Roles of children, isolation measures and socioeconomic status in SARS-CoV-2 household transmission: a retrospective cohort study in a southern US city

Abstract

Objective

There has been significant debate about both the role of children in SARS-CoV-2 community transmission as well as the potential harm or lack thereof of the virus in the paediatric population. We aimed to better understand how children are impacted by a SARS-CoV-2 household infection and to identify behavioural and social factors related to infection.

Methods

A cross-sectional study was conducted on households with children and a recent SARS-CoV-2 infection in Atlanta, GA. Participants were enrolled from December 2020 through November 2021. Salivary IgG against SARS-CoV-2 spike and nucleocapsid antigens was measured and household and infection data collected and analysed.

Results

49 households participated, including 121 adults and 57 children with 118 (66%) individuals having been infected with SARS-CoV-2. Mean SARS-CoV-2 secondary infection rate was 0.55 (95% CI 0.41 to 0.69) and median, 0.67 (range 0–1.0). Incidence rate did not differ significantly between households with paediatric versus adult index cases (incidence rate ratio (IRR) 0.97, 95% CI 0.63 to 1.49). However, sharing a bedroom or bathroom was associated with a higher incidence of secondary infection (IRR 2.36, 95% CI 1.49 to 3.74). Neighbourhood deprivation index was associated with higher household SARS-CoV-2 incidence, but this did not reach statistical significance.

Conclusions

Child index cases appeared as likely to result in household secondary SARS-CoV-2 infections as adult index cases, while sharing rooms was associated with higher incidence of infection. This study highlights the importance of isolation measures to prevent household transmission of SARS-CoV-2 as well as counters common thought that children are less likely to transmit infection.

WHAT IS ALREADY KNOWN ON THIS TOPIC

• Studies have had mixed results on children’s roles in household SARS-CoV-2 infection, with many showing that children were less likely to transmit the virus.

WHAT THIS STUDY ADDS

• We found that paediatric index cases are equally likely to result in secondary infection as adult index cases. In addition, our study supports isolation within the home as a strategy to reduce SARS-CoV-2 transmission.

HOW THIS STUDY MIGHT AFFECT RESEARCH, PRACTICE OR POLICY

• This study dispels the myths that children are less likely to spread SARS-CoV-2 and less likely to get infected when exposed. These data can be used to support public health strategies for future outbreaks of SARS-CoV-2 or similar viruses.

Introduction

Since the beginning of the COVID-19 pandemic, there has been significant debate about the role of children in community transmission as well as the potential harm or lack thereof of the virus in the paediatric population.1,5 Widespread lockdowns and school closures muddied the waters in 2020 and a significant portion of 2021, limiting interpretation of the data in children and their influence on community and household transmission. Were children not as susceptible to SARS-CoV-2, or was their exposure inherently less frequent, thus making it difficult to truly measure their contribution to the spread of SARS-CoV-2? Possibilities include that they were, in fact, less susceptible or that their exposure differed from adults due to non-pharmaceutical interventions (NPI) or, less likely, due to an inherent biological difference. Children could also be less symptomatic or have a shorter duration of viral shedding.

Early studies from the first few months of the pandemic found relatively low secondary infection rates (SIRs) in the household, ranging from 6% to 51%,⁶ with mixed findings on whether younger ages, like children, were associated with lower SIR.6,9 Subsequent studies, especially those in Europe and the USA, found consistently higher SIR, up to 70% in minority households in North Carolina.34 10,16 Three US studies and a large meta-analysis of evidence through August 2020 compared SIR of adult and paediatric index cases.13,1517 SIR of paediatric versus adult index cases was similar in situations where the index case could be identified.13,1517 And a meta-analysis published in 2022 analysing 95 articles (48 studies included in the meta-analysis) demonstrated a pooled household SIR from paediatric index cases lower than that of adult index cases (0.20 (95% CI 0.15 to 0.26) vs 0.64 (95% CI 0.64 to 0.85)), but this difference did not hold up with more recent SARS-CoV-2 viral variants.¹⁸ Other evidence has shown that children can have similar or even higher (for those less than 5 years) viral loads as adults and that they shed transmissible virus.19,25

