews and Meta-Analyses (PRISMA) guidelines and the protocol was registered on PROSPERO (CRD42022313960).

Definition
A household transmission cluster was defined as a group of ≥2 confirmed COVID-19 cases in cohabiting individuals where the diagnosis of 6 cases occurred within 2 weeks of each other. The index case, the primary case, was defined as the first person in households to be infected with SARS-CoV-2. Household contacts were defined as family members or close relatives who had unprotected contact with the index case but did not necessarily live together. The transmissibility of SARS-CoV-2 was empirically estimated by the secondary attack rate. The household secondary attack rate (SAR) was defined as the number of household secondary cases divided by total household contacts. Children were individuals <18. Notably, as for studies dividing the age groups by 10 years, individuals of 10-19 years old were included in the child group.

Search Strategy and Eligibility Criteria
A systematic retrieval was performed on two databases (PubMed and EMBASE) from inception to 20 April, 2022, using the key search terms: COVID-19, SARS-CoV-2, family characteristics, household transmission, and so on [details in Table S1], with no restriction on the language, date, study type or place of publication. Non-primary documents and modeling studies were excluded.
Depending on the study type and provided data, studies were included in two parts: case analyses of household pediatric infections and meta-analysis on the household SAR. Case analyses mainly included case reports focusing on individual household transmission of SARS-CoV-2. The personal information of index cases, household contacts, family relationships, and the disease progression of COVID-19 7 cases must be provided. While the SAR meta-analysis mainly included descriptive studies that had reported the household SARS-CoV-2 secondary attack rate in different age groups: at least two of the following were required: household contacts, household secondary cases, and SAR. Studies with insufficient data or possible duplicate cases were excluded.

Data Extraction and Quality Assessment
Two authors (Tian and Zhang) independently extracted the following information from each included study: author, country, study type, study period, case definitions, testing protocol, contact-tracing methods, demographic characteristic, COVID-19 data (exposures, index cases, household contacts, secondary infection cases, SAR), potential factors, and so on. Disagreements were resolved through consultation with the third author (Chen). To critically appraise the methodological quality of included studies, the JBI critical appraisal checklist was applied (Critical Appraisal Tools JBI, 2022). Each included study was scored independently by two authors (Tian and Zhang) and given an average point. Studies were ranked as high quality if they were scored >=10, medium if they were scored >=7-9, and low if they were scored <7.

8
All analyses were performed using R 4.1.2 software. The secondary attack rate (SAR) and its relative risk (RR) were calculated for each study. SARs were pooled with a random intercept logistic regression model after a Freeman-Tukey double arcsine transformation and RRs were pooled using a random-effects model with Der-Simonian and Laird weights. The within-study variation was estimated with the 95% confidence interval (CI) and the Higgin's and Thompsons' I 2 was used to assess heterogeneity between studies. Subgroup analyses were conducted to explore the source of heterogeneity. Publication bias was detected through performing funnel plot and Egger's test. P < 0.05 was considered statistically significant in all tests.

Results
As shown in the flow diagram in Figure S1, a total of 1632 records were identified through the data search and 236 articles were retrieved for full-text assessment. Finally, 95 articles were included in our analysis: 48 articles for household SAR meta-analysis and 47 articles for case analysis. Studies included in SAR meta-analysis was demonstrated in Table 1, of which 26/48 were of high quality and 22/48 were of medium quality according to the quality assessment in Table S2, and the full details of family clusters included in case analyses were shown in Table S3.
All included studies reported household COVID-19 from 18 countries and regions with a total of 1153693 participants (834613 adults and 9 319080 children).

Case Analyses of Household Pediatric COVID-19
In the case analysis of pediatric COVID-19, 47 articles were included, identifying 78 household transmission clusters. As shown in Table 2, only 10.3% (8/78) familial clusters were identified with a pediatric index case. These pediatric index cases only led to 7.7% (16/207) of all secondary cases, compared with the 92.3% of secondary cases caused by the adult index cases. Child contacts were identified as 29.8% (84/282) of all household contacts and reported in 60.3% (47/78) familial clusters. The child secondary infections only accounted for 30% (62/207) of all secondary infections, compared with the 70% as adults.

Household SAR of Child Contacts
Secondary infections of the pediatric household contacts were identified in 41 studies and the pooled SAR was 0.24 (95% CI: 0.18-0.30, I 2 =100%) [ Figure 1]. Publication bias was reported upon examination of a funnel plot (Egger Test, P = 0.021) [ Figure S2].

Household SAR comparison between Child and Adult Contacts
In household SAR comparison between child and adult contacts in 37 studies, children were demonstrated to be less likely to be infected by SARS-COV-2 than adults when exposed to household index cases (RR=0.74, 95%CI: 0.64-0.85, I 2 =97%) [ Figure 3]. No obvious publication bias was found in the funnel plot of Figure S6 (Egger test, P = 0.31).
Subgroup analyses of the comparison were performed on research periods and SARS-CoV-2 variants in

Household SAR of Child and Adult Index cases
A total of 18 studies reported the respective SAR of child and adult index cases in familial clusters. The estimated SAR of the child index case was 0.20 (95% CI: 0.15-0.26, I 2 =100%). As for the adult index cases, it was 0.36 (95% CI: 0.27-0.46, I 2 =100%). Compared with the adults, the child index cases were significantly associated with a lower possibility to transmit SARS-CoV-2 to their family members (RR = 0.64, 95% CI: 0.50-0.81, I 2 =96%) [ Figure 4].

