Study population and setting
The study was conducted on 391 COVID-19 cases confirmed by RT-PCR, identified by the Shenzhen Centre for Disease Control and Prevention in China, and 1286 of their close contacts. For primary cases, surveillance was conducted on returned travelers from Hubei regardless of symptoms, as well as through fever screening in the community and at clinics, and patients at hospitals.
Summary of Main Findings
The risk of severe symptoms was greater for males (odds ratio [OR] 2.5 [95% CI 1.1–6.1]) and increased slightly with age. For example, 60-69-year-olds had an increased risk compared to 50-59-year-olds (OR 3.4 [95% CI 1.4–9.5]).The median incubation period (time from infection to symptom onset) was estimated to be 4.8 days, with 90% of individuals developing symptoms between 1.6 and 14 days. The median time to recovery was estimated to be 20.8 days. Those with severe symptoms tended to have a longer time to recovery (time to first negative PCR result) – a median of 28.3 days (95% CI 25.3-31.6) compared to individuals with mild symptoms with a median of 20.1 days (95% CI 19.0-21.3). For each case, an average of 11.2% of household contacts and 6.6% of all contacts became infected (95% CIs 9.1–13.8 and 5.4–8.1 respectively). The risk of infection was not associated with age. Contacts showed symptoms a mean of 6.3 days after the initial case showed symptoms (serial interval), and each infection caused 0.4 detected infections on average, but there was large variability between individuals (the standard deviation for the serial interval was 4.2 days, and 80% of cases were caused by 8.9% of individuals). The authors built a model using these results, and found that surveillance alone would have to detect individuals causing 61% of onward transmission to stop the epidemic. This result suggests that additional measures such as physical distancing in the community are required to reduce the reproductive number to below 1.
Because PCR tests were conducted on contacts regardless of symptoms, the study captured cases across a range of disease severities and age groups, as well as individuals yet to show symptoms. The study was thus less susceptible to biases in healthcare seeking behaviors. Detailed travel and contact data enabled the authors to estimate the average number of infections caused by each individual, risks by age group and contact type, and the time between infections.
Contact tracing focused on close contacts only, and so the average number of secondary infections was likely to be higher than that estimated in the study. Some close contacts wearing masks (e.g. nurses) were excluded, so the results are of limited use for inferring transmission in a healthcare setting. The number of secondary cases would likely be different in different settings, depending on time to isolation, community physical distancing measures, and numbers of close contacts. Individuals were unlikely to be re-tested before the end of their 14-day isolation period, which could have overestimated the time to recovery.
This was one of the first studies to use detailed contact tracing data to estimate key quantities such as the serial interval (time between primary and secondary infections), incubation period (time between infection and symptom onset), and secondary attack rate (proportion of individuals infected) among close contacts. These estimates can be used in modelling studies to assess the impact of interventions such as physical distancing and digital contact tracing. Since PCR testing of close contacts of infected individuals was used to identify cases regardless of symptoms, there was less bias towards cases with severe disease, compared to studies relying on symptoms to identify infected individuals.