key: cord-0735965-1k46f4xz
authors: Reukers, Daphne F M; van Boven, Michiel; Meijer, Adam; Rots, Nynke; Reusken, Chantal; Roof, Inge; van Gageldonk-Lafeber, Arianne B; van der Hoek, Wim; van den Hof, Susan
title: High infection secondary attack rates of SARS-CoV-2 in Dutch households revealed by dense sampling
date: 2021-04-02
journal: Clin Infect Dis
DOI: 10.1093/cid/ciab237
sha: 6454e6325b2910ac5dd170d5986b542fff939b26
doc_id: 735965
cord_uid: 1k46f4xz

BACKGROUND: Indoor environments are considered one of the main settings for transmission of SARS-CoV-2. Households in particular represent a close-contact environment with high probability of transmission between persons of different ages and with different roles in society. METHODS: Complete households with a laboratory-confirmed SARS-CoV-2 positive case in the Netherlands (March-May 2020) were included. At least three home visits were performed during 4-6 week of follow-up, collecting naso- and oropharyngeal swabs, oral fluid, feces and blood samples for molecular and serological analyses of all household members. Symptoms were recorded from two weeks before the first visit through to the final visit. Infection secondary attack rates (SAR) were estimated with logistic regression. A transmission model was used to assess transmission routes in the household. RESULTS: A total of 55 households with 187 household contacts were included. In 17 households no transmission took place, and in 11 households all persons were infected. Estimated infection SARs were high, ranging from 35% (95%CI: 24%-46%) in children to 51% (95%CI: 39%-63%) in adults. Estimated transmission rates in the household were high, with reduced susceptibility of children compared to adolescents and adults (0.67; 95%CI: 0.40-1.1). CONCLUSION: Estimated infection SARs were higher than reported in earlier household studies, presumably owing to our dense sampling protocol. Children were shown to be less susceptible than adults, but the estimated infection SAR in children was still high. Our results reinforce the role of households as one of the main multipliers of SARS-CoV-2 infection in the population.

The first case of the coronavirus disease (COVID-19) emerged in Wuhan in December 2019 [1] .

Starting with an outbreak of pneumonia of unknown etiology, the causative agent severe acute respiratory syndrome coronavirus 2 (SARS-CoV-2) was identified early January 2020 [2] . Since then, the virus has spread rapidly across the world [3] , globally disrupting day-to-day life, causing substantial excess morbidity, mortality and economic recession [4, 5] .

Evidence from case and cluster reports shows SARS-CoV-2 is largely spread through respiratory droplets from infected persons, with proper distance and indoor air ventilation being significant factors reducing the risk of transmission [6] . Therefore, social distancing measures are important to reduce transmission, and most countries have instated strategies based on this premise. In the Netherlands, the first COVID-19 case was detected on February 27 [7] . In March, the Dutch government mandated a partial lockdown, characterized by social distancing, self-quarantine and self-isolation orders, closing of schools, bars, and restaurants, and urging people to work from home [8] . These measures generally increased the time spent at home. As household members live in close contact, it is difficult to attain a proper physical distance after a COVID-19 diagnosis of a household contact. In combination with evidence that a sizeable fraction of transmission events occur pre-symptomatically, the household constitutes a high risk setting for SARS-CoV-2 transmission [9] .

The infection secondary attack rate (SAR) of SARS-CoV-2 infection among household contacts is a useful measure to gauge the risk of transmission in this close-contact setting. It provides insight into the susceptibility of contacts and infectiousness of cases given certain characteristics, such as age, gender, household size, and severity of infection. Household studies performed in the first six months of the pandemic, mostly in China, found a relatively high household infection SAR of 15-22% [10] . In most countries, paediatric patients are underrepresented in the statistics of the COVID-19 outbreak and children usually exhibit mild symptoms [11, 12] . If children have lower susceptibility or infectiousness, this can have important implications for strategies to curb the spread of SARS-CoV-2.

