Modeling and Twitter-based Surveillance of Smoking Contagion

Nicotine, in the form of cigarette smoking, chewing tobacco, and most recently as vapor smoking, is one of the most heavily used addictive drugs in the world. Since smoking imposes a significant health-care and economic burden on the population, there have been sustained and significant efforts for the past several decades to control it. However, smoking epidemic is a complex and "policy-resistant" problem that has proven difficult to control. Despite the known importance of social networks in the smoking epidemic, there has been no network-centric intervention available for controlling the smoking epidemic yet.

The long-term goal of this work is the development and implementation of an environment needed for developing network-centric interventions for controlling the smoking contagion. In order to develop such an environment we essentially need: an operationalized model of smoking that can be simulated, to determine the role of online social networks on smoking behavior, and actual methods to perform network-centric interventions. The objective of this thesis is to take first steps in all these categories. We perform Twitter-based surveillance of smoking-related tweets, and use mathematical modeling and simulation techniques to achieve our objective.

Specifically, we use Twitter data to infer sentiments on smoking and electronic cigarettes, to estimate the proportion of user population that gets exposed to smoking-related messaging that is underage, and to identify statistically anomalous clusters of counties where people discuss about electronic cigarette a lot more than expected. In other work, we employ mathematical modeling and simulation approach to study how different factors such as addictiveness and peer-influence together contribute to smoking behavior diffusion, and also develop two methods to stymie social contagion. This lead to a total of four smoking contagion-related studies. These studies are just a first step towards the development of a network-centric intervention environment for controlling smoking contagion, and also to show that such an environment is realizable.