Towards a Characterization of Behavior-Disease Models

TLDR

Recent epidemic models increasingly use detailed census and mobility data to achieve household‑level granularity, yet systematic work on coupling behavioral changes with disease spread remains scarce, limiting their usefulness when population behavior shifts in response to epidemic awareness. This study proposes a characterization of prototypical self‑initiated social‑distancing mechanisms driven by local and non‑local prevalence information. The authors extend the SIR framework by adding behavioral classes whose transitions are coupled to disease spread, yielding a rich phase space with multiple epidemic peaks and tipping points that can be employed in data‑driven computational analyses of social adaptation.

Abstract

The last decade saw the advent of increasingly realistic epidemic models that leverage on the availability of highly detailed census and human mobility data. Data-driven models aim at a granularity down to the level of households or single individuals. However, relatively little systematic work has been done to provide coupled behavior-disease models able to close the feedback loop between behavioral changes triggered in the population by an individual's perception of the disease spread and the actual disease spread itself. While models lacking this coupling can be extremely successful in mild epidemics, they obviously will be of limited use in situations where social disruption or behavioral alterations are induced in the population by knowledge of the disease. Here we propose a characterization of a set of prototypical mechanisms for self-initiated social distancing induced by local and non-local prevalence-based information available to individuals in the population. We characterize the effects of these mechanisms in the framework of a compartmental scheme that enlarges the basic SIR model by considering separate behavioral classes within the population. The transition of individuals in/out of behavioral classes is coupled with the spreading of the disease and provides a rich phase space with multiple epidemic peaks and tipping points. The class of models presented here can be used in the case of data-driven computational approaches to analyze scenarios of social adaptation and behavioral change.