If children are not a significant driver in the transmission of SARS-CoV-2, the effect of measures on the spread of SARS-CoV-2 might not outweigh the harmful side-effects, such as the impact of school closures on the mental and social health of children. Previous household studies observed that the SAR was significantly higher for adult contacts compared to child contacts [10] . However, most studies only tested household contacts with COVID-19 related symptoms, only relied on RT-PCR in nasopharyngeal swabs, and did not perform any follow-up sampling. These studies may have missed mild, pre-or asymptomatic cases, especially in children [13, 14] . In the present study all household 4 contacts were tested as soon as possible after a laboratory-confirmed infection in the household was established, and subsequently followed-up for 4 to 6 weeks. A dense sampling strategy was employed, which included sampling from various anatomical sites while using multiple molecular and serological diagnostic methods to establish infection. This sampling strategy increases the chance of detecting all SARS-CoV-2 infected household contacts. It also increases the chance of determining transmission routes, including asymptomatic transmission, as accurately as possible [10] . The main aims of this study were to estimate secondary attack rates and to determine factors that impact susceptibility and infectiousness, stratified by age of household contacts.

This study is an update of a generic stand-by protocol drafted in 2006 to quickly initiate scientific research in the case of an outbreak of an emerging pathogen [15] . The generic protocol was tailored to the current COVID-19 pandemic with input from the WHO First Few Hundred protocol [16] . The generic and adapted study protocols were approved by the Medical-Ethical Review Committee of the University Medical Center Utrecht (NL13529.041.06). A prospective cohort study was performed following households where one household member was tested positive for SARS-CoV-2 in the period March 24-April 6 2020 (one household was included later on May 24).

Any person 18 years and older testing positive for SARS-CoV-2 who had at least one child in their household below the age of 18 and consented to be contacted for scientific research were reported by the Public Health Service of the region of Utrecht. We contacted this person (i.e. the index case) to request enrolment of the entire household in this study. Every household contact (persons living in the same house as the index patient) was to be enrolled in the study, except for contacts below the age of one year. Households were excluded if one or more of the household contacts did not want to participate in the study upfront, as in that case it would not be possible to fully determine household transmission patterns. 5 

Two research nurses performed the first home visit within 24 hours after inclusion to collect the informed consent forms and the first samples from all participants (see Table 1 for schedule of sample collection). Household contacts completed a questionnaire to collect demographic characteristics, medical history, travel history, anti-viral drug use, symptoms, symptom onset and hospital admission. Participants reported whether they had symptoms in the 2 weeks prior to the first visit. After the first visit, they filled in a symptoms diary for 2 weeks. A second visit was included at 2-3 weeks post-inclusion, and at the last home visit at 4-6 weeks post-inclusion, participants reported whether they had developed symptoms in the weeks between the second and third home visit. We defined three age strata: adults 18 years of age or older, adolescents 12 to 17 years of age (corresponding to secondary school age) and children 1 to 11 years of age (corresponding to day care and primary school age).

Total nucleic acid was extracted from the nasopharyngeal swab (NP), oropharyngeal swab (OP), oral fluid and feces specimens using MagNApure 96 with total nucleic acid kit small volume and elution in 50 µl. RT-qPCR was performed on 5 µl extract using TaqMan® Fast Virus 1-Step Master Mix (Thermo Fisher) on Roche LC480II thermal cycler with SARS-like beta coronavirus (Sarbeco) specific E-gene primers and probe as described previously [17] . As no other Sarbeco viruses are currently detected in humans, a positive Sarbeco E-gene RT-qPCR is validly taken as positive for SARS-CoV-2. The results of the NP and OP swabs were combined to one result: upper respiratory tract (URT) negative (NP and OP negative) or positive (NP and/or OP positive). For detection of antibodies against SARS-CoV-2 we used the Wantai total Ig ELISA as described previously [18] .

Laboratory-confirmed SARS-CoV-2 infection was defined as, either at least one positive PCR on any of the clinical samples taken during follow-up and/or detection of antibodies at any sampling timepoint. Every index case was by definition infected, as they had at least one positive PCR on an URT swab. 6 

The day of symptom onset as reported by the participant was set as the first day of illness.

Participants were considered symptomatic if at least one of the following symptoms occurred at any timepoint: respiratory symptoms (including sore throat, cough, dyspnoea or other respiratory difficulties, rhinorrhoea), fever, chills, headache, anosmia or ageusia, muscle pain, joint ache, diarrhoea, nausea, vomiting, loss of appetite or fatigue. For household contacts, symptom onset occurring more than 2 weeks prior to the first day of illness or first positive test result of the index case were considered not related to SARS-CoV-2 transmission within the household.