However, there are several limitations to pooling data from studies that use disparate methodologies. Many of the early studies did not use COVID-19 serology and may have missed asymptomatic infection. Not all studies required full participation of the entire household, which could have falsely inflated or deflated estimates of SIR. Larger studies using retrospective data from testing databanks have clear limitations since many infections go undiagnosed. Lastly, differences in country practices of isolation, like in China early in the pandemic, where most index cases were removed from the household, could also impact findings since SIR would be lower in households in which the index case was removed versus those in which the index case remained.

What these studies do tell us, though, is that children are clearly part of the transmission networks in households and the community. Children are index cases and infected contacts, and growing evidence disputes initial claims that the pandemic’s impact on children was minimal. As of 28 June 2023 when the US Centers for Disease Control (CDC) stopped updating the dataset, there were 1847 COVID-19 deaths in children aged 0–18 years.²⁶ In the most comprehensive meta-analysis of paediatric household transmission, there are very few determinants of transmission analysed: index case symptomatology, family relationships, co-morbidities, household size and sex,¹⁸ with larger comorbidities in contacts, symptoms in index case, and being a spouse of the index case all related to increased SIR. Larger households were associated with smaller SIRs.¹⁸ However, this is missing some other very important determinants of health and SARS-CoV-2 transmission such as socioeconomic status, mask use, vaccination and other factors that may affect transmission. Few studies take many factors into account at once. Our study location in Atlanta, GA, a diverse city ethnically, racially and economically, made an ideal setting to study household transmission, especially with several surges during the course of the first 1.5 years of the pandemic.

In this study, our objective was to better understand how children are impacted during a SARS-CoV-2 household infection, hypothesising that their role in household transmission or risk of infection did not differ significantly from adults. We also wanted to identify behavioural and social factors associated with risk of infection and spread within households. While new variants, like omicron, have continued to arise and prove more transmissible, this pre-omicron study provides critical information about risks of household transmission and how children have been impacted by the pandemic.

Methods

Overview and study sample

Index cases were identified through adult individuals who tested positive for SARS-CoV-2 infection at outpatient screening clinics at Emory Healthcare and the Southeast Permanente Group (Kaiser Permanente). We also advertised on social media and on Emory University listservs. Households were enrolled from December 2020 through November 2021 and were eligible if at least one member was under the age of 18 years and if one or more individuals tested positive for SARS-CoV-2 within 4–16 weeks prior to the study visit (dates of infection were late September 2020 through September 2021). For households to participate, all individuals living in the house at the time of infection were required to consent and enrol.

Data collection

Saliva samples were self-collected by participants (or parents of young children) using an Oracol (Malvern Medical Developments Ltd, Worcester, UK) device by rubbing a sponge in the gingival space on both sides of their mouth and were processed. This mode of collection focuses on gingival fluid. Saliva samples were stored maintaining cold chain and then shipped to the laboratory of Dr Christopher Heaney at Johns Hopkins Bloomberg School of Public Health. The salivary samples were then tested on a Luminex platform for IgG antibodies that bind to the SARS-CoV-2 nucleocapsid (N), receptor binding domain (RBD) and spike (S) proteins as previously described.²⁷ A salivary sample was defined as SARS-CoV-2 IgG positive if reactive to the GenScript N antigen and either the RBD or S Spike proteins as previously determined during validation of the assay.²⁷ A positive antibody to N antigen differentiates infection from vaccination as these antibodies would not be present from vaccination with mRNA or adenoviral-adjuvanted vaccines. Total IgG was measured in samples testing negative for SARS-CoV-2 IgG to exclude false negatives due to insufficient sample quality. For IgG against the N antigens, this assay has shown a sensitivity of 98% and specificity of 99%, and with kinetics, that is, duration of detection, also highly correlated to serum responses.^{27 28} Pisanic et al described the assay and its validation (as compared with serum measurements) of these gingival fluid multiplexed antibody results for several antigens and antibody subtypes, including IgG to N antigen used in this study.²⁷ Samples without meeting a validated threshold of total IgG were classified as indeterminate. Study participants completed individual and household surveys through RedCap after the study visit²⁹ and included individual demographics, symptoms, pre-existing medical conditions, exposures, vaccination status and a household survey describing infection timelines and isolation procedures in the household.

