I've seen a few tweets recently about how R_0 is the mean of a distribution (via @nntaleb) and how its dispersion is important to understand (via @DFisman & @C_Althaus).
This is very true (for #nCoV2019 & otherwise), and it's why I posted this graphic last week. [THREAD] 1/x
This is very true (for #nCoV2019 & otherwise), and it's why I posted this graphic last week. [THREAD] 1/x
@nntaleb @DFisman @C_Althaus As y'all may recall, R_0 is the *average* number of people a new case will infect in a fully susceptible population... But in a given population, the number of ways R_0 can be 2 (as in the above visualization) is essentially countless because *each person is different*. 2/x
@nntaleb @DFisman @C_Althaus We have different biology and different behaviors that impact our individual likelihood of passing on the infection in question. This is why some people may infect lots of people and some people may infect no one at all, yet yield an R_0 of 2 (see Scenario 2 in the viz). 3/x
@nntaleb @DFisman @C_Althaus In situations like Scenario 2, we end up with a lot of "dead ends" with respect to chains of transmission because most people don't infect anyone. In situations like Scenario 3 though, we have fewer dead ends and a higher chance of sustained chains of transmission. 4/x
@nntaleb @DFisman @C_Althaus Both scenarios yield an R_0 of 2 – but they mean different things for sustained transmission. This is why understanding this underlying distribution (or dispersion) for @nCoV2019 will be important. Information on contact networks and infection networks can help with this. 5/x
@nntaleb @DFisman @C_Althaus @Ncov2019 As of now, I don't think we know which scenario #nCoV2019 best represents. However, @DFisman has pointed out via @ProMED_mail that #SARS exhibited overdispersion [promedmail.org/promed-post/?i…].
My colleagues & I found similar results with #MERS too. [ncbi.nlm.nih.gov/pmc/articles/P…].
6/x
My colleagues & I found similar results with #MERS too. [ncbi.nlm.nih.gov/pmc/articles/P…].
6/x
@nntaleb @DFisman @C_Althaus @Ncov2019 @ProMED_mail If this ends up *also* holding true for the related #nCoV2019, we may expect to see shorter chains of transmission (and a smaller overall outbreak size) than a situation where the underlying distribution isn't as dispersed – despite the fact that R_0 seems moderate in value. 7/x
@nntaleb @DFisman @C_Althaus @Ncov2019 @ProMED_mail But that's still a big "if". We don't know yet, and that's why this is a space to watch. It's also why R_0 should be considered a measure of *potential* transmissibility. When R_0 > 1, sustained transmission may be possible – but the underlying distribution matters too. 8/x
@nntaleb @DFisman @C_Althaus @Ncov2019 @ProMED_mail Until we have more information on said distribution, what matters is that the vast majority of estimates of R_0 for #nCoV2019 have been >1... But because overdispersion is still a possibility, such estimates should be viewed as a call for planning & preparedness – not panic. 9/9
