Close
About
FAQ
Home
Collections
Login
USC Login
Register
0
Selected
Invert selection
Deselect all
Deselect all
Click here to refresh results
Click here to refresh results
USC
/
Digital Library
/
University of Southern California Dissertations and Theses
/
Tobacco and marijuana surveillance using Twitter data
(USC Thesis Other)
Tobacco and marijuana surveillance using Twitter data
PDF
Download
Share
Open document
Flip pages
Contact Us
Contact Us
Copy asset link
Request this asset
Transcript (if available)
Content
i
Tobacco and marijuana surveillance using Twitter data
by
Anuja Majmundar
A Dissertation Presented to the
FACULTY OF THE USC GRADUATE SCHOOL
UNIVERSITY OF SOUTHERN CALIFORNIA
In Partial Fulfillment of the
Requirements for the Degree
DOCTOR OF PHILOSOPHY
(PREVENTATIVE MEDICINE (HEALTH BEHAVIOR RESEARCH))
December 2020
ii
Dedication
To my kind-hearted husband, Shankar.
For being there for me through it all.
For bringing out the best in me.
For doing everything in his power to create a safe space for us to dream, learn, and grow.
To all those who are resilient in the face of structural inequality.
For staying in the game.
For beating the odds.
For setting the bar high for the rest of us.
One day at a time.
iii
Acknowledgements
I am grateful to my dissertation committee chair, Dr. Mary Ann Pentz, for her unwavering
support and excitement for my ventures into the unfamiliar territories of learning; her quiet efforts,
sometimes unbeknown to me, in making sure I had the right resources to make progress. To my entire
dissertation committee: Dr. Jennifer B. Unger, for lending her voice on issues that affected me personally,
offering constant encouragement and feedback, and inspiring me to become a better researcher; Dr. Tess
Boley-Cruz, for her kindness, direction, and encouragement in promoting my work; Dr. Jon-Patrick
Allem, for being generous with his time and efforts to guide me in research scholarship and career
decisions, for reviewing the first drafts of my papers, and for leading by example when it came to
pursuing research as an engaging and exciting career, and Dr. Pablo Barberá, for his helpful feedback,
timely advice, and empathy during a time when factors beyond my control affected me. I am also grateful
to USC Tobacco Center of Regulatory Science staff members for their support. To my dear friends,
Georgia Christodoulou, for her constant support and efforts in propping me up when I needed it, and Jessi
Tobin for our shared love of self-deprecating humor and for keeping it real. To my family for the values.
I would also like to acknowledge that this dissertation took shape in the midst of the COVID-19
pandemic, the vaping epidemic, socio-political turmoil, hiring freezes, and my pregnancy. And the lamps
of learning continue to burn.
This dissertation was also supported by the National Cancer Institute, and the FDA Center for
Tobacco Products (CTP) (Grant # U54 CA180905). This dissertation does not necessarily represent the
official views of the NIH or FDA.
iv
Table of Contents
Dedication ....................................................................................................................................... ii
Acknowledgements ........................................................................................................................ iii
List of Tables .................................................................................................................................... v
List of Figures .................................................................................................................................. vi
Abstract ......................................................................................................................................... vii
Introduction .................................................................................................................................... 1
Chapter 1: Monitoring Twitter discourse about health-effects of vaping, 2018-2019.............. 12
Abstract ............................................................................................................................................................... 12
Introduction ......................................................................................................................................................... 13
Methods ............................................................................................................................................................... 15
Results ................................................................................................................................................................. 17
Discussion ........................................................................................................................................................... 20
Limitations........................................................................................................................................................... 22
Conclusion ........................................................................................................................................................... 22
Chapter 2: Associations between categories of health effects of e-cigarettes and stance
towards e-cigarettes .................................................................................................................... 24
Abstract ............................................................................................................................................................... 24
Introduction ......................................................................................................................................................... 25
Methods ............................................................................................................................................................... 27
Results ................................................................................................................................................................. 31
Discussion ........................................................................................................................................................... 34
Limitations........................................................................................................................................................... 37
Conclusion ........................................................................................................................................................... 37
Chapter 3: Twitter Surveillance at the Intersection of the Triangulum ..................................... 38
Abstract ............................................................................................................................................................... 38
Introduction ......................................................................................................................................................... 39
Methods ............................................................................................................................................................... 41
Results ................................................................................................................................................................. 46
Discussion ........................................................................................................................................................... 51
Limitations........................................................................................................................................................... 55
Conclusion ........................................................................................................................................................... 55
Concluding Remarks ..................................................................................................................... 56
References .................................................................................................................................... 58
v
List of Tables
Table 1. Predominant categories of health effects and example keywords .................................. 16
Table 2. E-Cigarette-related keywords ......................................................................................... 17
Table 3. Descriptive statistics of Post-Stance towards e-cigarettes .............................................. 32
Table 4. Adjusted multinomial logistic regression estimates (Hypothesis 1) ............................... 33
Table 5. Adjusted multinomial logistic regression estimates (Hypothesis 2) ............................... 34
Table 6. List of tobacco, marijuana, and e-cigarette search keywords ......................................... 41
Table 7. Definitions of key topics at each intersection of the Triangulum ................................... 44
Table 8. Predominant topics at each intersection of the Triangulum ........................................... 51
vi
List of Figures
Figure 1. Conceptual framework .................................................................................................... 9
Figure 2. Percentage distribution of categories in 2019 ............................................................... 18
Figure 3. Percentage distribution of categories in 2018 ............................................................... 19
Figure 4. Percentage distribution of categories of health effects during 2018 and 2019 .............. 20
Figure 5. Inclusion criteria and sampling ..................................................................................... 30
Figure 6. Topics at the intersection of combustible tobacco and marijuana ................................. 47
Figure 7. Topics at the intersection of e-cigarettes and marijuana ............................................... 48
Figure 8. Topics at the intersection of e-cigarette and combustible tobacco ................................ 49
Figure 9. Topics at the intersection of e-cigarettes, marijuana, and combustible tobacco ........... 50
vii
Abstract
Electronic-cigarette (e-cigarette) use (vaping) has increased sharply among U.S. youth and adults,
signaling the emergence of a new public health epidemic. While emerging findings suggest that vaping
may facilitate smoking cessation, it may also result in adverse health effects. Additionally, evolving
patterns of vaping behaviors and the introduction of new products compatible with nicotine and marijuana
among never and current smokers further complicate the challenge of assessing health effects of these
products, devising effective strategies for tobacco regulation, and communicating scientific findings with
the public. Social media surveillance of real-time, naturally occurring public conversations about the
health effects of vaping offer an opportunity to capture timely insights that complement and extend
findings from traditional research methodologies. This dissertation: (1) monitors Twitter discourse about
predominant categories of health effects of e-cigarettes over two years (2018 and 2019), (2) examines
associations between categories of health effects of e-cigarettes discussed on Twitter and subsequent
stance toward e-cigarettes (pro-, anti-, neutral-), (3) explores key topics of discussions pertaining to co-
occurring mentions of marijuana, tobacco and e-cigarettes based on the Triangulum framework, which
has been adopted by the state of California to investigate the interrelated influences of use of these
substances. This dissertation found that: (1) In 2019, Death was most commonly referenced followed by
Neurological and Respiratory health effects in 2019. In 2018, Neurological was the most common health
effect, followed by Mental Health and Death, (2) Majority of posts were pro-e-cigarettes. Posts pertaining
to Neurological health effects were predominantly pro-e-cigarettes followed Mental Health, and
Respiratory health effects. Pain was significantly less likely to be pro-e-cigarette compared Mental Health
effects, (3) Person tagging was the most predominant topic across all intersections of the Triangulum
(marijuana, tobacco and e-cigarettes). Discussions about Illicit products in the marketplace was one of the
key topics at in the intersection of e-cigarettes, combustible tobacco, and marijuana, Blunts and Cigars at
the intersection of combustible tobacco and marijuana, Product features and undesirable Smell among e-
cigarette and marijuana-related conversations, and Flavors among e-cigarette and combustible-related
discussions. Future health communication campaigns may consider enhancing awareness of health risks
associated with adverse outcomes related to the predominant categories of health effects: Death, Mental
Health, and Neurological. Another point of intervention to curb e-cigarette use could be when e-cigarette
users discuss health-related concerns linked to a subsequent Pro-Stance towards e-cigarettes online. By
examining the intersections of marijuana and other tobacco products, this study offers inputs for
designing comprehensive FDA regulations including regulating product features that may be appealing
and improving enforcement efforts to curb sales of illicit products. Future public health interventions for
tobacco products may warrant including leveraging discussions about undesirable aspects of tobacco use
such as the undesirable smell of smoking and adverse health outcomes.
1
Introduction
Background and Significance
Social media platforms have emerged as popular public venues for sharing and seeking health
information, especially among youth (Fox, 2011; Hausmann, Touloumtzis, White, Colbert, & Gooding,
2017; Majmundar, Allem, Boley Cruz, & Unger, 2018; Vance, Howe, & Dellavalle, 2009; Vraga & Bode,
2018). Social media users contextualize their discussions about tobacco products in light of information
from news articles (Lazard, Wilcox, Tuttle, Glowacki, & Pikowski, 2017), personal experiences (Allem,
Dharmapuri, Unger, & Cruz, 2018), marketing messages (Allem, Cruz, et al., 2018), peer discussions
(Chu et al., 2018; Chu et al., 2015). Tobacco products also include e-cigarettes, defined as battery-
operated devices in the shape of a cigarette, cigar, or pen that aerosolize nicotine and other additives
without tobacco combustion (National Cancer Institute, 2019). E-cigarette users, in particular, actively
use social media to obtain and share information about these products (Link, Cawkwell, Shelley, &
Sherman, 2015; Pepper et al., 2017), including discussing health effects experienced from product use
(Park & Conway, 2017). As such, social media platforms offer opportunities to investigate online
discourse about diverse aspects of e-cigarettes and inform public health policies and interventions about
emerging trends in the evolving e-cigarette landscape.