A differentiation was made between mild, moderate and severe COVID-19 based on self-reported symptoms or hospital admission [19] . We defined mild COVID-19 cases as laboratory confirmed cases showing any clinical symptoms. Moderate COVID-19 cases showed clinical signs of pneumonia, including dyspnea, and severe COVID-19 cases reported dyspnea and consulted a health professional (e.g. an emergency room) for their symptoms or reported having been admitted to the hospital for COVID-19.

In every household, a primary case (the most likely first case of the household) was determined based on laboratory confirmation, symptom onset and travel history. As a default, the index case was considered the primary case, and a household contact was only considered the primary case if they had a laboratory-confirmed SARS-CoV-2 infection with a symptom onset at least 2-14 days before the index case. Household contacts who had symptoms more than 2 weeks before the index case were included in the analysis as susceptible to COVID-19, unless they were already seropositive at day 1.

Household secondary attack rates (SARs) were estimated excluding the index case (i.e. the labconfirmed person that led to inclusion of the household in the study), but including the primary case.

This corresponds to common practice as reliable information on the primary case in the household often is lacking [20, 21] . To take the clustered nature of the data into account, SARs were estimated with a logistic regression using generalized estimating equations (GEEs), with household as the unit of clustering and assuming an exchangeable correlation structure. Analyses were performed using 7 three age strata as defined above, and using covariates sex, household size, and severity of infection of the index case. Model selection was based on the Quasi Information Criterion for small sample sizes (QICc). Analyses were carried out in R (version 3.6.0) using the geepack (version 1.5.1) and emmeans (version 1.3.1) packages [22] [23] [24] [25] [26] .

Next to the estimates of the SAR we analysed the data using the final size distribution of a stochastic SEIR transmission model. In this model persons are classified as susceptible (S), infected but not yet infectious (E), infected and infectious (I), or recovered and immune (R). Appeal of these analyses are that the estimated parameters have a biological interpretation (susceptibility, infectiousness), that final-size distributions are invariant with respect to the latent-period distribution, and that different assumptions on the distribution of the infectious period can be incorporated [27, 28] . With respect to the contact process, we assumed frequency-dependent transmission as this mode of transmission is preferred over density-dependent transmission by information criteria (not shown) [29] . Time is rescaled in units of the infectious period, and we assumed a realistic variation in the infectious period, corresponding to an infectious period of 6-10 days. Here, because households were included only if an infected person was present, the final-size distributions needed to be conditioned on the presence of an infected index case if the index case was not also the primary case [28] . Such conditioning was applied for 17 households (Figure 1 ). Model selection was performed using LOOIC (leave-one-out cross validation information criterion), a measure for predictive performance [29, 30] . Estimation was performed in a Bayesian setting using Hamiltonian Monte Carlo implemented in Stan (version 2.21.2) [31] . Details will be made available on our digital repository (https://github.com/mvboven/COVID-19-FFX).

The Medical Ethical Review Board of the University Medical Centre Utrecht reviewed and approved the study protocol (NL13529.041.06). All participants above the age of 12 gave written informed consent. Parents or guardians of participating children below the age of 16 gave written informed consent for participation, for children 12-16 both parents and children had to give consent. 8 

Fifty-five households were included, with 242 total participants. Each household had an index case (55 participants) and the other 187 participants were household contacts ( Table 2) . Household size varied from three to nine persons (Figure 1 ). Index cases were predominantly female (n=40, 72.7%), and health care worker (n=41, 75.9%). Seven index cases were admitted to the hospital before or during participation in the study, and none of the other cases in the household required hospitalization. In 10 of the 55 households, the index case was observed not to be the primary case; in nine households this was determined to be another adult, and in one household, an adolescent contact. In 17 households, no transmission took place, and in 11 households every member got infected with SARS-CoV-2 ( Figure 1 ). In children, fewer SARS-CoV-2 infections were found compared to adolescent and adult household contacts. In total, 51% of adults, 46% of adolescents, and 30% of children got infected. Children and adolescent household contacts were less often symptomatic than adult contacts. Adults were also more likely to have a severe infection (37%) compared to adolescents (15%) and children (5%).

