Tian et al. provide a framework for assessing population- level interventions of disease outbreaks through the construction of counterfactuals in a large-scale, natural experiment assessing the efficacy of mild, but early interventions compared to delayed interventions. The technique is applied to the recent SARS-CoV-2 outbreak with the population of Shenzhen, China acting as the mild-but-early treatment group and a combination of several US counties resembling Shenzhen but enacting a delayed intervention acting as the control. To help further the development of this framework and identify an avenue for further enhancement, we focus on the use and potential limitations of compartmental mod- els. In particular, compartmental models make assumptions about the communicability of a disease that may not per- form well when they are used for large areas with multiple communities where movement is restricted. To illustrate this phenomena, we provide a simulation of a directed percolation (outbreak) process on a simple stochastic block model with two blocks. The simulations show that when transmissibility between two communities is severely restricted an outbreak in two communities resembles a primary and secondary outbreak potentially causing policy and decision makers to mistake effective intervention strategies with non- compliance or inefficacy of an intervention.
Statistics and Its Interface
Full text attached
Kane, M. J., & Gilani, O. (2021). The need to incorporate communities in compartmental models. Statistics and Its Interface, 14(1), 29-32.