Emerging empirical evidence suggests that e-cigarette may offer harm reduction among current
smokers (Berry et al., 2019; Villanti et al., 2018) but may also cause increased harm among non-smokers
(McMillen, Klein, Wilson, Winickoff, & Tanski, 2019). Widespread misperceptions about the relative
health risks (perceptions about comparative health risks between e-cigarettes and other tobacco products)
and absolute health risks (perceptions about long-term and short-term health risks from e-cigarette use) of
e-cigarette use (vaping) are a public health concern (Huang et al., 2019; Majeed et al., 2017). Studies
comparing relative perceived health risks of e-cigarettes and combustible cigarettes offer mixed findings.
For instance, some findings report that e-cigarettes are perceived to be less risky than combustible
cigarettes (Choi & Forster, 2013; Wackowski & Delnevo, 2015) and less likely to cause diseases such as
2
lung cancer or heart disease (Pepper, Emery, Ribisl, Rini, & Brewer, 2015). Similar findings suggest that
e-cigarettes are also considered to be the least risky and addictive compared to other tobacco products
such as combustible cigarettes and cigars (Berg et al., 2015; Roditis, Delucchi, Cash, & Halpern-Felsher,
2016). On the other hand, studies have also found that e-cigarettes are perceived to be just as harmful as
cigarettes (Majeed et al., 2017). In general, findings about relative perceived risks of e-cigarettes point to,
in part, inaccurate perceptions about the health risks associated with these products. Evidence comparing
absolute perceived risks of e-cigarettes and cigarettes, suggests that e-cigarettes are associated with lower
absolute perceived harm and addictiveness for e-cigarettes (Cooper, Loukas, Harrell, & Perry, 2017).
Timely investigations examining perceived health effects of e-cigarettes can offer actionable insights for
future public health and health communication interventions addressing e-cigarette use.
Several factors may introduce or reinforce misperceptions about the health effects of e-cigarettes.
These include social media exposure to targeted e-cigarette marketing promotions (Allem, Cruz, et al.,
2018; Kirkpatrick et al., 2019), pro-e-cigarette advocacy group messages (Allem et al., 2016), automated
account (or bots)-perpetuated misinformation (Allem, Ferrara, Uppu, Cruz, & Unger, 2017; Jamison,
Broniatowski, & Quinn, 2019), peer-network narratives about these products (Ayers et al., 2017),
inaccurate public interpretation of scientific findings and abuse liability (Huang et al., 2019; Majeed et al.,
2017). There is also a lack of knowledge about the mechanisms and ingredients of e-cigarettes among the
public (Coleman et al., 2016) and health providers (England et al., 2014; Gorzkowski, Whitmore,
Kaseeska, Brishke, & Klein, 2016; Pepper, McRee, & Gilkey, 2014). Exposure to misleading health
claims from marketers or retailers (Basáñez, Majmundar, Cruz, Allem, & Unger, 2019; Klein et al., 2016;
Wagoner et al., 2019) may also contribute to these misperceptions.
Gaps in the understanding of perceived and absolute health effects and risks of vaping may result
in a disproportionate burden of adverse health effects or abuse liability among vulnerable populations.
These populations may include those that are particularly at risk of vaping such as adolescents (Soneji,
Barrington-Trimis, et al., 2017), pregnant women (Oncken et al., 2017; Wigginton, Gartner, & Rowlands,
3
2017), and , sexual gender and ethnic minority youth (Hammig, Daniel-Dobbs, & Blunt-Vinti, 2017;
Hoffman, Delahanty, Johnson, & Zhao, 2018). Vulnerable adolescents, in particular, who are susceptible
to or use e-cigarettes and/or combustible tobacco are exposed to and engage with tobacco-related social
media more than their peers (Hébert et al., 2017).
In clinical settings, doctor-patient interactions about e-cigarettes are rarely initiated by doctors
despite the fact that doctors are the most trusted sources for e-cigarette-related health information (Correa
et al., 2018; Singh et al., 2017; Wackowski, Bover Manderski, & Delnevo, 2015). Surveys of medical
practitioners also suggest that they are uncertain about the health effects of and effectiveness of these
products for smoking cessation (Kollath-Cattano et al., 2019; Pepper et al., 2014). Additionally, medical
practitioners’ lack of knowledge about emerging trends associated with vaping is also an additional
barrier in facilitating doctor-patient communication in clinical settings (Peterson, Fisher, & Zhao, 2018).
Capturing organic, public discussions about the health effects of e-cigarettes, otherwise unreported in
primary care or clinical settings, can provide timely inputs for tobacco control and prevention
interventions.
Need for social media surveillance to characterize the diversity of health effects of vaping
The e-cigarette landscape is characterized by a stream of new products (Zhu et al., 2014), copycat
products (Kirham, 2018), and unregulated products (Blood, 2019) compatible with a wide variety of
substances such as marijuana and tetrahydrocannabinol (THC) wax (Majmundar, Kirkpatrick, Cruz,
Unger, & Allem, 2019). This complicates the assessment of long-term and short-term health effects of
these products (Benowitz & Goniewicz, 2013). In an evolving e-cigarette landscape, characterizing
emerging health effects from vaping is crucial to inform future clinical investigations examining links
between e-cigarette ingredients and mechanisms with resulting health effects. Social media surveillance
insights, in this context, can help address this crucial aspect of the vaping epidemic in the United States.
Surveillance of public perceptions of health effects of e-cigarettes through social media can
potentially offer a compliment to traditional methods used to understand perceived and reported health
4
effects of e-cigarettes. Emerging evidence from traditional method-based inquiries including
pharmacological and survey research suggests that exposure to e-cigarette aerosols is associated with
multiple health risks such as cancer, respiratory and cardiovascular diseases (Grana, Benowitz, & Glantz,
2014; National Academies of Sciences Engineering and Medicine (U.S.). 2018) However, several gaps
exist. For instance, current research examining health effects of e-cigarettes does not capture the diversity
of e-cigarette devices and e-liquids (Benowitz & Goniewicz, 2013; Schroeder & Hoffman, 2014; Walley,
Wilson, Winickoff, & Groner, 2019). E-cigarettes vary in their design, customizability, and engineering
(Zhu et al., 2014). E-liquids vaporized by these devices for inhalation also vary in their composition of
flavorings, moisture-retaining substances (i.e., propylene glycol or glycerol), nicotine and other additives
(Etter, Zather, & Svensson, 2013; Tierney, Karpinski, Brown, Luo, & Pankow, 2016). These variations in
devices and e-liquids determine the amount and quality of nicotine, toxicants and harmful aerosol
released, all of which, have direct consequences for health (Prochaska, 2018). As evidenced by past
research, real-time surveillance of organic public discussions about e-cigarette-related health perceptions
on social media can address the challenges related to the rapidly changing landscape of e-cigarette
products and use behaviors (Ayers et al., 2017). For example, social media surveillance highlighted trends
in dual tobacco product use (Allem, Ferrara, et al., 2017), appeal of device features such as small size
(Kavuluru, Han, & Hahn, 2019) and compatibility with cannabis and tobacco (Majmundar, Kirkpatrick, et
al., 2019), and early mentions of use of Juul among youth in schools (Allem, Dharmapuri, Unger, et al.,
2018).
Twitter, in particular, is used by 24% of US adults (24% of women, 23% of men, 24% of white
individuals, 26% of African American individuals, and 20% of Hispanic individuals) (Perrin & Anderson,
2017). About 46% of Twitter users are estimated to be on the platform daily (Perrin & Anderson, 2017).
Approximately, 32% of adolescents report using Twitter daily (Anderson & Jiang, 2019). This
dissertation demonstrates the utility of collecting data from Twitter to characterize and understand the
communicative and behavioral aspects of sharing perceived health-effects of e-cigarettes on social media.
5
Perceived health effects of e-cigarettes on social media and stance toward e-cigarettes
Perceived health effects of e-cigarettes on social media may have implications for attitudes
towards and patterns of tobacco product use. Social media users may engage with diverse types of
information, including posting, about the health effects of e-cigarettes. Such engagement is likely
associated with overall stance toward e-cigarettes and is an underexplored area of research.
The Elaboration Likelihood Model (ELM) offers a useful framework to explain social media
communication-related processes (Teng, Khong, & Goh, 2014). According to ELM, individuals spend
different levels of effort to cognitively process or elaborate messages depending on their level of
involvement and interest in the topic (Petty & Cacioppo, 1984; Petty & Cacioppo, 1986). When people
are highly involved in a topic and are able to weigh the strengths or weaknesses of an argument, there is a
high likelihood of cognitive elaboration of a message. When the likelihood of such elaboration is high,
information processing will occur via the central or cognitive route, which in turn is linked to strong and
persistent attitudes (Haugtvedt & Petty, 1989). When elaboration likelihood is low, messages are
processed through the peripheral or experiential route. Processing through the peripheral route involves
little cognitive effort because it relies on cues such as instant assessment of a source's credibility and other
quick shortcuts such as visual appeal. Research suggests that attitudes associated with the peripheral route
are temporary and not as predictive of subsequent behavior as those emerging from the central route of
cognitive processing (Petty & Cacioppo, 1983).