Overall estimated household infection secondary attack rate (SAR) was 43% (95%CI 33%-53%). In univariable analysis only age was significant at the 5% level (p=0.036), while sex (p=0.11), household size (p=0.64), being a healthcare worker (p=0.28) and severity of infection of the index case (p=0. 30) were all not significantly associated with the outcome (p>0.10). In a multivariable analysis that included sex and age group (child, adolescent, adults), being a child was strongly associated with decreased probability of infection (p=0.006), while female sex was not significantly associated with Table 1 ). 9 

Building on the results of the SAR estimates, we analyzed transmission models that differed with respect to assumptions on the susceptibility and transmissibility of age groups (children, adolescents, adults). Table 3 shows the results. In the unstructured model the transmission rate was estimated at 1.2 per infectious period, which implies that an infected person infects on average 1. 

Estimated household infection secondary attack rates in our study were high (43%) and substantially higher than reported in earlier studies (reviewed in [10] ). Transmission model analyses corroborated this finding and in addition revealed that children below the age of 12 had reduced susceptibility compared to adolescents and adults. Neither household size, nor severity of infection of the index case had a significant impact on household infection SAR or transmission in households.

Our study confirmed that the infection SAR for SARS-CoV-2 is higher than the infection SARs of related emerging coronaviruses such as SARS-CoV (6.0%) or MERS-CoV (3.5%) [10] . High estimated SARs are also in line with observations from surveillance data and cluster reports that the household is the most frequently reported setting of infection [32] . Our study differs from earlier household studies for SAR-CoV-2 in that we observed substantially higher SARs [10, 33] . In fact, only a few studies reported estimates that were somewhat similar to our estimates (32%-38% vs 43%) [34, 35] .

A systematic review showed that estimated infection SARs increase with frequency of testing [33] , and that most of the earlier studies analyzed existing data from contact tracing procedures performed by local public health services, monitored household contacts only during quarantine without additional follow-up, or only tested symptomatic household contacts. A few studies did test 10 all household contacts irrespective of symptoms. However, none had a follow-up of more than 4 weeks, or an increased sensitivity for case finding through assessment of multiple sample types and diagnostic methods. Thus, we believe that our estimates of SARs may be more representative of the true household infection SARs than those presented in earlier studies.

A systematic review based on 18 studies concluded that children are at lower risk of infection than adults, although there was substantial heterogeneity in study design and in population characteristics [36] . Potential contributing factors include age-specific differences in the balance between innate and adaptive immune responses [37, 38] , more concomitant viral infections or cross-immunity to other coronaviruses in adolescents and adults [39] [40] [41] , and physiological differences in the respiratory tract of children as compared to adults [38, 42] . Irrespective of the cause, a relevant question is, how infectious are infected children to other household members, especially as in some studies the severity of infection of the index case was associated with higher infectiousness [10, 33] . We did not find evidence for lower transmissibility of children compared to adolescents and adults, but it should be noted that our study (and for that matter, most other household studies) may be underpowered to detect even moderate differences in age-specific transmissibility.

We would also like to mention two related limitations of our analyses. For inclusion of households our study depended on the prevailing testing policies and infected population in different age groups. This may well have resulted in a high likelihood of selecting (symptomatic) adult index cases, such that index cases may not be representative of infections in the population. Standard practice for estimating SAR partly solves this problem by taking into account all secondary infections in the household while leaving the index case out from the analyses [21] . Related to this is the fact that the index case may not always be the primary (i.e. first sequential) case in the household [20, 21] , such that standard estimates of SARs may not be indicative of transmission routes in the household. A previous household study tried to solve this issue by including index cases and excluding primary cases [43] , but this introduces bias as it would artificially increase the SAR of the prevailing type of the index cases (i.e. adults). In a sensitivity analysis, we reran the analyses using logistic regression by excluding both the primary and index case, and found relatively small impact on the estimated SARs in different age groups (not shown). The transmission model analyses do not suffer from these problems, but they do require that the primary case in the household can be identified, and this is often not possible in retrospective household analyses. Finally, we did not collect information about behavioral parameters that might have influenced transmission within households. However, we assume that households adhered to the prevailing social distancing and 11 other control measures. Wearing masks was not one of the measures advised by the Dutch government at that time. Factors we did take into account, which might make isolation and quarantine more difficult, such as household size, were not significantly associated with the outcomes.

