Networks for smoking dynamics

Over the years, multiple policies have helped considerably decrease the prevalence of smoking worldwide. The majority of these policies have been at the population level. Despite the current progress in tobacco control resulting from these policies, the rate of decline of smoking prevalence is decreasing. Therefore, there is a need for new models that can capture the system’s full complexity and act as test-beds to develop better policies and control measures.

Recent network-based studies have shown that smoking is a behavioural contagion. In particular, smoking behaviour can spread through social ties and depending on the social tie, the probability of both smoking initiation and quitting changes [2, 3]. Additionally, relapsing into smoking is more likely when quitters engage with a smoker group [2]. Even though empirical research has established the contagious nature of smoking behaviour, current models do not incorporate this completely. Most of these models are Ordinary Differential Equation (ODE) models and do not account for network topology or consider all possible empirically proven interactions.

We identify the following contributions in this paper. Firstly, we develop an agent-based model (ABM), which considers both spontaneous terms and interactions between agents to study the spread of smoking. This model can be used to develop network-based intervention strategies and policies for tobacco control. Secondly, we explore the effect of the underlying network topology on smoking dynamics. We find that the underlying network structure affects smoking dynamics considerably. The ABM on the real-world network accurately replicates historical data, while the ABM on a Fully-Connected (FC) network performs much worse than other networks. Finally, we show that policymakers can use Lancichinetti-Fortunato-Radicchi (LFR) networks which have community structure embedded in them to model smoking behaviour when the actual network of the local population is unavailable.