Performance of engagement on social media indicates relevance or interest and involves effort
and attention. For instance, in order to post an opinion about a news article on social media, an individual
might need to first consider his or her stance about that topic. These choices involve active elaboration.
Online peer-engagement (i.e., peers liking pictures on Instagram) is associated with increased brain
activity among adolescents, notably in areas responsible for social cognition and memories (Sherman,
Payton, Hernandez, Greenfield, & Dapretto, 2016). It is likely that posting about a topic on social media
will route messages to be centrally processed, determining the subsequent nature of attitudes.
6
Theory also supports that attitudes are highly predictive of behaviors. According to the Integrated
Model for Behavioral Prediction (IMBP) (Fishbein & Cappella, 2006; Fishbein & Yzer, 2003), among
other factors, exposure to media is associated with nature of attitudes, which, in turn, are highly predictive
of behavioral intentions. Past studies examining e-cigarette risk perceptions and attitudes, however, are
based on self-reported responses (Barnett et al., 2013; Heinz et al., 2013; Pokhrel et al., 2016). Organic
and real-time social media data offer an opportunity to infer stance toward e-cigarettes over time and test
associations between the nature of health messages posted and subsequent stance towards these products.
Recently, there has been a surge of reports of vaping-related illnesses in the media, which is likely to
stimulate discussions on this dynamic topic of health effects of e-cigarettes (Rimmer & Iacobucci, 2019).
Social media data, in this context, can allow researchers to capture a diversity of health issues discussed
and use data science methods to synthesize the information for public health practitioners. In other words,
it is possible to track whether posting (or tweeting) about specific topics of conversations related to health
effects of e-cigarettes are associated with the overall stance or attitudes (pro-, anti-, neutral-) towards e-
cigarettes.
Perceived health-effects of e-cigarettes from the lens of evolving use-behaviors
Perceived health effects of e-cigarettes when viewed in the context of abuse liability and appeal,
can offer important clues about emerging vaping behaviors. The Triangulum (Figure 1) offers a
comprehensive framework for understanding evolving patterns of tobacco use at the intersection of
tobacco, marijuana (reference term inclusive of cannabis and THC references) and electronic vaporizers
(McDonald, Popova, & Ling, 2016; Morean, Kong, Camenga, Cavallo, & Krishnan-Sarin, 2015;
T.R.D.R.P., 2016). Research suggests that tobacco products (e.g., blunts) and electronic vaporizers are
often used as delivery devices for marijuana use (Budney, Sargent, & Lee, 2015). Dual or poly-use of
tobacco, marijuana with vaporizers (or vaping marijuana) is also prevalent (Bello et al., 2019; Huh &
Leventhal, 2016). Such behaviors can potentially enhance appeal of tobacco and marijuana products, and
complicate isolation of e-cigarette-related health effects from other substances during public health risk
7
assessments. In other words, co-use or co-administration of tobacco, marijuana and electronic vaporizers
can make it challenging to scientifically assess and communicate adverse health effects attributable to
aerosolization processes, juice composition, nicotine or marijuana (Prochaska, 2018). This has direct
implications for prevention and tobacco control efforts.
From a prevention standpoint, individuals may mistakenly attribute health effects of one
substance (e.g., marijuana) to vaping, which might be challenging to address given lack of conclusive
evidence. Attitudes towards vaping may also be strongly associated with vaping other substances such
that positive attitudes toward vaping may contribute to vaping marijuana or dual use of e-cigarettes and
combustible cigarettes. For instance, emerging evidence suggests that individuals vaping marijuana do so
because of their preference for vaping (over smoking combustible tobacco) that offers benefits such as the
ability to hide or conceal the device, decreased smell, and increased convenience (Morean, Lipshie,
Josephson, & Foster, 2017). Research also demonstrates that vaping is associated with subsequent
marijuana use (Dai, Catley, Richter, Goggin, & Ellerbeck, 2018), and that e-cigarette use initiation is
associated with subsequent progression to combustible tobacco use (Leventhal et al., 2015). Marijuana
use has also been associated with subsequent initiation, persistence and relapse of combustible tobacco
use (Weinberger, Platt, Copeland, & Goodwin, 2018).
From a regulatory standpoint, it is crucial to inform FDA’s product standards procedures
(requirements that determine whether tobacco products meet a specific standard for the protection of
public health) and product review protocols (requirements for e-cigarette manufacturers to provide
evidence that their products are not harmful to health) in light of delivery devices such as open-system
pod mods that are used with nicotine e-liquids and other substances such as marijuana and THC wax.
Such concerns have led to the state of California’s TRDRP research program to evaluated and at the
intersection or reciprocal influences of tobacco, marijuana and electronic vaporizers (Husten & Deyton,
2013). Examples of such devices include hookah pens that deliver aerosolized flavored aldehydes, with or
without nicotine; flavored little cigars now available in electronic versions; open-system pod mods that
8
allow users to aerosolize liquid THC and nicotine (Budney et al., 2015; Kostygina, Glantz, & Ling, 2016;
McDonald et al., 2016). These evolving vaping behaviors and technological innovations may work in
combination with the changing regulatory landscape (e.g., legalization of marijuana) and marketing
influences to enhance product appeal (Baggio et al., 2014; Dutta-Bergman, 2004; Kostygina et al., 2016;
Majmundar, Kirkpatrick, et al., 2019).
Past research also makes a case for studying combustible tobacco, marijuana and e-cigarette use
together, rather than separately to inform regulations and interventions (McDonald et al., 2016). While
long-term health risks of such emerging vaping behaviors are largely undetermined (Budney et al., 2015;
Rabin & George, 2015), public health surveillance insights can offer crucial points of departure for
investigating the regulatory, behavioral and pharmacological aspects of this trend. Surveillance of health
effects in the context of co-occurring mentions of e-cigarettes, tobacco and marijuana on social media can
provide valuable insights for further investigations.
Conceptual framework
This research informs the intersection of tobacco, e-cigarettes, and marijuana by (a)
characterizing categories of perceived health effects of e-cigarettes, (b) investigating associations between
these categories and subsequent attitude or stance (pro-, anti-neutral-) towards e-cigarettes, and (c)
examining patterns of co-occurring mentions of marijuana, tobacco, and e-cigarettes in the context of e-
cigarette-related health effects across multiple years. The following illustration summarizes the
framework of this research:
9
Figure 1. Conceptual framework
Overview of dissertation studies
The primary goal of this dissertation is to characterize and examine relationships between
commonly referenced health effects of e-cigarettes, subsequent stance toward e-cigarettes, and evolving
e-cigarette-use behaviors by leveraging two years of organic, public conversations on Twitter.
While it is established that there are widespread misperceptions about the health effects of e-
cigarettes, there is limited understanding of the prevailing perceptions about the health effects of e-
cigarettes. Study 1 addresses this gap by describing themes of discussions about health-effects of e-
cigarettes over two years (2018 and 2019). Based on earlier research (Callahan-Lyon, 2014; Glantz &
Bareham, 2018; Grana et al., 2014; Hua, Alfi, & Talbot, 2013b; Stratton, Kwan, Eaton, & National
Academies of Sciences Engineering and Medicine (U.S.). Committee on the Review of the Health Effects
of Electronic Nicotine Delivery Systems, 2018), this study determined the following apriori categories of
health effects: cancer (e.g., lung cancer), respiratory (e.g., wheezing), cardiovascular (e.g., cardiac arrest,
chest pain), pain (e.g., chronic back pain), nervous (e.g., dizziness), stress, mental health (e.g., anxiety)
and other health effects. We also do not know whether posting about specific topics of health effects of e-
cigarettes (e.g., pain, cancer) is associated with a specific stance toward e-cigarettes (pro- or anti-). Study
Perceived health-
effects of e-cigarettes
E-cigarettes
Combustible
tobacco
Marijuana
Attitudes
about
e-cigarettes
Behavior
10
2 investigates the relationship between categories of health effects and overall stance toward e-cigarettes
(pro-, anti-, neutral-) while demonstrating a methodology to quantify meaningful concepts from text as
data while taking into account their temporal order. Study 3 investigated key topics of conversation at the
intersection of tobacco, marijuana and e-cigarettes. It also demonstrates a novel surveillance approach
based on a well-known public health framework to illuminate trends related to changing patterns of
tobacco use. Taken all together, findings from this dissertation research will address questions of public
health and regulatory significance.
FDA Public Health Education Efforts: Findings will provide inputs for more targeted and tailored
mechanisms of tobacco prevention messages. Addressing specific perceived health effects can play an
important role in enhancing message specificity, relevance and overall tailoring of health messages
(Kreuter & Wray, 2003). By using systematic approaches of surveillance, this dissertation will provide
inputs related to public perceptions of health effects of e-cigarettes for future health communication
campaigns and prevention interventions (Study 1). An understanding of the downstream effects of posting
about health effects of e-cigarettes (Study 2) can help health communication researchers prioritize
predominant perceived health effects of e-cigarettes that are associated with a positive stance towards e-
cigarettes and leverage commonly referenced health effects associated with a negative stance. To address
emerging use behaviors, Study 3 identifies key topics of discussion at the intersection of the tobacco
Triangulum to inform future online, targeted and tailored health communication campaigns. Past work
has developed near real-time delivery mechanisms that target individuals posting pro-tobacco messages
on Twitter (Deb et al., 2018). Similar online interventions can target tailored health messages at groups
that engage with social media content at the intersection of the Triangulum and address prevailing
misinformation or misperceptions related to those emerging tobacco-use behaviors.