Definition of study variables

A case of SARS-CoV-2 infection was defined as a reported positive molecular test for SARS-CoV-2, a reported positive SARS-CoV-2 antigen test or a positive SARS-CoV-2 salivary antibody test as defined above. A household index case was defined as the individual who received the first confirmed SARS-CoV-2 test result or had the earliest symptom onset date among the household cluster. A household contact was defined as anyone living in the same residence as the index case at the time of infection. SIR was defined as the proportion of secondary cases in a household infected from the index case over the total number of household contacts. If an index case could not be identified, meaning that more than one individual in the family started having symptoms on the same day, or in the case of asymptomatic infections, tested positive on the same day, then they were not included in the risk factor analyses. Zip codes were used to determine the neighbourhood deprivation index (NDI) as a marker of socioeconomic vulnerability. The NDI is a measure of 13 socioeconomic variables of the community incorporating measures of community wealth, housing conditions, education among others.³⁰ Households were assigned to the bottom 20th percentile of NDI or above this cut-off.

Data analysis

Descriptive statistics were performed using frequencies, mean and median where appropriate. Bivariate statistical analysis using t-test measure of association compared SIR across different categories. Then, χ² or t-test was used to compare associations of different factors with either being an index case, an infected contact or a non-infected contact. Lastly, a multivariable Poisson regression model accounting for clustering of households was built using the main exposure variable (paediatric vs adult index cases) and those variables with a p value<0.20 on the unadjusted Poisson analyses. We then calculated the incidence rate ratios (IRRs) for each variable. Lastly, we also performed a logistic regression model to identify household factors related to SIR above the median and SIR below the median, while controlling for confounders. All statistical analyses were performed using SAS (Cary, NC, V.9.4).

Results

Participant characteristics and secondary infection rates

A total of 50 households provided informed consent; however, one household did not complete all parts of the study and was not included in the final results and analysis. Due to IRB restrictions, we were not able to document or record reasons for non-participation of eligible individuals who were contacted at Emory Healthcare and Kaiser Permanente. Out of the remaining 49 enrolled households, a total of 121 adults and 57 children were enrolled in this study (table 1), with 118 (66%) individuals having been infected with SARS-CoV-2 and 33 (57%) infected children. There was a total of 69 secondary infections with a mean SIR of 0.55 (95% CI 0.41 to 0.69) and median of 0.67 (range 0–1.0). On bivariate analysis, the mean SIR was not significantly different between households with a paediatric index case (0.57) and households with an adult index case (0.54), p=0.89 (figure 1). There were 13 index cases under the age of 18 (specific ages: 3, 5, 8, 10, 13, 15, 16 years).

Table 1. Demographic and clinical characteristics of individuals from the 49 final families participating in the study, describing frequency unless otherwise noted.

