Contact Tracing and Transmission Dynamics

Paired data of index case and close contacts were extracted from the contact tracing database and outbreak investigation reports. For a family cluster, the index case was determined based on the temporality of symptom onset and review of the epidemiological link. A secondary case was excluded from the paired data if the beginning of exposure was after symptom onset of the secondary case (only applied when the secondary case was symptomatic). For health care contacts, the date at exposure would be the date at admission of the case if the exact date at exposure was not recorded.
Incubation period and serial interval were estimated using the contact tracing data in Taiwan and publicly available data sets globally (eMethods in the Supplement). We used the Bayesian hierarchical model to increase the stability in small-sample estimation. The exposure window period was defined as the period between the first and last day of reported exposure to the index case based on contact investigation. Following the WHO, we defined the secondary clinical attack rate as the ratio of symptomatic confirmed cases among the close contacts.¹⁹ We analyzed the dynamic change of secondary clinical attack rate in relation to symptom onset of the index case (days <0, 0-3, 4-5, 6-7, 8-9, or >9).
The percentage of missing information was small (7.0% for age, 6.1% for sex, and 3.3% for time from onset to exposure; Table 1). In the univariable analysis of secondary clinical attack rate by different exposure characteristics (eg, age), close contacts with missing information in that particular exposure attribute were excluded. All statistical tests were 2-sided with an α level of .05. All confidence intervals (CIs) were 95%. R software (R Foundation for Statistical Computing) and RStan (Stan Development Team) were used for data management and analysis.