Tobacco regulations: First, findings will inform formulation of comprehensive Food and Drug
Administration (FDA) product standards and product review requirements. The FDA has the authority to
develop product standards - requirements that tobacco products meet a specific standard for the protection
11
of public health. These include restriction or elimination of harmful constituents or evaluation of product
appeal (Husten & Deyton, 2013). FDA’s product reviews require e-cigarette manufacturers to provide
evidence that their products are not harmful to health. As per the most recent finalized FDA guidance
(F.D.A., 2019), manufacturers of new e-cigarette products (not commercially marketed in the United
States as of February 15, 2007; including product modifications) will be required to submit applications
for premarket tobacco product applications (PMTA). Under PMTA, manufacturers furnish evidence about
associated health-effects, if any, from randomized controlled trials testing the new tobacco products or
comparable products. The furnished evidence could be in the form of meta-analysis of existing literature
or from primary research directly pertinent to the new tobacco product.
Findings from this research will offer insights on the perceived health-effects of e-cigarettes that
can be addressed in the PMTAs (Study 1). Knowledge about which of these categories translate to
downstream effects in the form of pro- or anti- stance toward e-cigarettes will indicate potential points of
appeal that warrant regulatory considerations (Study 2). Insights from health-related discussions in the
context of the tobacco Triangulum will also highlight trends in abuse liability (Study 3).
A key innovative feature of this dissertation is its motivation in harnessing the potential of large-
scale online social media data to contribute to the evidence base about the harm of vaping products and
subsequently inform future FDA regulatory decisions regarding regulating vaping products. By using
cutting-edge data science methods of surveillance, this dissertation will inform the paradigm of tailored
public education framework and FDA’s product review framework and products standards for e-
cigarettes.
12
Chapter 1: Monitoring Twitter discourse about health-effects of
vaping, 2018-2019
Abstract
Background: Despite rising popularity of electronic nicotine delivery systems (ENDS) or e-cigarette use
(vaping), there is a limited understanding of public discussions about the health effects of these products.
Timely social media surveillance of commonly referenced health effects of e-cigarettes can capture the
public’s evolving understanding of associated health risks in a rapidly changing e-cigarette landscape.
The current study characterizes public conversations about the health-effects of e-cigarettes on Twitter.
Methods: Twitter posts with references to health-effects of vaping on Twitter (n= 524,159 posts) were
collected over two years (2018-2019). Rule-based classifiers based on a dictionary of medical terms and
corresponding colloquial terms were used to classify Twitter posts to the following categories: Cancer,
Respiratory, Cardiovascular, Neurological, Pain, Stress, Mental Health, Gastrointestinal, Weight,
Pregnancy/In-utero, Injury, Immunity, Death, and Other health effects.
Results: In 2019, Death was most commonly referenced (n= 77719 posts, 11.98%), followed by
Neurological effects (n= 67504 posts, 10.40%) and Respiratory (n= 48875, 7.52%) health effects in 2019.
In 2018, Neurological (n= 44345 posts, 11.51%) was the most common topic, followed by Mental Health
(n= 31227 posts, 8.11%) and Death (n= 27040 posts, 7.02%).
Conclusion: References to Death, Neurological and Respiratory health effects were predominant e-
cigarette-related health categories of conversation in the past two years. Characterizing public
understanding of health effects of vaping may be crucial inputs for health intervention campaigns and in
clinician settings. Future campaigns can enhance awareness of health risks associated with adverse
outcomes related to the predominant categories of health effects: Death, Mental Health, and Neurological.
13
Introduction
Despite rising popularity of electronic nicotine delivery systems (ENDS) or e-cigarette use
(vaping) (Ayers, Ribisl, & Brownstein, 2011), there is a limited understanding of public discussions about
the health effects of these products. While scientific evidence about the long-term health effects is
inconclusive, emerging findings suggest that vaping is linked to cancer, cardiovascular, respiratory, and
other health issues (Callahan-Lyon, 2014; Glantz & Bareham, 2018). Recent findings also suggest that
youth may deny that e-cigarette are a health hazard (Martinez, Hughes, Walsh-Buhi, & Tsou, 2018) and
perceive improved health after use (Hart et al., 2018). Such misperceptions have important implications
for abuse liability, wherein positive perceived health effects may encourage experimentation and
recreational e-cigarette use (Blundell, Dargan, & Wood, 2018; Budney et al., 2015; Hart et al., 2018).
Timely surveillance of e-cigarette-related public conversations can provide insight to a rapidly changing
landscape (Barrington-Trimis & Leventhal, 2018).
The current e-cigarette landscape may create adverse implications for public health. E-cigarettes
devices are available in the form of open-system pod mods (modifiable devices compatible with a range
of e-liquid solutions such as nicotine or cannabis solutions), closed-system pod mods (prefilled e-liquid
cartridges inserted into a closed-system e-cigarette device), disposable pods (single-use pods), and other
versions (e.g., tanks) (Galstyan, Galimov, & Sussman, 2018; Majmundar, Kirkpatrick, et al., 2019). These
devices vary in their engineering and design (Zhu et al., 2014), which determines the nature of aerosol
released during use, including quality and quantity of nicotine, toxicants, and carcinogens (Prochaska,
2018), E-liquids, solutions aerosolized by e-cigarette devices to produce vapor, also vary in their nicotine
concentrations, quantity of moisture-retaining substances (i.e., propylene glycol or glycerol), flavors and
other substances such as cannabis (Etter et al., 2013; Tierney et al., 2016).
This diversity in e-cigarettes and e-liquids may enhance appeal, lower perceived risk, and
encourage experimentation – all of which may have serious implications on health. For instance, Allem,
Dharmapuri et al. (2018) demonstrated that the compact and small Juul pods are used stealthily in schools
14
during class time. Majmundar et al (2018) revealed that open-system pod mods are increasingly
referenced in the context of both nicotine and cannabis substances, which raises concerns about abuse
liability. A recent analysis of e-liquid-related conversations on social media found that key topics of
conversations included flavors, cannabis, and juice composition, among others (Allem, Majmundar,
Dharmapuri, Cruz, & Unger, 2019). Research using social media trace data comprising of organic public
perceptions about health effects of e-cigarettes is warranted.
Current literature offers a helpful overview of the public’s overall understanding of relative risks
of e-cigarettes compared to other tobacco products. Studies suggest that e-cigarettes are considered to be
less harmful (Choi & Forster, 2013; Majeed et al., 2017; Wackowski & Delnevo, 2015), least addictive,
and least risky compared to combustible tobacco products (Berg et al., 2015; Roditis et al., 2016). A gap
in literature pertains to a characterization perceived health effects of e-cigarettes to offer inputs regarding
specific health effects worth highlighting in future health communication campaigns and in patient-
provider settings.
Social media surveillance offers a rapid method to describe topics of conversations about e-
cigarettes. As noted in previous research, Twitter conversations capture organic conversations at a low
cost and without the prime of a researcher (Allem, Dharmapuri, Unger, et al., 2018; Allem et al., 2016;
Allem, Ferrara, et al., 2017; Allem, Cruz, et al., 2018; Vance et al., 2009). Surveillance research drawing
data from social media platforms has been used to characterize different aspects of e-cigarettes in the past.
For example, Ayers and colleagues used Twitter data to describe reasons for e-cigarette use (Ayers et al.,
2017). They found that the predominant reasons for vaping was that individuals liked the social image
portrayed during vaping (e.g., looking “cool”). Other studies have characterized e-cigarette-related
conversations on social media to reveal targeted marketing practices through the use of cartoons (Allem,
Cruz, et al., 2018; Kirkpatrick et al., 2019), misleading messages highlighting e-cigarettes as ‘vitamin
delivery’ devices (Basáñez et al., 2019), and negative responses from the public and pro-tobacco groups
to e-cigarette regulations and health education campaigns (Allem et al., 2016; Harris et al., 2014; Lazard
15
et al., 2017). Examining aggregated Twitter posts about health-effects of e-cigarettes, in essence, can
serve as a large focus group on this issue.
In the present study, public conversations about the health-effects of e-cigarettes were
characterized using Twitter posts over two years (2018 and 2019). Based on prior research (Callahan-
Lyon, 2014; Glantz & Bareham, 2018; Glasser et al., 2017; Grana et al., 2014; Hua et al., 2013b; National
Academies of Sciences Engineering and Medicine (U.S.). 2018; Pisinger & Dossing, 2014; Rhoades et
al., 2019), categories of health effects were determined apriori. We identified and distinguished key
categories by year (2018, 2019). To the best of our knowledge, only one study to date has examined the
reported health effects of vaping as described on online discussion forums (Hua, Alfi, & Talbot, 2013a).
Findings from this research will provide an updated perspective on public conversations about the health-
effects of e-cigarettes.
Methods
Twitter (https://Twitter.com/) posts containing e-cigarette related terms (e.g., ‘vape’, ‘e-
cigarettes’, ‘Juul’) drawn from previous research (Ayers et al., 2017; Chu, Allem, et al., 2019; Chu,
Allem, Cruz, & Unger, 2016), were obtained from January 1, 2018 to December 31, 2019 using Twitter’s
Streaming Application Program Interface (API). Data cleaning procedures included removal of retweets,
non-English posts, posts from bot accounts (Allem & Ferrara, 2016), duplicate posts that were not
retweets, spam posts and promotional posts. Text in the posts were converted to lower case, and features
such as punctuation, special characters, hyperlinks, and hashtags were removed. This resulted in an initial
sample of 5,248,063 e-cigarette-related posts (2018: 2,516,664 posts, 2019: 2,731,399 posts).