Variables	N	%
Total sample (n)	178	—
Age, median (IQR)	17.5 (8.0–40.0)	—
Participants less than 18 years old	57	32.0
Sex
Male	77	43.3
Female	101	56.7
SARS-CoV-2 infection confirmed (PCR or anti-nucleocapsid IgG)	118	66.3
SARS-CoV-2 negative test	50	28.1
SARS-CoV-2 test indeterminate*	10	5.6
Asymptomatic infection†	60	35.7
Race/ethnicity‡
Non-Hispanic White	97	54.5
Non-Hispanic Black	76	42.7
Hispanic, Latino, Spanish origin	8	4.5
Asian or Pacific Islander	14	7.9
Other race	2	1.1
Health insurance
Yes	173	97.2
No	5	2.8
Healthcare worker	14	7.9
Annual Household Income
Less than US$25 000	10	5.6
US$25 001–US$50 000	16	9.0
US$50 001–US$100 000	30	17.9
US$100 001–US$200 000	65	36.5
Greater than US$200 000	48	27.0
Declines to disclose/unknown	9	4.8
Neighbourhood deprivation index (NDI)
>20th percentile	114	64.0
Lowest 20th percentile	51	28.7
Unknown NDI	13	7.3
COVID-19 vaccine (2 doses)	35	19.7
Household infection time period
Pre-Delta	107	60.1
Delta	71	39.9

PCR: polymerase chain reaction test for SARS-CoV-2 viral detection.

Anti-nucleocapsid IgG: positive total immunoglobulin G to the nucleocapsid antigen.

Indeterminate: insufficient total immunoglobulin to determine presence/absence of anti-SARS-CoV-2 antibodies.

^*Indeterminate.

^†Total sample of 168 (10 not included as infection not able to be determined).

^‡Participants could choose more than one response for race/ethnicity.

Figure 1. Comparison of secondary infection rate (SIR) among (a) symptomatic index cases versus asymptomatic index cases, (b) paediatric versus adult index cases, (c) household masking most or all of the time versus some or none of the time and (d) sharing either a bedroom or bathroom in the household versus sharing neither. Total n=49 households.

Of the 178 participants, 76 (42.7%) identified as Non-Hispanic Black; 8 (4.5%) from Hispanic, Latino or Spanish origin; 14 (7.9%) identified as Asian or Pacific Islander and 2 (1.1%) identified as other. Most (97.2%) study participants had health insurance and a quarter came from households reporting annual household incomes of less than US$50 000. Among those infected with SARS-CoV-2, some of the most common symptoms reported were fatigue, headache, fever, nasal congestion, chills, dry cough and loss of smell and taste (online supplemental table 1). 60 (35.6%) individuals with SARS-CoV-2 infection were asymptomatic, of which 17 were diagnosed retrospectively by the salivary ELISA. At the time of enrolment, about a third of participants (n=60, 35.7%) had received at least one dose of the COVID-19 vaccine but only 35 (19.7%) had received the two-dose series (table 1).

Risk factors for household infection

Figure 1 pictorially shows differences in household SIR across the following variables: index case symptomology, age of index case and household isolation practices. For those households where members wore masks all or most of the time, SIR (0.42) was significantly lower than in households where members never, occasionally or rarely wore masks (0.76) (p=0.02). The SIR for those who shared at least one bedroom and/or bathroom (0.76) was significantly different from those who do not share a bedroom and/or bathroom (0.27, p=0.001).

Using a Poisson analysis to compare household incidence of individual infections of SARS-CoV-2, having a paediatric index case in the household was not associated with higher incidence of infection on unadjusted analysis (IRR 0.98, 95% CI 0.68 to 1.42, table 2). When paediatric index cases were divided into younger than 10 (ages 3–9) and older (ages 10–17), this did not statistically significantly alter the IRR, although the youngest group was associated with lower incidence (IRR 0.72, 95% CI 0.47 to 1.11) when compared with adult index cases, while the older children index case household had a higher incidence than households with adult index cases (IRR 1.56, 95% CI 0.85 to 2.83). In addition, children were infected at rates that did not differ significantly from adults.

Table 2. Factors associated with being an index case, infected non-index case or uninfected, using χ² analyses, comparing to the reference value of each variable.