Our results confirm and reinforce the household is a main transmission source of SARS-CoV-2. Our results also underscore the need, not only for isolation of infected household members, but also for prompt and effective quarantine of household contacts. 12 Figure 1 Legend:

Overview of transmission of SARS-CoV-2 within households. Each row represents a household and each square a household member. The grey squares are the index cases and the most likely primary case is indicated by a black border. Blue squares indicate uninfected household members and red squares infected household members, with lighter colors indicating a younger age group. The squares are ordered by age, but the first two squares are always two spouses and the parents/guardians of the children. 

ProMED International Society for Infectious Diseases. UNDIAGNOSED PNEUMONIA -CHINA (HUBEI): REQUEST FOR INFORMATION. Archive Number: 20191230.6864153. Available at

Early Transmission Dynamics in Wuhan, China, of Novel Coronavirus-Infected Pneumonia

Timeline of WHO's response to COVID-19

Effects of Physical Distancing to Control COVID-19 on Public Health, the Economy, and the Environment

Excess all-cause mortality during the COVID-19 pandemic in Europe -preliminary pooled estimates from the EuroMOMO network

Ann Intern Med 2020. 7. National Institute for Public Health and the Environment (RIVM)

Minister Arie Slob, Minister Tamara van Ark. Kamerbrief met nieuwe aanvullende maatregelen om de COVID 19 uitbraak te bestrijden

Transmission heterogeneities, kinetics, and controllability of SARS-CoV-2. medRxiv : the preprint server for health sciences

Household Transmission of SARS-CoV-2: A Systematic Review and Meta-analysis

COVID-19 in 7780 pediatric patients: A systematic review

Systematic review of COVID-19 in children shows milder cases and a better prognosis than adults

Epidemiology of COVID-19 Among Children in China

Utility of the first few100 approach during the 2009 influenza A(H1N1) pandemic in the Netherlands

The First Few X (FFX) Cases and contact investigation protocol for 2019-novel coronavirus (2019-nCoV) infection, Version 2

Detection of 2019 novel coronavirus (2019-nCoV) by real-time RT-PCR

Differences in Antibody Kinetics and Functionality Between Severe and Mild Severe Acute Respiratory Syndrome Coronavirus 2 Infections

Design and Analysis of Vaccine Studies

Design and Analysis of Vaccine Studies

R: A language and environment for statistical computing

The R package GEEPACK for generalized estimating equations

Estimating equations for association structures

Estimated Marginal Means, aka Least-Squares Means

The distribution of general final state random variables for stochastic epidemic models

Estimation of household transmission rates of pertussis and the effect of cocooning vaccination strategies on infant pertussis

Practical Bayesian model evaluation using leave-one-out crossvalidation and WAIC

A Widely Applicable Bayesian Information Criterion

A Probabilistic Programming Language

What settings have been linked to SARS-CoV-2 transmission clusters?

The household secondary attack rate of SARS-CoV-2: A rapid review

COVID-19 Testing, Epidemic Features, Hospital Outcomes, and Household Prevalence

Transmission Potential of Asymptomatic and Paucisymptomatic Severe Acute Respiratory Syndrome Coronavirus 2 Infections: A 3-Family Cluster Study in China

Susceptibility to SARS-CoV-2 Infection Among Children and Adolescents Compared With Adults: A Systematic Review and Meta-analysis

Characterization of pre-existing and induced SARS-CoV-2-specific CD8+ T cells

Why is COVID-19 less severe in children? A review of the proposed mechanisms underlying the age-related difference in severity of SARS-CoV-2 infections

Why is COVID-19 so mild in children?

COVID-19 is milder in children possibly due to cross-immunity

Possible causes for decreased susceptibility of children to coronavirus

Physiological advantages of children against COVID-19

Household Transmission of SARS-CoV-2 in the United States

Refers to the reference class, viz. adults. Unit: per infectious period (see Methods)

Transmissibility and susceptibility are relative to adults, and include intrinsic differences between groups (e.g., differences in viral loads) and varying rates at which contacts are made

Figure 1

We thank the Public Health Service Utrecht for assistance in the recruitment of households. We Weert for support during the early stages of this study, household visits, and laboratory analyses.We would like to thank Johan Reimerink in particular for his contributions to developing the serological assays, and Chris van Dorp (Los Alamos National Laboratory) for his support coding the final size distribution in Stan.None of the authors has any potential conflicts of interest to disclose.