To identify posts containing health-related references to e-cigarettes, rule-based classifiers were
used to categorize text based on two sources. The first source comprised of relevant terms from the
Unified Medical Language System® (UMLS) Consumer Health Vocabulary (CHV), (Bodenreider, 2004)
(Zeng & Tse, 2006). CHV includes lay terms, recorded in Electronic Health Records, used by patients to
describe their health conditions (e.g., lay term for cardiovascular disease is ‘heart disease’) (n=13479
16
keywords). The second source comprised of a list of informal terms corresponding to the keywords in the
first source described above (e.g., informal term of ‘inebriation’ is ‘drunk’). This second source of
keywords (n=177 terms) was generated by two independent coders. Keywords were also expanded to
account for misspellings or shorthand (e.g., ‘heart ‘spelled as ‘hrt’) based on a manual scan of a subset of
randomly selected tweets. Keywords form the above two sources encompassed health diseases and
symptoms (referred to as ‘health effects’ from hereon). The analytic sample comprised of 524,159 e-
cigarette-related posts with mentions of health-effects (2018: 206941 posts, 8.26% of vape-related tweets;
2019: 314964 posts, 11.5% of vape-related tweets).
Next, each post in the analytic sample was classified at least one of the apriori categories (e.g.,
respiratory, pain). Apriori categories of health effects were: Cancer, Respiratory, Cardiovascular,
Neurological, Pain, Stress, Mental Health, Gastrointestinal, Weight, Pregnancy/In-utero, Injury,
Immunity, Death, and Other health effects. In some cases, one post was categorized under more than one
topic (e.g., “vaping makes me feel dizzy but less stressed”, maybe classified under ‘Neurological’ and
‘Stress’). Please see Table 1 for example keywords associated with each category.
Table 1. Predominant categories of health effects and example keywords
Health Categories Example keywords
Neurological Coma, dizzy, lightheaded
Mental health PTSD, ADHD, jittery
Death Die, kill, lost life
Injury Injury, rupture, wound, bruise
Respiratory Cough, wheeze, black lung
Pain Painful, achy, cramping
Cancer Cancer, tumor, malignant
Gastrointestinal Belly, belch, vomit, puke
Cardiovascular Stroke, heart attack, blood pressure
Weight-related Fat, obese, weight, stoutness
Stress Stressed, cortisol
17
Immunity Flu, common cold, allergy
Pregnancy/In-utero Pregnant, preggers, miscarriage
Other Anemia, jaundice, mumps
The analyses relied on public, anonymized data, and adhered to the privacy policies, terms and
conditions, terms of use of Twitter. This study was performed under Institutional Review Board approval
from the lead author’s university. No Twitter posts are reported verbatim in this study to protect the
privacy of the users. For a complete list of e-cigarette-related keywords, see Table 2.
Table 2. E-Cigarette-related keywords
'ecig', 'ecigs', 'ecigarette', 'ecigarettes', 'e-cigarette', 'e-cigarettes', 'e-liquid', 'e-liquids',
'eliquid', 'eliquids', 'ejuice', 'ejuices', 'e-juice', 'e-juices', 'vaping', 'vapes', 'vape', 'vaper', 'Juul',
'juuling'
Results
In 2019, Death was most commonly referenced (n= 77719, 11.98%) followed by references to
Neurological (n= 67504, 10.40%) and Respiratory (n= 48875, 7.52%) health effects in 2019. The
predominant categories remained the same even in terms of the total number of references (after including
overlapping categories). A breakdown of the distribution of tweets across health categories (including
overlapping categories) during 2019 is presented in Figure 2.
18
Figure 2. Percentage distribution of categories in 2019
In 2018, Neurological health effects (n= 44345, 11.51%) was most common, followed by Mental
Health (n= 31227, 8.11%) and Death (n= 27040, 7.02%). A breakdown of the distribution of tweets
across health categories (including overlapping categories) during 2018 is presented in Figure 3. In terms
of the total number of references (after including overlapping categories), Neurological was the most
common category (n= 44485, 14.40%), followed by Mental Health (n= 42568, 11.05%) and Pain (n=
28922, 10.10%).
death 11.98
neurological 0.42 10.40
respiratory 1.48 0.45 7.53
mental health 0.30 0.44 0.22 7.09
injury 0.53 0.27 1.55 0.12 6.20
other 0.42 0.53 0.92 0.40 0.67 5.00
pain 0.13 0.23 0.18 0.50 0.09 0.45 4.11
gastrointestinal 0.12 0.19 0.22 0.14 0.19 0.23 2.04 3.00
cancer 0.32 0.18 0.89 0.11 0.12 0.30 0.07 0.08 3.00
cardiovascular 0.12 0.10 0.20 0.12 0.06 0.21 0.09 0.04 0.12 1.44
weight 0.05 0.07 0.05 0.04 0.02 0.07 0.03 0.04 0.01 0.02 1.40
stress 0.03 0.07 0.05 0.31 0.02 0.09 0.18 0.03 0.02 0.05 0.02 1.28
immunity 0.04 0.02 0.10 0.03 0.03 0.08 0.03 0.02 0.01 0.01 0.00 0.00 0.61
pregnancy.
Inutero
0.01 0.01 0.01 0.01 0.01 0.02 0.01 0.01 0.01 0.00 0.00 0.00 0.00 0.34
TOTAL 15.96 13.38 13.85 9.82 9.87 9.40 8.15 6.34 5.24 2.59 1.85 2.15 0.97 0.44
death neurological respiratory
mental
health
injury other pain gastrointestinal cancer cardiovascular weight stress immunity
pregnancy/
inutero
19
Figure 3. Percentage distribution of categories in 2018
Figure 4 provides an overview of distribution of categories in 2018 and 2019. Compared to 2018,
Death, Mental Health, and Respiratory were more prevalent categories in 2019. While Death, Mental
Health, and Neurological were one of the most predominant categories during 2018 and 2019, Respiratory
health effects emerged as part of the top five categories in 2019.
neurological 11.51
mental health 0.69 8.11
death 0.22 0.16 7.02
injury 0.22 0.11 0.27 6.93
respiratory 0.26 0.18 0.32 0.28 6.20
other 0.54 0.45 0.27 0.94 0.62 5.84
pain 0.28 0.53 0.10 0.09 0.12 0.57 5.20
cancer 0.26 0.23 0.19 0.21 0.86 0.37 0.18 5.12
gastrointestinal 0.20 0.18 0.11 0.23 0.21 0.26 2.76 0.11 4.01
cardiovascular 0.06 0.06 0.06 0.07 0.12 0.15 0.06 0.33 0.03 2.15
weight 0.06 0.05 0.04 0.03 0.05 0.12 0.04 0.07 0.05 0.02 2.08
stress 0.07 0.28 0.03 0.02 0.04 0.10 0.14 0.04 0.04 0.03 0.01 1.75
immune system 0.02 0.02 0.02 0.03 0.18 0.12 0.02 0.01 0.02 0.01 0.00 0.00 0.97
pregnancy
inutero
0.01 0.01 0.03 0.01 0.01 0.02 0.01 0.01 0.01 0.00 0.01 0.00 0.00 0.29
TOTAL 14.40 11.05 8.82 9.42 9.45 10.36 10.10 8.01 8.21 3.16 2.65 2.55 1.42 0.41
neurological
mental
health
death injury respiratory other pain cancer
gastro-
intestinal
cardiova
scular
weight stress immunity
pregnancy/
inutero
20
Figure 4. Percentage distribution of categories of health effects during 2018 and 2019
Discussion
Findings from this research demonstrate the utility of social media data in informing public
surveillance and health interventions. In general, references to Neurological, Death, Mental Health, and
Respiratory, were some of the most predominant categories in the past two years. Compared to 2018,
Death, Mental Health, and Respiratory health effects were more prevalent in 2019. While Death, Mental
Health, and Neurological health effects were one of the most predominant categories during 2018 and
2019, the topic of respiratory health effects emerged as part of the top five categories in 2019.
Characterizing public understanding of health effects of vaping may be crucial inputs for health
intervention campaigns and in clinician settings. Future campaigns can enhance awareness of adverse
outcomes related to the predominant categories of health effects: Death, Mental Health, and Neurological.
While recent vaping prevention campaigns (F.D.A., 2020; Truth Initiative, 2019) used narrative health
communication strategies to highlight the specific health outcomes of vaping, this study suggests
additional categories of health effects. Evidence also suggests that addressing risks associated with e-
cigarette use is an important topic in a patient-primary care provider setting (Peterson et al., 2018). This
study may also inform specific categories of health effects that may be addressed in a patient-provider
setting. Categories of health effects such as Cancer, Cardiovascular diseases that were not as predominant
may also warrant adequate consideration, while discussing the potential long-term health risks of vaping.
21
There is limited work on e-cigarette use patterns and health perceptions among vulnerable
populations who suffer from mental health disorders. Chronic smokers who also suffer mental illness
such as depression and anxiety reported high satisfaction from vaping (Pratt, Sargent, Daniels, Santos, &
Brunette, 2016). Perceived barriers to vaping in these populations include vaping being an unsatisfactory
substitute for psychiatric medicines, contributing to drug interactions, nicotine addiction and health risks
associated with certain e-liquids (Sharma, Wigginton, Meurk, Ford, & Gartner, 2016). Findings from this
study reveals that mental health-related references were one of the most predominant categories during
2018 and 2019. Future health research may consider investigating e-cigarette-use prevalence, trajectories,
and health risks among those suffering from mental health disorders.