Variable	Index case, n (%) N=49	Infected non-index, n (%) N=69	Uninfected, n (%) N=50	Infected non-index vs uninfected
				IRR	95% CI	P value
Age, mean, SD	31.4 (15.4)	24.3 (17.8)	24.9 (17.9)	--	--	0.87*
Age<18 years	13 (27)	20 (29)	15 (30)	0.98	0.68 to 1.42	0.93
Sex
Female	19 (39)	31 (45)	23 (46)	1.02	0.72 to 1.42	0.92
Race
Non-White	24 (50)	32 (46)	22 (44)	1.04	0.75 to 1.45	0.81
Paediatric index case				0.98	0.68, 1.42	0.91
Symptomatic index	29 (60)	43 (61)	27 (54)	1.14	0.80, 1.62	0.46
Comorbidity	12 (25)	19 (40)	9 (18)	1.24	0.85 to 1.79	0.27
Healthcare professional	6 (12.2)	6 (9)	2 (4)	1.32	0.73 to 2.39	0.35
Delta	21 (43)	27 (39)	19 (38)	1.02	0.72 to 1.44	0.91
Fully vaccinated (2 doses)	11 (23)	12 (17)	12 (24)	0.83	0.54 to 1.29	0.42
Always or frequently masked indoors	34 (69)	41 (59)	40 (80)	0.69	0.49 to 0.97	0.03
Members shared bathroom, bedroom or both	28 (57)	52 (75)	16 (32)	2.29	1.56 to 3.38	<0.001
Annual Income*:
Less than 50k	9 (19)	8 (12)	6 (14)	0.93	0.55 to 1.56	0.78
NDI group†
Bottom 20%	13 (29)	23 (36)	11 (23)	1.30	0.91 to 1.87	0.15

P values <0.05 considered significant and shown in bold. (49 households; 168 individuals).

^*Missing 8.

^†Missing 12.

NDI, neighbourhood deprivation index.

However, we found several statistically significant differences in IRR on unadjusted analysis (table 2). Living in a house where members frequently or always wore masks indoors was associated with reduced infection (IRR 0.69, 95% CI 0.49 to 0.97; p=0.03). Households where members shared a bedroom, bathroom or both were a factor strongly associated with secondary SARS-CoV-2 infection (IRR 2.29, 95% CI 1.56 to 3.38). Living in an area at the bottom 20th percentile of NDI was positively associated with infection among household contacts of the index case, but did not reach statistical significance (OR 1.30, 95% CI 0.91 to 1.87). The index case having symptoms and the likely variant were not associated with a higher incidence of secondary infection. Likewise, being vaccinated or having a comorbidity also was not associated with a higher or lower incidence of secondary infection.

The Poisson regression analysis used paediatric index case as the main exposure and adjusted for mask use, room sharing, low social vulnerability and clustering of the households. Age of the index case (paediatric vs adult) was not statistically associated with a higher incidence. Sharing a bathroom/bedroom retained significance in the adjusted model with more than twice the incidence of COVID-19 infection in individuals where household members shared rooms (IRR 2.36, 95% CI 1.49 to 3.74). Masking did not maintain statistical significance and being in the bottom 20% of NDI neighbourhoods was not associated with higher incidence.

Finally, we also conducted a logistic regression of household factors by dichotomising SIR to above the median of 0.67 and those below the median. All SIRs above the median were 1.0, so essentially the analysis compared those households where all contacts were infected to the rest of the households. Factors strongly associated with a high SIR (aOR>2) included sharing a bedroom or bathroom, identifying as non-white and having a vaccinated member, while those households with higher SIR were half as likely to mask in the household and live in an area above the 20% NDI. However, only sharing rooms reached statistical significance (OR 19.13, 95% CI 2.94 to 124.37), similar to the Poisson regression (online supplemental figure 1).

Discussion

Our study demonstrated a high SARS-CoV-2 SIR in households with children (mean of 0.55, 95% CI 0.41 to 0.69) when asymptomatic infections were measured through a salivary antibody assay. Not only was this higher than many SIRs reported in the literature,³¹ this study adds to the evidence, contrary to common thought in 2020, that the risk of infection from a paediatric index case was not statistically different from the risk from an adult index case on unadjusted and adjusted analyses. In addition, among household contacts of index cases, paediatric and adult infections were not statistically different (IRR 0.98, 95% CI 0.68 to 1.42).

