The impact of household physical distancing and its timing on the transmission of SARS-CoV-2

ABSTRACT
Understanding the formation and dynamics of within-household contact
networks is crucial for accurate modeling of infectious disease transmission.
In household studies, contact network information is usually collected in
samples, although this feature is usually overlooked by statistical network
models. In the first part of this talk, we present a flexible framework for
jointly modeling multiple networks using Exponential Random Graph
Models (ERGMs), designed to appropriately handle and integrate
information from network samples. We then use this framework to examine
detailed contact patterns in 280 U.S. households with SARS-CoV-2 index
cases, incorporating longitudinal symptom and contact data, and employing
ERGMs to elucidate behavioral responses to infection awareness. By
calibrating a two-level mixing model to empirical infection data, we show
that behavior change—specifically, reducing physical contact at symptom
onset—can substantially decrease household transmission, lowering
secondary infections by over 35\% in households of 4-5 members.
This illustrates how principled network modeling can enhance predictive 
simulation and inform targeted interventions at the household scale.