Findings pertaining to references to Pain and vaping, although not predominant, may also warrant
consideration in future investigations. Research suggests that vaping maybe be associated with pain
management when co-used with marijuana (Wilsey et al., 2013; Wilsey et al., 2016). Future
investigations may consider identifying and addressing high perceived benefits of vaping for pain
management. References to pain may also be syndromic in nature. In other words, in line with recent
work analyzing trends in emergency department codes for EVALI cases in 2019 (Hartnett et al., 2019),
pain-related symptoms were one of the most common as a consequence of vaping. More research is
needed to clarify the implications of vaping on physiological pain.
In 2018 and 2019, long term or chronic conditions associated with Cancer and Cardiovascular
health effects were less commonly referenced compared to immediate term effects tied to Injuries and
Deaths. While the EVALI outbreak may have drawn public attention to relatively short-term health risks
of vaping such as shortness of breath and chest pain, potentially longer-term health risks are also of
concern. FDA’s Future health communication campaigns may consider increasing awareness of both
long-term and short-term health effects of vaping among vulnerable populations. The FDA may also
consider including short-term and longer-term health risks of vaping in health warning messages on e-
cigarette packages and promotions on social media. Future research may also explore self-reported health
22
effects of e-cigarettes associated with high perceived short-term and long-term health risks among
adolescents and young adults to be addressed in prevention campaigns and interventions.
Per the most recent finalized FDA guidance (F.D.A., 2019), manufacturers of new e-cigarette
products (not commercially marketed in the United States as of February 15, 2007; including product
modifications) will be required to submit applications for premarket tobacco product applications
(PMTA). Under PMTA, manufacturers furnish evidence about associated health-effects, if any, from
randomized controlled trials testing the new tobacco products or comparable products. The furnished
evidence could be in the form of meta-analysis of existing literature or from primary research directly
pertinent to the new tobacco product. Findings from this research offer insights on commonly discussed
and/or experienced health effects of e-cigarettes that may be addressed in the PMTA applications.
Limitations
This study drew data from Twitter and findings may not generalize to other social media
platforms. Findings may also not represent data from individuals with private Twitter accounts. The time
range of the data is January 2018 to December 2019, and findings may not generalize to other years.
While many health effects were considered to understand predominant categories of conversations,
specific references to biological basis of these effects (e.g., mentions of blocked arteries in the context of
heart dysfunction) may not be captured in this study.
Conclusion
Findings from this two-year investigation of the most common health-related categories of vaping
on Twitter revealed that references to Death, Neurological, and Respiratory health effects are some of the
most predominant categories in the past two years. Characterizing public understanding of health effects
of vaping may be crucial inputs for health intervention campaigns and in clinician settings. Future
campaigns can enhance awareness of health risks associated with adverse outcomes related to the
predominant categories of health effects: Death, Mental Health, and Neurological.
23
24
Chapter 2: Associations between categories of health effects of e-
cigarettes and stance towards e-cigarettes
Abstract
Background: E-cigarette users at times discuss the health effects of e-cigarette use on social media.
Health interventions could be strategically designed to communicate the known health risks from e-
cigarette use when e-cigarettes users are taking part in these discussions and communicating a positive
stance about e-cigarettes. This study examined the associations between perceived health effects from e-
cigarette use and subsequent stance (pro-, anti-, neutral-) toward e-cigarettes expressed on Twitter.
Hypotheses: (A) Twitter posts about categories pertaining to: (a) Pain, (b) Stress are more likely to be
pro-e-cigarette compared to Mental Health effects, (B) Twitter posts about categories pertaining to: (a)
Cancer, (b) Cardiovascular, (c) Neurological, (d) Other effects are less likely to be anti-e-cigarette
compared to Respiratory health effects.
Methods: E-cigarette-related Twitter posts from 2000 unique accounts pertaining to one of seven
Categories of Health Effects of E-Cigarettes: Neurological, Pain, Stress, Cancer, Cardiovascular, Mental
Health, Other effects were identified using the Unified Medical Language System medical dictionary and
selected using stratified sampling procedures. Next, Baseline- and Post-Stance toward E-Cigarettes for
each Twitter account was identified based on the temporal order of e-cigarette-related posts not pertaining
to their health effects, and annotated in terms of anti-, pro-, or neutral-stance. Finally, adjusted
multinomial regression analysis was used to assess associations between Categories of Health Effects of
E-Cigarettes and Post-Stance toward E-Cigarettes.
Results: The majority of posts were classified as Pro-Stance toward E-Cigarettes (n=982) followed by
Anti-Stance (n=509) and Neutral-Stance (n=509). Neurological was commonly associated with a Pro-
Stance (n=302, 30.75%) followed Mental Health, and Respiratory. Findings did not support the
hypotheses. Contrary to the hypothesis, Pain was significantly less likely (RRR = 0.52, p=0.003) to be
pro-e-cigarette compared Mental Health Effects.
Conclusion: This study uses temporal analysis of organic vaping-related Twitter discussions to describe
the varying positions toward e-cigarette use as they relate to perceived health effects from use. While
some e-cigarette users discuss a health issue in a negative frame towards e-cigarettes, these attitudes were
not universal. Interventions could be designed around discussions of posts to Twitter however findings
may only be limited to those with negative stances. Further research is needed to identify ways to
communicate the health risks of e-cigarette use.
25
Introduction
Social media platforms have emerged as popular public venues for discussions about health-
related topics(Fox, 2011; Hausmann et al., 2017; Vance et al., 2009; Vraga & Bode, 2018). Individuals
share their health-related intentions, perceptions, concerns, and attitudes (Orr, Baram-Tsabari, &
Landsman, 2016; Zhang & Yang, 2014), which may be driven by unanticipated (e.g., policy changes) or
anticipated events (e.g., personal decision to quit tobacco) (Orr et al., 2016; Park & Conway, 2017).
Tobacco-related discussions, in particular, are contextualized in light of information from news articles
(Lazard et al., 2017), personal experiences (Allem, Dharmapuri, Unger, et al., 2018), marketing messages
(Allem, Cruz, et al., 2018), and peer discussions (Chu et al., 2018; Chu et al., 2015). Past work suggests
that e-cigarette users actively use social media to obtain and share information about these products (Link
et al., 2015), including discussing health effects experienced from product use (Park & Conway, 2017).
Health interventions could be strategically designed to communicate the known health risks from e-
cigarette use when e-cigarettes users are taking part in these discussions.
Health-effects of e-cigarettes differ in their severity and outcomes. Additionally, not all health
effects may be viewed as adverse consequences of e-cigarette use, for instance, studies suggest that
individuals who vape, do so to manage their allergies, pain(Hart et al., 2016; Wilsey et al., 2013; Wilsey
et al., 2016), and stress(Gibson et al., 2018). As such, some health effects of e-cigarette use may be
associated with a positive stance toward e-cigarette use whereas adverse effects may be linked to a
negative stance toward vaping. Stance in this context, is conceptualized as an expression of attitudes
towards a specific target or issue (Charles, 2003). For instance, individuals posting about first-hand or
second-hand adverse health events such as dizziness or nausea due to vaping, might harbour a negative
stance toward e-cigarettes. Alternatively, those experiencing anxiety-relief may harbour a positive stance
toward e-cigarettes. Regardless, there may be an opportunity to target tailored tobacco messages for
individuals discussing the health effects of e-cigarette use.
26
This study examined the various categories of perceived health-related effects from e-cigarette
use and, at the same time, distinguishes between pro-, negative- and neutral- stance toward e-cigarettes
expressed subsequently. Social media data offer opportunities to track trajectories of organic
conversations about health effects of e-cigarettes and infer overall stance towards these products. Unlike
self-reported data that suffer from limitations of response bias (Rosenman, Tennekoon, & Hill, 2011),
text-based social media conversations are unprimed by the researcher, and offer opportunities to
understand the public’s attitudes and experiences in their own words. Additionally, unlike sentiment
analysis where the goal is to assess whether a text is positive or negative, this study identifies stance
toward e-cigarette use, which involves detection of overall favourability of a target, which may or may
not be mentioned explicitly(Krejzl, Hourová, & Steinberger, 2017).
This study builds on categories of health effects identified in Study 1. It leverages Twitter data to
test the following hypotheses:
Hypothesis 1: Post about categories pertaining to: (a) pain, (b) stress are more likely to be pro-e-
cigarette compared to those posting about mental health effects.
This is due to recent research suggesting that vaping is associated with pain management when
co-used with marijuana, (Wilsey et al., 2013; Wilsey et al., 2016) and reducing stress.(Gibson et al., 2018)
In contrast, chronic smokers who also suffer from adverse mental health including depression and
anxiety, actively shared concerns about barriers to vaping, including unsatisfactory substitute for
psychiatric medicines, drug interactions, nicotine addiction and health risks associated with certain e-
liquids (Sharma et al., 2016).
Hypothesis 2: Post about categories pertaining to: (a) cancer, (b) cardiovascular, (c) neurological,
(d) other health effects are less likely to be anti-e-cigarette compared to those posting about
respiratory health effects.
27
While harmful vaping-related health consequences such as cancer, respiratory and cardiovascular
effects may be perceived negatively in general (Pepper et al., 2015), there has been a sharp increase in
adverse respiratory health effects of vaping that hamper normal lung functioning among e-cigarette users
(Rimmer & Iacobucci, 2019; Stiles, 2019). Given valence of news coverage is also known to influence e-
cigarette-related attitudes and beliefs (Tan, Lee, Nagler, & Bigman, 2017), widespread discussions about
and news media coverage of respiratory effects of vaping are likely to be more anti-vaping than other
adverse health effects such as cancer and cardiovascular diseases.