Unlike our findings, most data from the first 1–2 years of the pandemic suggested that paediatric index cases were much less likely to lead to other household infections1,39 32 and that kids were less likely to get infected, research that supported reopening schools in some instances.⁵ These studies that identified fewer paediatric than adult index cases were likely impacted by NPIs (eg, school closures, lockdowns) and limited by methodologic factors such as limited sampling, undertesting of asymptomatic cases and use of large databases to identify cases and contacts. Additionally, many studies did not enrol the entire household, which may have resulted in underestimation of SIR.^{10 33 34} Despite these limitations and external confounders, studies by Pitman-Hunt et al and a meta-analysis in 2020 concluded that index cases in children were infrequent transmitters of SARS-CoV-2 and paediatric household contacts were less likely to acquire infection.^{33 35} Interestingly, even a study with an SIR of 0.59 from paediatric index cases concluded that children were not driving transmission because they only made up 8% of the index cases in their study.³⁶ While our SIR for paediatric index cases was very similar to that study, we had a higher percentage of paediatric index cases (27% of our households), showing that the epidemiologic context (local transmission, degree of societal opening, adults’ larger amounts of social interactions, study design) could explain a different number of paediatric index cases across studies more than an inherent resistance to infection.³⁶

Aside from comparing paediatric to adult household index cases, we also wanted to explore other factors that may influence household transmission such as use of NPI, race and socioeconomic factors like residence in a low NDI locale. We found an association with inconsistent or no use of NPI and a higher likelihood of SARS-CoV-2 infection, an association that was present in both unadjusted and adjusted analyses (tables2 3). Sharing a bathroom or bedroom was associated with a more than double incidence of secondary infection in the household. While not universal across studies, many showed statistically significant associations between NPI usage and lower onward transmission in households or sharing spaces and a higher SIR,337,39 as was the case in our study. And while some studies have found symptomatic index cases associated with greater household transmission,³² this was not the case in our study despite the fact that we had a high proportion of asymptomatic index cases (table 2). Two CDC studies had opposite findings, with one study of three states finding no difference in household transmission based on symptom status of index case. On the other hand, a different CDC study on household transmission after an outbreak at summer camp showed a strong association with symptomatic paediatric cases and household transmission^{22 40}

Table 3. Adjusted Poisson Regression analysis with incidence rate ratios of secondary SARS-CoV-2 infection in 49 households, controlled for clustering of households.

Variable	IRR	95% CI
Paediatric (<18 years) index case	0.97	0.63 to 1.49
Sharing a bathroom or bedroom	2.36	1.49 to 3.74
Masking all/most of the time	0.73	0.48 to 1.11
Bottom 20% of NDI	1.05	0.69 to 1.63

P values <0.05.considered significant and shown in bold.

IRR, incidence rate ratio; NDI, neighbourhood deprivation index.

Vaccination status of household contacts is another important factor. Findings have generally shown a reduction in SIR with vaccination of contacts,3941,43 such as in a large Danish database study of household contacts where a protective effect of vaccination in the context of Delta variant household transmission was found.⁴⁴ However, another large longitudinal study did not detect a difference in likelihood of household transmission based on vaccination status of household contacts.⁴⁵ It is likely that the number of vaccinated individuals was overall too low to show a benefit in our study given the time period (late 2020 to late 2021) and the fact that only adults were able to get vaccinated for most of the study period.

Our study occurred during the alpha and delta waves, and as we do not have molecular epidemiologic data, we classified the household infections by time period (pre-Delta or Delta era). Our findings were similar to others during the pre-Omicron era, where either the time period or known variant was not associated with differences in household.^{39 40} In fact, one of these studies taking place over a similar time period included 74 families with children and investigated both acquisition of infection in the household as well as risk factors for transmission within the household.⁴⁰ While this study did not find a difference in variants (Delta and pre-Delta), unlike our study, it did not find any difference in mask use or social distancing within the household. That study’s different approach to measuring infection, surveying participants about their practices and analysing the data may have impacted the lack of consistency with our study.

