|
Abstract:
|
As soon as the outbreak of SARS-CoV-2 became an issue of worldwide concern, attempts have been made to estimate the incubation time distribution, i.e. the time from infection to symptom onset. These estimates have been used as surrogate of the latency time distribution to decide upon the length of the quarantine period after infection. Data on latency time, the time from infection to start of infectiousness, are harder to obtain and are usually based on longitudinal PCR test results. Estimates of both distributions are prone to bias due to lack of structured data collection. Furthermore, both the time origin and the end point are typically interval censored. Often parametric distributions are assumed. With the upper tail of the distribution being of interest, this may give seriously biased estimates and too narrow confidence intervals. We discuss data structure and data quality issues. As example we use data from individuals diagnosed with SARS-CoV-2 in Vietnam, both community transmissions and imported cases. We compare estimates based on parametric and nonparametric approaches.
|
