There is increasing interest in superspreading for COVID-19, not least due to an excellent article by @kakape on the topic 1/9 sciencemag.org/news/2020/05/w…
Mathematically, we are interested in the variation of the number of secondary cases that are generated by one infected index case. This variation can be described by a negative binomial distribution using the dispersion parameter k. 2/9
Small values of k mean that only a few infected individuals contribute to most transmissions. Ebola, SARS and MERS have k values around 0.2. Assuming R0 = 2, this would result in 62% of cases not transmitting at all, while around 15% of cases cause 80% of transmissions. 3/9
High values of k mean steadier transmission without much superspreading. For example, k for influenza is thought to be around 1. Assuming R0 = 2, this would result in 33% of cases not transmitting at all, while around 40% of cases cause 80% of transmissions. 4/9
So what is k for COVID-19? In January, @JulRiou and I estimated a median k = 0.54. There was still considerably uncertainty and it remained unclear whether the transmission characteristics were more akin to SARS or influenza. 5/9 eurosurveillance.org/content/10.280…
A more recent study from @JustinLessler and colleagues using data from Shenzhen, China, estimated a similar k of 0.58. 6/9 thelancet.com/journals/lanin…
Another study from @sbfunk, @AdamJKucharski and colleagues obtain a k around 0.1. wellcomeopenresearch.org/articles/5-67 A quick analysis of a published transmission cluster from Austria results in k = 0.16. 7/9 ages.at/service/servic…
The value of k for COVID-19 remains unclear, and it will likely depend on social contact structures and environmental settings. I would argue that COVID-19 does transmit in a relatively steady and efficient way while also having the potential for superspreading. 8/9
Preventing these superspreading events from happening - and conducting thorough contact tracing should they happen - can play an important role in avoiding new flare ups of COVID-19. 9/9
