Coupled behaviour-disease modelling for insights into pandemic waves, control measures, and emergence of social norms

Abstract

Events such as the COVID-19 pandemic, as well as many historical examples from older infectious diseases, have presented strong evidence of behavioural responses shaping the dynamics of infectious disease transmission. While some epidemiological models account for this feedback interaction between the disease and social dynamics, others do not place such a high level of importance on this phenomenon. Mathematical models that couple disease dynamics and behavioural dynamics have recorded key successes, making valuable additions to the field and science in general. However, many of these useful coupled behaviour-disease models are currently limited in their direct application to real-world cases due to a lack of validation against empirical data, especially population behaviour data. This challenge has been significantly addressed by the recent increase in behavioural data collection. This thesis takes advantage of this development to model the SARS-CoV-2 virus dynamics by constructing five coupled behaviour-disease models of varying complexity and comparing them to a conventional seasonal SIR disease model. While we fit these models to empirical data, we also formulate a similar coupled model in the context of altruistic punishment with the purpose of understanding how contact precautionary measures, such as social distancing, become social norms. Our results demonstrate that behavioural changes influenced by the decision to use or not use non-pharmaceutical interventions (NPIs), when incorporated with the simplest assumptions, perform better than the disease model in the study case of some European nations. Thus, there is a revelation of some level of collective social norm regarding NPIs. We established that such norm formation is driven by a set of necessary conditions: a low cost of mitigation and a significant cost to impact ratio to deter non-mitigating strategies. Coupled models of higher levels of complexity can be considered in this context in comparison to one another for the purposes of fitting and prediction. Our findings show that while these models can provide very useful information, they are more useful on a case specific basis. Given the right incentives, our results also show the possibility of local disease elimination when the strength of social norms for contact precautionary measures is high by driving transmission down, motivated by individuals' actions. These findings can be advantageous to individuals, social groups, institutions, and governments for public health event planning and optimising their decision-making techniques for the best outcome.