Findings should inform the development of future interventions aimed at targeting members of
the public in a position amenable to receiving information about the health risks from e-cigarette use.
Methods
This section is divided into Data, Measures, Procedures, and Analytic Approach.
Data
Twitter (https://Twitter.com/) posts containing e-cigarette-related terms (e.g., ‘ecig', ‘e-
cigarettes’) were obtained from January 1, 2018 to December 23, 2019 using Twitter’s Streaming
Application Program Interface (API). All data collected is publicly available, that is, anyone with an
internet connection can view the post, unless the account deleted it.
Twitter account ids from Study 1’s analytic sample of 309,607 unique user accounts pertaining to
health-effects of e-cigarettes were retrieved. Please see Figure 5 for an overview of the study design
(inclusion criteria). For every Twitter account in the sample the following sequences were examined: An
initial post unrelated to health effects of e-cigarettes established a baseline, a subsequent health effects-
related post (the explanatory variable), and a third an e-cigarette post unrelated to health effects to
measure an outcome. As a result, the analytic sample consisted of 17,127 posts from 5,709 unique Twitter
accounts.
28
For each account, post pertaining to health-effects of e-cigarettes was classified into one of eight
categories of health effects (cancer, neurological, respiratory, cardiovascular, stress, pain, mental health,
other health effects) (accomplished in Study 1). For each account, e-cigarette-related posts not pertaining
to health posted prior to and after the health-effect-related post were labelled in terms of anti-, pro-, or
neutral-stance.
Next, stratified sampling procedures based on categories of health-effects were used to create a
final representative analytic sample consisting of 2000 user accounts (consisting of 2000 posts with
mentions of health-related effects of e-cigarettes and 4000 e-cigarette-related posts not pertaining to
health-effects of e-cigarettes). The initial proposed approach was to use machine learning to auto-label a
larger dataset, this method yielded sub-optimal prediction results (F-scores) (see Appendix for more
details). As an alternative method, a smaller manually annotated representative sample was used to
investigate the hypotheses.
Measures
The primary outcome variable, Post-Stance toward E-Cigarettes, was defined as a pro-, anti- or
neutral- position toward e-cigarettes subsequent to posting about health-effects of e-cigarettes. The
primary explanatory variables were individual categorical variables pertaining to Categories of Health-
Effects of E-Cigarettes (e.g., cancer, pain, stress). Covariates included Baseline-Stance toward E-
cigarettes, defined as a pro-, anti- or neutral- position toward e-cigarettes prior to posting about health-
effects of e-cigarettes, and Mentions of Marijuana in posts pertaining to health-effects of e-cigarettes
(Yes/No).
Procedures
The primary explanatory variables pertaining to Categories of Health Effects of E-Cigarettes were
drawn from Study 1, that used rule-based classifiers on e-cigarette-related posts to classify them to
appropriate health-effect-related topic(s). The outcome variable Post-Stance toward E-cigarettes and
covariate, Baseline Stance toward E-cigarettes, comprised of manually annotated data. Two independent
29
coders labelled 400 e-cigarette-related posts (unrelated to health-effects of e-cigarettes), in terms of their
stance toward e-cigarettes (pro-, anti- or neutral- toward e-cigarettes). Inter-coder reliability was
determined to be excellent (Cohen’s Kappa: 0.97). The codebook for manual annotation procedures,
clarified the difference between Stance and Sentiment. For instance, a positive sentiment or expression
may be used to convey an anti-e-cigarette stance (e.g., ‘I appreciate your argument in supporting ecig
products which are harmful to health’). Similarly, a negative sentiment may be used to convey a pro-e-
cigarette stance (e.g., ‘Telling me not to vape is not going to work’). The manual annotation training also
emphasized that determining the stance towards e-cigarettes involved an understanding of figurative
language that may involve broader interpretation of rhetorical features such as sarcasm, humor, and irony.
For instance, the sentence, ‘The last time I checked, I did not think that vaping was cool’, includes
sarcasm used to convey an anti-vaping stance.
Rule-based classifiers were used to determine presence of marijuana-related keywords (e.g.,
‘marijuana, ‘weed’) in e-cigarette-related posts pertaining to health-effects of e-cigarettes (Yes/No). Posts
to Twitter with at least one marijuana-related keywords were classified as ‘Yes’, and posts having no
references to marijuana-related keywords were classified as ‘No’.
Analytic approach
The unit of analysis was individual post to Twitter. Adjusted multinomial regression analyses
were conducted to evaluate the association of Categories of Health-Effects of E-Cigarettes identified in
Study 1 (indicator variable) and Post-Stance toward E-Cigarettes. Covariates included Mentions of
Marijuana in the posts pertaining to health-effects of e-cigarettes, and Baseline Stance toward E-
Cigarettes.
30
Figure 5. Inclusion criteria and sampling
Notes:
• HR= E-cigarette-related posts to Twitter that include keywords pertaining to health-related effects
• ECNH= E-cigarette-related posts to Twitter not pertaining to health-effects of e-cigarettes
E-cigarettes-related posts,
unrelated to health effects of e-
cigarette use posted at baseline
(Baseline ECNH)
E-cigarette-related posts
pertaining to health effects
of e-cigarette use
(HR)
Classify posts in terms of
stance towards e-
cigarettes (pro-, anti-,
neutral-)
Classify in terms of
Neurological, Pain,
Stress, Cancer,
Cardiovascular,
Mental Health, Other
effects
E-cigarettes-related posts,
unrelated to health effects of e-
cigarette use subsequent to HR
(Post ECNH)
Classify posts in terms of
stance towards e-
cigarettes (pro-, anti-,
neutral-)
For each Twitter account
Baseline-Stance
(Covariate)
Health category
(Explanatory variable)
Post-Stance
(Outcome variable)
Stratified sampling of user accounts by health categories
(n=2000 user accounts; n=2000 HR posts, 2000 Baseline-
Stance posts and n=2000 Post-Stance posts)
31
Results
Table 3 presents descriptive statistics of Post-Stance towards E-Cigarettes. Pro-Stance was more
predominant (n=982, 49.1%) followed by Anti-Stance (n=509, 25.25%) and Neutral Stance (n=509,
25.25%). Neurological (30.75%) was the most common category among Pro-Stance posts followed by
Mental Health (24.85%), Respiratory (14.26%), Pain (9.57%), Stress (5.30%), Cardiovascular (1.32%),
Cancer (4.79%), and Other health effects (9.16%). Among Anti-Stance posts, Neurological (37.92%) was
the most common category followed by Mental Health (15.32%), Respiratory (15.32%), Pain (9.57%),
Cancer (4.79%), Stress (3.73%), Cardiovascular (2.55%) and Other health effects (9.63%). A majority of
Pro-Stance, Anti-Stance, or Neutral-Stance did not mention marijuana in their posts.
32
Table 3. Descriptive statistics of Post-Stance towards e-cigarettes
Pro-E-
Cigarettes
(n=982)
Anti-E-
Cigarette
(n=509)
Neutral-E-
Cigarette
(n=509)
p-value
Categories of health
effects
0.001*
Neurological 302 (30.75%) 193 (37.92%) 156 (30.65%)
Mental Health 244 (24.85%) 78 (15.32%) 84 (16.50%)
Respiratory 140 (14.26%) 78 (15.32%) 84 (16.50%)
Pain 94 (9.57%) 49 (9.63%) 69 (13.56%)
Stress 52 (5.30%) 19 (3.73%) 16 (3.14%)
Cardiovascular 13 (1.32%) 13 (2.55%) 16 (3.14%)
Cancer 47 (4.79%) 30 (5.89%) 31 (6.09%)
Other health effects 90 (9.16%) 49 (9.63%) 53 (10.41%)
Baseline-Stance
towards E-Cigarettes
0.001*
Pro-E-Cigarettes 747 (76.07%) 98 (19.25%) 135 (26.52%)
Anti-E-Cigarettes 98 (9.98%) 311 (61.10%) 73 (14.34%)
Neutral-E-Cigarettes 137 (13.95%) 100 (19.65%) 301 (59.14%)
Mentions of marijuana
0.18 Yes 77 (7.84%) 28 (5.50%) 31 (6.09%)
No 905 (92.16%) 481 (94.50%) 478 (93.91%)
*p<0.05
33
Adjusted multinomial regression analysis revealed that Pain was significantly less likely to be
pro-e-cigarette (RRR = 0.52, p=0.003) compared to Mental Health. Please see Table 4 for more details.
Table 4. Adjusted multinomial logistic regression estimates (Hypothesis 1)
Pro-E-Cigarettes
(n=390)
Anti-E-Cigarettes
(n=146)
RRR CI p-value RRR CI p-value
Categories of health effects^
Stress 0.98 0.50-1.91 0.96 0.90 0.39-2.12 0.82
Pain 0.52 0.33-0.80 0.003* 0.65 0.37-1.14 0.13
Baseline-Stance towards E-
Cigarettes^^
Pro-E-Cigarettes 8.85 5.65-
13.87
0.001* 1.22 0.63-2.35 0.56
Anti-E-Cigarettes 3.41 1.73-6.74 0.001* 19.12 9.77-37.41 0.001
Mentions of marijuana^^^
Yes 1.88 0.87-4.03 0.11 1.06 0.38-2.99 0.11
*p<0.05
^ Reference category: Mental health effects
^^ Reference category: Neutral-E-Cigarettes
^^^ Reference category: No
34
Adjusted multinomial regression analysis revealed Cancer, Cardiovascular, Neurological, or
Other health effects were less likely to be associated with an Anti-Stance than Respiratory health effects
but not significantly so. Please see Table 5 for details.
