Skip to main content

Implication of backward contact tracing in the presence of overdispersed transmission in COVID-19 outbreak

Our take —

In this study, available as a preprint and thus not yet peer reviewed, authors modeled the propagation of COVID-19 cases and estimated the effectiveness of different contact tracing strategies in identifying secondary cases in the context of overdispersion. Model results suggest that adding backward contact tracing is 2-3 times more efficient at identifying cases than forward tracing alone. Authors demonstrated that backward tracing is a valuable strategy for developing a better epidemiologic understanding of SARS-CoV-2 transmission because it is more likely to capture cases generated from a common source than forward tracing. However, further work is needed to understand the conditions under which additional backward tracing is cost-effective and feasible.

Study design


Study population and setting

Overdispersion refers to the phenomenon where there is a lot of variability in the number of secondary cases generated from a primary case (i.e., one primary case may infect one other person while another may infect six). “Backward contact tracing” is a disease control strategy where the contact history of multiple cases up to 14 days prior to symptom onset is collected; the goal is to identify one upstream primary case. This contrasts with “forward contact tracing,” which aims to identify all contacts of a confirmed case and whether these contacts become secondary cases. Authors developed a branching process model, which is used to model reproduction of disease in given populations, to estimate the effectiveness of forward contact tracing alone compared to a strategy that uses a combination of both forward and backward tracing in the presence of overdispersed SARS-CoV-2 transmission. Authors assumed the index case that triggers either forward or backward tracing was identified through symptom-based surveillance. Effectiveness was measured by the number of third generation cases averted and the overall relative reduction in third generation cases.

Summary of Main Findings

Results from model simulations suggest that backward tracing is highly effective in identifying clusters of secondary transmission. Forward tracing, on average, is only able to identify the mean number of secondary cases. Adding backward tracing to a forward tracing strategy was more effective than forward tracing alone, increasing the number of identified secondary infections by a factor of 2-3. The higher the degree of overdispersion, the greater the absolute number of cases averted through backwards tracing: the highest degree of overdispersion assumed in the model resulted in 2-3 times more third generation cases detected than the lowest degree assumed.

Study Strengths

The model assumed any primary case is initially unknown in order to simulate how such contact tracing would operationalize in the real world.


The model assumed that second generation cases were traced and quarantined before becoming infectious, when in reality many cases may not quarantine until after becoming infectious. Authors also assumed independently identified cases would not have the same primary case. If these cases did actually come from the same primary case, tracing and contact efforts would be unnecessarily duplicated. Authors did not explicitly examine the impact of backward and forward case identification timing, only the probability that they would be. It is unclear how this parameter may impact results. Despite relative benefits of backwards tracing, limited resources may preclude its use in some settings.

Value added

This modeling study demonstrated the benefit of including backwards contact tracing efforts to other contact tracing strategies. As secondary cases are more likely to come from a cluster rather than generate a new cluster, backwards contact tracing has the potential to be extremely valuable in efforts to mitigate SARS-CoV-2 transmission when resources are available.

This review was posted on: 11 January 2021