Discussion
Analyses of the household transmission of SARS-COV-2 will certainly facilitate a better understanding of the transmission chain and Based on these previous research, larger quantities of articles were included in our study. With more timely articles, more comprehensive analyses were conducted. Besides the total pooled household SAR of child contacts and index cases, subgroup analyses were also performed in different SARS-CoV-2 variants and different periods, as well as the transmissibility comparison between child and adult contacts. To the best of our knowledge, almost no previous meta-analyses have been conducted on the pediatric household transmission of different SARS-CoV-2 variants.
As our results show, both the child index cases and secondary cases only comprised a small proportion of the household transmission in case analyses, which suggested that children were unlikely to be the main source of SARS-COV-2 in familial clusters. In the total unclassified results of SAR meta-analyses, lower household transmissibility was demonstrated in both pediatric index cases and contacts compared to adults.
This was consistent with these previous meta-analyses (Grijalva, et al., 2020;Madewell, et al., 2021Madewell, et al., , 2020Zhu, et al., 2021). All these 15 findings are implying that children are less vulnerable to SARS-COV-2 than adults. Just as large previous data shown, the old adults also had a higher SAR than the young. Contrary to Zhu's analysis (Zhu, et al., 2021), significant difference was found between children <10 and children >10 in our analyses and a recent population-based cohort study also suggested a higher transmissibility of SARS-COV-2 in younger children than the older (Paul, et al., 2021). However, this difference still lacked statistical power because of the limited included studies and relatively little advantage, and negative results also occurred in our comprehensive analyses. Therefore, more future studies are still required.
Notably, some new findings were found in subgroup analyses on household SAR of different periods and SARS-COV-2 variants. In the early period of the pandemic (the Wild type mainly dominated during 2019-2020), a relatively low household SAR was observed in children (10%-30%) and child contacts usually had lower transmissibility than adults. However, with the emergence of some new variants (Alpha and Delta) from the beginning of 2021, household SAR in children seemed to be higher than before (30%-40%). Consistent with our results, many epidemiologic studies had pointed out that children and adolescents had become more susceptible to these new variants (Allen, et  Interpretation of the result in determinant assessment should be more conservative considering the high heterogeneity. A higher SAR was observed in the symptomatic index cases than those asymptomatic. Extensive evidence has proved that mild or asymptomatic patients are less contagious than those with typical clinical symptoms (Cevik, et al., 2021;Heald-Sargent, et al., 2020;Luo, et al., 2020). A larger household size might be associated with a lower SAR. One possible reason may be that large families usually have a low average age and young people tend to be less susceptible. The spouse relationship emerged as a susceptible group in our result. Chaw suggested that intimate relationships with frequent interaction and prolonged proximity in a closed environment were risk factors . But negative outcome occurred in the parent-child relationship, which might result from the children's low vulnerability. Household contacts with comorbidities or female contacts were found to be more susceptible, which were also reported in many large population studies (Lyngse, et al., 2022;Prunas, et al., 2022).
There are several limitations of our systematic review and meta-analysis. Firstly, articles included in case analyses were limited and relatively insufficient, larger datasets or more scientific methods are necessary for a more accurate prevalence assessment. In meta-analyses, some included studies were of retrospective or cross-sectional type and the information of index cases and contacts was mainly obtained from contact tracing datasets. Therefore, the determination of the case status might be uncertain, especially asymptomatic child index cases were often mistakenly identified as secondary cases, distorting transmission pathways. The epidemiological information was self-reported and subject to recall bias and response bias. In addition, the SAR would be overestimated for not excluding infection resource outside the household and was also underestimated in studies where only the symptomatic contacts were tested. Because of the data insufficiency, many other potential determinants associated with the SAR were not investigated in detail, such as the incubation and infectious periods and public lockdown policy, and subgroup analyses of child index cases were not conducted either. Lastly and most importantly, high unexplained heterogeneity in our analyses constituted an important obstacle when interpreting the results. This might be attributed to the great variation in the design of studies: different definitions of index cases and contacts, inconsistent testing protocols and follow-up time, sociodemographic factors, and so on. Many

2021).
All of these implied a multitude of related factors and substantial differences among populations. Therefore, the generalizability of our results is limited: compared with the quantitative results, the qualitative conclusions might be more reliable.

Conclusion
Although children were demonstrated not to be dominant in the household transmission, their transmissibility of SARS-CoV-2 appeared on the rise as new variants emerge. Given the potentially serious complications of pediatric COVID-19, vaccination research and implementation in children remain a must.