Characterising the distribution of incubation time and latency time

Funder: Wellcome Trust (OUCRU core funding), University of Leiden (NL)

Principal Investigators: Ronald Geskus, Nhat Le, Marc Choisy (for the OUCRU COVID-19 Modeling Group)

Collaborator(s): Pham Quang Thai, epidemiology department, National Institute of Hygiene and Epidemiology (NIHE); Vera Arntzen and Marta Fiocco (University of Leiden).

Location of activity: Ho Chi Minh City and Hanoi, Viet Nam


The time from infection to becoming infectious (latency time) and to developing symptoms (incubation time) are key characteristics of any infection. For SARS-CoV-2, data to estimate both quantities have not been collected in a systematic way. As a consequence, information is incomplete, and there is a high risk of bias. For most individuals, the exact time of infection is not known. At most, we have information on the exposure interval during which someone became infected. Also, a large percentage of individuals remains asymptomatic, and these individuals are more likely to be missed in data collection.


Study design

From the start of the pandemic until July 2021, Viet Nam has performed active contact tracing of all community infected individuals and quarantined these “F1 contacts” in supervised locations. Since March 2020, all individuals that entered Viet Nam have been quarantined. Vietnam has had four major outbreaks, caused by at least three different variants, the original variant, the alpha variant and the Delta variant. In collaboration with NIHE and the Ho Chi Minh City Centre for Disease Control, we combine data from different sources and outbreaks to create a unique data set with representative information on the initial stages of the SARS-CoV-2 infection and disease process.

Contact tracing data provides information on the window of exposure. For the latency time, we assume that individuals become infectious when SARS-CoV-2 RNA becomes detectable. This information is obtained from the longitudinal polymerase chain reaction (PCR) test results.

For the incubation time, we use the time when patients first report symptoms that are suggestive of SARS-CoV-2 infection; individuals that remain asymptomatic are excluded. Data on the latency time are doubly interval-censored. Not only the time origin (infection) but also the event time (detectable RNA) is at best known to lie within an interval: the presence of detectable RNA is known each time a PCR test is done. For the incubation time, we have exact information on the time of becoming symptomatic for most individuals; then, only the time origin is interval-censored. We estimate both distributions by maximising the likelihood for (doubly) interval-censored data. We compare estimates for the different variants.



The length of the quarantine period is based on estimates of the incubation time distribution because data on the latency time is lacking. Our project will fill this gap by estimating the latency time distribution. Furthermore, latency time and incubation time are important components for mathematical models that quantify and predict the spread of SARS-CoV-2.


Outputs to date:

Pham Quang Thai, Rabaa M, Duong Huy Luong et al. The First 100 Days of Severe Acute Respiratory Syndrome Coronavirus 2 (SARS-CoV-2) Control in Viet Nam. Clinical Infectious Diseases [Internet] 2020;72(9):e334-e342. Available from: