Epidemics and cumulants

This recent paper ( Beyond R0: the importance of contact tracing when predicting epidemics ) provides an interesting new mathematical framework for thinking about the spread of epidemics, based on cumulants.

It shows that the epidemiologist’s focus on R0 is shortsighted, and that higher moments of the generating function of the social graph are important in understanding the spread of epidemics on a network (duh!). However it itself suffers from limitations in how it seeks to apply this insight.

Instead, there should be much more focus on determining the characteristics of this network community-per-community (and the paper hints at good interventions depending on context), while also modeling inter-community transmissions. The mathematical framework of cumulants should be better in handling the propagation of modeling errors.

One can see epidemiologists are starting to go in this direction, for instance in this Harvard paper, because they know household composition is super important for COVID (see their household composition samping, which has a different impact on Italy, China, etc).

It is noteworthy that this would jibe very well with the sociological response/epidemic contours pushed by sociologists and medical anthropologists elsewhere in this forum, as well as the multiscale approach that was used during Ebola.

Interestingly, if this approach was used, it would allow extracting vastly more useful data from the Korean public case database by sociologists, as they could better understand relevant communities there, or even we could try to isolate different community’s characteristics of propagation (for instance through churches).

This could all inform on how contact tracing apps might need to be deployed (or not).

See also https://royalsocietypublishing.org/doi/10.1098/rsif.2007.1100

And https://arxiv.org/pdf/2004.05272.pdf