High variability in transmission of SARS-CoV-2 within households and implications for control

Abstract

Background Severe acute respiratory syndrome coronavirus 2 (SARS-CoV-2) poses a high risk of transmission in close-contact indoor settings, which may include households. Prior studies have found a wide range of household secondary attack rates and may contain biases due to simplifying assumptions about transmission variability and test accuracy. Methods We compiled serological SARS-CoV-2 antibody test data and prior SARS-CoV-2 test reporting from members of 9,224 Utah households. We paired these data with a probabilistic model of household importation and transmission. We calculated a maximum likelihood estimate of the importation probability, mean and variability of household transmission probability, and sensitivity and specificity of test data. Given our household transmission estimates, we estimated the threshold of non-household transmission required for epidemic growth in the population. Results We estimated that individuals in our study households had a 0.41% (95% CI 0.32%– 0.51%) chance of acquiring SARS-CoV-2 infection outside their household. Our household secondary attack rate estimate was 36% (27%– 48%), substantially higher than the crude estimate of 16% unadjusted for imperfect serological test specificity and other factors. We found evidence for high variability in individual transmissibility, with higher probability of no transmissions or many transmissions compared to standard models. With household transmission at our estimates, the average number of non-household transmissions per case must be kept below 0.41 (0.33–0.52) to avoid continued growth of the pandemic in Utah. Conclusions Our findings suggest that crude estimates of household secondary attack rate based on serology data without accounting for false positive tests may underestimate the true average transmissibility, even when test specificity is high. Our finding of potential high variability (overdispersion) in transmissibility of infected individuals is consistent with characterizing SARS-CoV-2 transmission being largely driven by superspreading from a minority of infected individuals. Mitigation efforts targeting large households and other locations where many people congregate indoors might curb continued spread of the virus.

SARS‐CoV‐2 infection in dogs and cats is associated with contact to COVID‐19‐positive household members

“…Large variations in transmissibility and superspreading by a minority of individuals have been observed in humans (Toth et al, 2021). In our study, we did not find a difference in SARS-CoV-2 prevalence in pets for different HH sizes or composition (i.e., number of humans, cats and/or dogs).…”

“…Several studies in humans have proven that the risk of SARS‐CoV‐2 transmission within HHs can be high and that a larger HH size is associated with seropositivity (Reukers et al., 2022 ; Warszawski et al., 2022 ). Large variations in transmissibility and superspreading by a minority of individuals have been observed in humans (Toth et al., 2021 ). In our study, we did not find a difference in SARS‐CoV‐2 prevalence in pets for different HH sizes or composition (i.e., number of humans, cats and/or dogs).…”

Monogamy as protection against COVID‐19?: Non‐monogamy stigma and risk (Mis)perception

“…Thus, it is inappropriate to apply messages about monogamy (i.e., that having fewer and psychologically close sex partners is safer) to COVID‐19 (i.e., that having fewer and psychologically closer interaction partners is safer). Reducing the number of contacts, one has increases one's risk for COVID‐19, but a more important predictor of acquiring COVID‐19 is the amount of time spent in the presence of others (Toth et al., 2021 ).…”

“…Moreover, even after COVID‐19 tests became available, it was still impossible to know a person's status at any given time. COVID‐19 tests are not always accurate due to false negatives, missed detection windows, or because a partner contracted COVID‐19 after taking the test (Toth et al., 2021 ). Without knowing the infection status of one's household members, one might incorrectly presume that household members pose less risk of COVID‐19 transmission than do individuals with the same risk factors who live outside of the household.…”

Quantifying individual-level heterogeneity in infectiousness and susceptibility through household studies

“…When household studies evaluate heterogeneity at all, it is commonly by stratifying cases by a pre-specified individual trait (like host age) [62]. Reed-Frost models have been extended to "multitype" versions in which distinct classes of individuals have di↵erent infectivities [63,64] and "collective" versions where infectiousness varies continuously across individuals [63,65,66], and have been used to estimate jointly the SAR and variation in infectiousness for the 1918 influenza pandemic [67], 2009 "swine flu" (H1N1) [68], and COVID-19 [69,70]. A stochastic, discrete-generation model chain binomial has the benefit of admitting analytic solutions for the outbreak size distribution in a finite, well-mixed population like a household, and is often mathematically equivalent to continuous time models [63,71].…”