One benefit of our study was the generalisability of our research in this geographic area and similar urban areas, as shown by the facts that the study was racially and demographically representative of the surrounding (table 1) metropolitan Atlanta and that the mean SIR in this study was comparable to another study in Atlanta (SIR=0.45).³⁴ Another benefit was the incorporation of socioeconomic factors such as living in an area with a low NDI, race/ethnicity and household income. We hypothesised that low NDI, low-income or non-White race would be associated with an increased household SIR, based on the known racial and economic disparities that have been evident throughout the pandemic.46,48 While none achieved statistical significance at a p value of <0.05, the analyses did show point estimates consistent with this hypothesised higher risk of infection/high SIR in those living in the lowest NDI neighbourhoods (tables2 3). While many of the initial house studies did not incorporate associations of household transmission with racial, ethnic and socioeconomic factors, this has been more commonly studied in recent studies, many of which have noted differences consistent with our hypotheses.1639 49,51 Cerami et al found an SIR of 70% in minority households versus 52% in white households (p=0.05)¹⁶ while Donnelly et al demonstrated a higher SIR in Asian and Latino households compared with white ones.³⁹ We did not find other studies using NDI as a marker of vulnerability or deprivation.

Another benefit of our study was the use of a non-invasive salivary multiplexed antibody test that was very appealing to families, many of whom said they would not have participated if we had to take a blood sample from kids. Furthermore, the multiple epitopes of the assay allowed for the differentiation of vaccine versus infection-derived immunity, important as the vaccine was rolled out mid-study.²⁸ This allowed for accurate identification of asymptomatic cases by using anti-nucleocapsid reactivity, which has shown high correlation with serology and a sensitivity of 98% and 99%, respectively, with the N antigens used in this study.^{27 28} In fact, 40% of our index cases were asymptomatic, consistent with other studies that found high proportions of asymptomatic cases.^{36 52 53} Asymptomatic children have also been found to be associated with transmission in prior research, showing the importance of detecting asymptomatic cases in both research and public health surveillance.^{54 55} While we did not find other studies using a similar assay in household transmission studies, it has been used in surveillance of antibody reactivity in other settings, such as a nursing home.⁵⁶

Limitations of this study included the possible loss of significance when adjusting for multiple confounders, especially given that our sample was limited to 49 households. It may have been difficult to tease out other factors associated with increased household transmission in this scenario. Furthermore, it is possible that some of the asymptomatic household cases that were identified by serology could have been misclassified as a case if they had had SARS-CoV-2 infection at some point in the past, prior to the household cluster or between the index case infection and enrolment. However, the likelihood of this having a significant impact is low since three-quarters of the cases had complementary symptoms or PCR done at the time of household cluster. Another limitation is that this study occurred prior to the most transmissible variant, Omicron, emerging. However, again, this may have also been a strength since we were better able to compare factors related to transmission, factors that could have been masked with a more transmissible variant. Lastly, indeterminate results of the salivary IgG test, which were more common in young children, may have underestimated the number of household cases. However, as these cases were spread out among the households, it is unlikely that it affected the analyses. Further investigation on the performance of this test in infants and toddlers and ways to improve sensitivity is necessary to take full advantage of this non-invasive sampling method.

Conclusion

In conclusion, we found high household infection rates in households with children during the pre-Omicron era, SIRs in paediatrics versus adult infection that were not statistically different and several factors associated with higher household infections like sharing rooms with infected individuals and inconsistent or no mask use. Since much of SARS-CoV-2 transmission occurs among those in close proximity to each other, such as in household units, isolation and mask wearing are key tools to limiting transmission.

Data availability statement

Data are available upon reasonable request.