Curbing the use of electronic cigarettes (e-cigarettes) and vaping among youth is a priority of public health agencies and health services to prevent this population from health complications and reduce the loss of productivity and healthcare costs in the future. For public health analysts concerned about the use of tobacco and e-cigarettes, it is insightful to look at how social media might play a role in affecting beliefs and behaviors. Further, to understand the complete picture of the social media landscape on tobacco and e-cigarettes, it is also important to consider how youth interact with and respond to conversations about e-cigarettes across platforms. Armed with this information, appropriate, actionable, and effective social media-based intervention campaigns and policies can be implemented to combat the use of these harmful products.
Rather than utilizing the traditional approaches of simulation modeling and survey-based methods for public health research, which can be slower to collect and analyze relevant data, social media itself provides a unique opportunity to gather more timely, real-time information about youth’s daily activities. Although social media mining methods are routinely leveraged to understand customer behavior about brands, their use in public health research is underexplored. Faculty across multiple colleges at George Mason imagined that utilizing this non-traditional data could inform policy evaluation and implementation programs faster than what had been done in the past.
Thanks to a grant from 4-VA@Mason, a team with broad expertise in social media mining, public health policy, and epidemiology came together to examine the best way to move forward. Lead PI Hemant Purohit, Associate Professor in the College of Engineering and Computing, Department of Information Sciences & Technology at Mason, who studies Social Media Mining, connected with fellow Mason faculty member, Hong Xue, Associate Professor in the College of Health and Human Services, Department of Health Administration and Policy. Purohit and Xue then reached out to involve Bernard Fuemmeler of Virginia Commonwealth University’s Massey Cancer Center in an advisory role. Fuemmeler is a Professor in the Department of Family Medicine & Population Health, and Scientific Director for Health Communication and Digital Innovations. The team aimed to develop a social data-driven approach toward informing and evaluating intervention designs and policymaking to prevent e-cigarette/vaping use.
With the help of a former undergraduate student, Anuridhi Gupta (who has subsequently transitioned to the doctoral program at Mason) they began by reviewing theories used for guiding behavioral health interventions in public health literature and identified the relevant theory for this research, specifically Social Cognitive Theory. Based on this framework, they designed a study to look at a range of user intent classes in the conversations on e-cigarettes and vaping by examining behaviors across multiple social platforms, in particular, Twitter, Reddit, and YouTube. The team then created a novel resource of labeled data from Twitter and Reddit conversations to support research and development of social media mining tools to aid intervention designs for curbing e-cigarette/vaping usage.
From there, the team performed an extensive experimental analysis to examine the capability of machine learning-based automated classification systems to categorize social media posts into relevant intent classes, which could inform the development of a scalable analytical tool to assist public health analysts. Further, they identified insights on promoting or discouraging e-cigarettes and vaping on Twitter and Reddit by analyzing social media posts categorized with relevant intent classes using topic modeling and psychometric techniques. The analysis indicated that Accusational posts were the most prevalent on Twitter, indicating that the public often undermines the credibility of information sources, agencies, and officials by blaming them. Similarly, Anecdotal posts were the most prevalent on Reddit.
“Without this multidisciplinary research, it would not have been possible to achieve these insights and the resulting data-driven framework for scalable social media analytics in this short timeframe. We are grateful to the 4-VA grant program for supporting us to kickstart our mission for healthy society and well-being,” Purohit said.
The research team has submitted a manuscript to a journal with a focus on a novel intent mining framework to understand online conversations on e-cigarettes to inform social media-based intervention design. Further, the team is currently working on a plan to write a National Institutes of Health proposal based on this preliminary work as a promising direction for data-driven policymaking for public health.