Study population and setting
The study objective was to estimate the risk of secondary transmission within households and the associated risk factors for secondary infection. From January to February 2020 in Zhuhai, China, the study identified 104 probable/confirmed COVID-19 cases in 46 households, and enrolled 35 people who met inclusion criteria: 1) a confirmed case, 2) the only index case in their household, and 3) lived with others in their household. All 35 index cases and their 148 contacts completed study questionnaires, and the 148 household contacts were then prospectively followed for 21 days for symptom monitoring, and naso-/oropharyngeal swabs were collected every 3 to 7 days during the study period to test for SARS-CoV-2 infection.
Summary of Main Findings
Among the 35 households, 22 (62.9%) had a secondary infection, with 48 secondary cases (of which 10.4%, n=5 were asymptomatic) among the 148 exposed contacts, with an estimated secondary infection rate of 32.4% (95% CI: 22.4 – 44.4%). Using a multivariate regression model, they found having an underlying medical condition increased the odds of susceptibility of infection by 5.99 times (95% CI: 1.81 – 19.83), direct exposure to Wuhan increased the odds of infection by 4.14 times (95% CI: 1.24 – 13.68), having no protective measures after the index case developed symptoms increased the odds by 4.95 (95% CI: 1.59 – 15.39), and sharing a vehicle with them increased it by 4.37 (95% CI: 1.80 – 10.58). Notably, whether the index case wore a mask, whether the index case self-isolated while indoors, and the number of hours spent at home with the index case were not associated with the odds of secondary infection. Using a log-normal model for household secondary cases, the estimated median incubation period was 4.3 days (95% CI: 3.4 – 5.3 days), and the serial interval between successive cases was estimated at 5.1 days (95% CI: 4.3 – 6.2 days).
The study had prospective follow-up of households which allowed for estimation of the secondary infection rate and serial interval to better understand transmission, and to construct an epidemiological curve of the date of symptom onset, diagnosis, and hospital admission within the sample. Authors also collected a number of demographic and clinical variables among not only the household contacts, but the index cases (e.g. symptoms), and the household (e.g. ventilation/disinfection, number of contacts in the house, etc.). Using naso-/oropharyngeal swabs allowed them to confirm case status among the contacts, as opposed to relying on clinical diagnosis based on symptoms, which allowed for asymptomatic cases to also be included.
To construct their major multivariate model, the study used different models for the index case, contact case, and household, and then based on these results, took the most significant variables to construct their final model. This is not a valid approach to integrate these multi-level variables, and likely leads to confounding, especially given that the final model was based on significance level and not a priori defined confounders reasonably thought to play a role. In addition, they had a relatively small sample size, which is reflected in wide confidence intervals, and may lead to an underpowered analysis. They also assumed that secondary infections were exposed via the index case only, which is an oversimplification of potential transmission routes and may overestimate secondary attack rates. There may also be selection bias in the households that were selected, and households with more cases may be more likely to seek medical care and evaluation, which could again overestimate the secondary infection rate.
The study has prospective follow-up of households in order to determine the secondary infection rate and serial interval and determined potential risk factors for secondary infection.