Serial interval of SARS-CoV-2 was not constant, but was shortened over time by non-pharmaceutical interventions

Like Comment

It was back in February that we first noticed the phenomenon of shortened serial intervals, i.e., the time between illness onsets in successive cases in a transmission chain, when we were working on one of our earlier projects 1. Under that project, we were constructing transmission pairs with then available contact tracing data on COVID-19 for mainland China. The main research questions were to address on estimation of several essential parameters, including super-spreading events, serial intervals, and hazard of infection. In fact, serial interval is crucial for understanding the transmission of an epidemic and to estimate several other important epidemiological parameters including generation time, basic reproduction number (R0), instantaneous reproduction number (Rt) etc. The Rt is a measure of real-time transmissibility and defined as the average number of infectees per infector at time . Till date, the serial interval assumed to be constant over time and Rt was estimated with this single serial interval distribution.

We noticed a diverse range of estimations for serial interval for COVID-19 has been reported by numerous studies. These estimates were based on analyzing transmission pair data of different length and using different forms of probability distribution. Even with sufficient sample size for the transmission pairs and using the same distribution, the serial intervals were found to be significantly different. Therefore, we were encouraged to answer: why were these serial interval estimates that much diverse? Apart from the sample size and distributions used, we considered that the temporal scale of data could have driven the length of serial interval. We simultaneously, took up this project 2, focusing on the serial intervals.

We started to evaluate the serial interval at temporal scale i.e. during pre-peak weeks, peak weeks, and post-peak weeks, and found that the serial intervals shortened significantly as epidemic progressed. Therefore, we coined a term daily ‘effective serial interval’ which is a real time measure of serial interval that accounts the effects of the potential factors over time.

Next we explored the possible reasons for this temporal pattern of serial intervals. In theory, serial interval depends on the infectiousness profile of the infector, the incubation period of the infectee as well as the properties of contacts, e.g., contact patterns, structure of contacts. Therefore, the underlying mechanism of observing shortened serial intervals might be the result of these non-pharmaceutical interventions (NPIs), implemented over time. These NPIs either truncate the infectiousness profile of infector, or lessen the risk of getting effective contact for infectee.  Hence, we were encouraged to test these hypotheses by different model analyses, including probabilistic models, simulation models and regression models. All these models shout along with other NPIs, timely isolation of infectors had led to shortened serial intervals of COVID-19 in China and the evidential agreement of the results from these models analyses likely generalize in other locations.

An important implication of our work is its influence on estimating the effective serial interval to correct the real-time transmissibility (i.e. estimating Rt) over using a single constant serial interval. Our findings open up two main research questions:

  1. To develop a convenient framework for estimating corrected transmissibility (Rt) by accounting for the real-time metric of effective serial intervals for infectious diseases in general.
  2. Effective serial intervals provide better measurement of instantaneous transmissibility (Rt) as it includes the effects of possible drivers of transmission. Therefore, it is interesting to evaluate the effectiveness of these NPIs by assessing how the corrected Rt reflects the effects of NPIs at temporal and spatial scale.

Answering these questions would be helpful to policy makers by offering a real-time information on the impact of public health measures, hence, these could be taken up by our team and elsewhere as the future course of research advancement.

 References

 1             Xu, X. K. et al. Reconstruction of Transmission Pairs for novel Coronavirus Disease 2019 (COVID-19) in mainland China: Estimation of Super-spreading Events, Serial Interval, and Hazard of Infection. Clin Infect Dis, doi:10.1093/cid/ciaa790 (2020).

2             Ali, S. T. et al. Serial interval of SARS-CoV-2 was shortened over time by nonpharmaceutical interventions. Science, doi:10.1126/science.abc9004 (2020).

Sheikh Taslim Ali

Research Assistant Professor, School of Public Health, The University of Hong Kong

My primary research area is the Infectious Diseases Epidemiology, particularly respiratory viruses including influenza, RSV, SARS CoV-2 viruses. The main focus of my research is to understand the disease dynamics through stochastic modeling and statistical inference of the outbreak; estimation epidemiological parameters via time-series modelling under Bayesian framework. I develop stochastic simulations (Monte Carlo) of infectious disease dynamics, with the aim of determining how the effectiveness of containment and mitigation policies depend on disease severity. Currently, I am working on developing the general mechanistic models to address the seasonality, prediction and forecasting of influenza virus transmission modulated by the different potential drivers (meteorological, environmental and social).

No comments yet.