Table 5. Adjusted multinomial logistic regression estimates (Hypothesis 2)
Pro-E-Cigarettes
(n=592)
Anti-E-Cigarettes
(n=363)
RRR CI p-value RRR CI p-value
Categories of health
effects^
Cancer 0.66 0.36-1.22 0.19 0.86 0.45-1.64 0.64
Cardiovascular 0.45 0.18-1.10 0.08 1.00 0.42-2.40 0.98
Neurological 1.00 0.69-1.47 0.98 1.06 0.70-1.61 0.76
Other 0.92 0.55-1.50 0.72 0.97 0.56-1.68 0.92
Baseline-Stance
towards E-
Cigarettes^^
Pro-E-Cigarettes 14.39 10.11-
20.48
0.001* 2.94 1.94-4.47 0.001*
Anti-E-Cigarettes 2.80 1.81-4.35 0.001* 11.04 7.39-16.49 0.001*
Mentions of
marijuana^^^
Yes 1.08 0.56-2.10 0.24 1.18 0.58-2.40 0.47
*p<0.05
^ Reference category: Respiratory health effects
^^ Reference category: Neutral-E-Cigarettes
^^^ Reference category: No
Discussion
The prevalence of posts to Twitter with a Pro-Stance was predominant in the sample. Pro-Stance
posts predominantly mentioned Neurological effects followed by Mental Health, and Respiratory.
Contrary to our hypothesis, posts about Pain were significantly less likely to be pro-e-cigarette compared
to Mental Health. Cancer, Cardiovascular, Neurological, or Other health effects were not significantly
35
less likely to be associated with an anti-e-cigarette stance compared to Respiratory. One of the strengths
of this research lies in demonstrating a unique approach to understanding health-related attitudes by
conceptualizing meaningful constructs from text-based data while considering their temporal order. Taken
all together, findings characterize health effects and subsequent stance toward e-cigarettes to inform for
future public awareness interventions targeting discussions about health effects of e-cigarettes that are
typically predominantly pro-e-cigarette.
Findings may warrant consideration in developing strategies for real-time social media health
communication campaigns. Recent efforts include health communication Twitter bots (automated social
media accounts that post content and interact with other accounts) that identify accounts associated with
pro-tobacco posts and intervene with personalized tobacco prevention messages(Deb et al., 2018).
Campaigns with automated components can target Twitter accounts discussing potential health effects of
e-cigarettes and associated with a pro-stance toward e-cigarettes, to spread about potential health risks in
that context. For example, such bots could address discussions about improvement in Respiratory
functioning from vaping, by stating potential health risks in that context (e.g., lung inflammation).
Additionally, such efforts can also leverage categories associated with a negative stance to curb e-
cigarette use.
Pro-tobacco messages have been found to outnumber anti-tobacco messages on social media
(Allem, Ramanujam, et al., 2017; Huang, Soto, Fujimoto, & Valente, 2014; Huang, Kornfield, Szczypka,
& Emery, 2014). While paid tobacco advertisements are prohibited on Facebook, Twitter, Instagram and
Google, social media users can post and/or engage with pro-e-cigarette content in the form of news
articles, discussion forums, and organic (non-paid) posts about e-cigarette use from network peers.
Vulnerable populations including adolescents engaging with tobacco-related content online are more
likely to initiate tobacco use and offer low support for tobacco regulations (Majmundar, Chou, Cruz, &
Unger, 2018; Soneji, Pierce, et al., 2017; Unger et al., 2018). Results from this study support previous
findings and add to the literature by further characterizing such e-cigarette-related content in terms of
36
categories of health effects. Neurological health effects were commonly mentioned among accounts
associated with pro- and anti-e-cigarette stance. Past work suggests that adolescents are not always aware
of the extent of nicotine present in e-liquids (Alexander, Williams, & Lee, 2019), and use e-cigarettes for
reasons attributable to nicotine-related effects, including feeling awake and being able to focus (Sidani et
al., 2019), which may be associated with a pro-e-cigarette stance. Findings from this study may warrant
consideration in future health communication campaigns demystifying misperceptions about e-cigarette-
related health effects.
Several other factors may shape overall stance towards e-cigarettes. Public misperceptions about
the relative and absolute health risks of Electronic Nicotine Delivery Systems (ENDS) or e-cigarettes may
contribute to overall positive stance toward vaping (Huang et al., 2019; Majeed et al., 2017). Exposure to
pro-vaping media content may also normalize pro-vaping attitudes (Krauss et al., 2015). Evolving
patterns of vaping may also create a favorable stance towards vaping given those who vape marijuana
perceive it as a healthier and convenient option compared to smoking marijuana (Lee, Crosier,
Borodovsky, Sargent, & Budney, 2016). Examples of such behaviors include hookah pens that deliver
aerosolized flavored aldehydes, with or without nicotine; flavored little cigars now available in electronic
versions; open-system pod mods that allow users to aerosolize liquid THC and nicotine (Budney et al.,
2015; Kostygina et al., 2016; McDonald et al., 2016). Association of vaping (e-cigarette use) with
respiratory, cardiovascular and other chronic diseases such as cancer (Callahan-Lyon, 2014; Grana et al.,
2014; Stratton et al., 2018), may also create a negative stance toward vaping. Device features enhancing
product appeal such as vaporization of cannabis and disposable pod mods may drive a positive stance
toward e-cigarette despite adverse health effects. E-liquid characteristics such as flavors or PG/VG
content which determines the volume of vapor clouds, may also be other potential factors. Severity or
intensity of health effects may also play an important role in shaping subsequent stance toward vaping.
Future efforts may explore ways in which perceived health effects interact with product features, vaping
behaviors, and risk perceptions to shape subsequent stance toward vaping.
37
The unit of analysis in this study was individual Twitter user account. It is possible that accounts
associated with health advocates, public health or healthcare organizations in an individual’s network,
posting about health-effects of e-cigarettes, may be important antecedents to consider while examining
the relationship between categories of health effects posted by individuals and their subsequent stance
toward vaping. Past work suggests that posts originating from accounts owned by professional sources
such as healthcare organizations or government agencies, are shared the most and treated positively in a
health context(Love, Himelboim, Holton, & Stewart, 2013). Future studies can examine ways in which
product characteristics such as e-cigarette device architecture, heating mechanisms may introduce or
enhance appeal and abuse liability and warrant additional review as part of FDA’s protocols.
Limitations
Data selection procedures could have introduced sampling bias as accounts making posts
unrelated to health-effects of e-cigarettes were excluded. Accounts posting one or fewer times post about
e-cigarettes were also excluded from this study. As a result, findings may not be generalizable to all
Twitter users. Findings may not apply to conversation from other years outside the sampling period.
Findings may not generalize to other social media platforms.
Conclusion
This research leverages text-based Twitter data to understand health-related attitudes.
Neurological was the most commonly mentioned category associated with Pro-Stance toward E-
Cigarettes. A majority of user accounts classified as pro-e-cigarette, anti-e-cigarette, or neutral-e-cigarette
did not mention marijuana in their posts. Contrary to our hypothesis, Pain was significantly less likely to
be associated with a pro-e-cigarette stance compared to Mental Health Effects. Cancer, Cardiovascular,
Neurological, or Other health effects were not significantly less likely to be associated with an anti-e-
cigarette stance compared to Respiratory. Findings will inform future health communication campaigns.
38
Chapter 3: Twitter Surveillance at the Intersection of the
Triangulum
Abstract
Objective: To describe key topics of discussions related to each intersection of the Triangulum, a
framework depicting the intersections of e-cigarette, combustible tobacco, and marijuana-related
discussions, using publicly available data from Twitter.
Method: Twitter posts containing marijuana, electronic cigarette and combustible tobacco terms were
collected from January 1, 2018, to December 23, 2019. Posts to Twitter with co-occurring mentions of
keywords associated with the Triangulum was defined as an intersection (e-cigarettes and combustible
tobacco, combustible tobacco and marijuana, electronic cigarettes, and marijuana, electronic cigarettes
and combustible tobacco). For each intersection of the Triangulum, text classifiers and unsupervised
machine learning was used to identify topics in posts (n=999,447).
Results: Person tagging (Twitter accounts tagging other accounts in posts) was the most predominant
topic across all intersections. Discussions about Illicit products in the marketplace was one of the key
topics at in the intersection of e-cigarettes, combustible tobacco, and marijuana, Blunts and Cigars at the
intersection of combustible tobacco and marijuana, Product features among e-cigarette and marijuana-
related conversations, and Flavors among e-cigarette and combustible-related discussions.
Conclusion: By examining intersections of marijuana and other tobacco products, this study offers inputs
for designing comprehensive FDA regulations including regulating product features associated with
appeal and improving enforcement to curb sales of illicit products. Future public health interventions for
tobacco products may warrant including leveraging discussions about undesirable aspects of tobacco use
such as undesirable smell of smoking and adverse health outcomes and addressing prevailing
misinformation. This study also demonstrates the utility of Twitter data for surveillance of complex and
evolving health behaviors.
39
Introduction
Rising e-cigarette use in the US (Fadus, Smith, & Squeglia, 2019; US Department of Health
Human Services, 2016) is accompanied by the proliferation of new products that may enhance appeal and
abuse liability (T.R.D.R.P., 2016). Today, e-cigarette use has rapidly evolved to have the ability to deliver
a variety of substances to the user, such as cannabis (term inclusive of marijuana, THC and other related
terms in this paper) and nicotine. Examples of such devices include hookah pens that deliver aerosolized
flavored aldehydes, with or without nicotine; flavored little cigars now available in electronic versions;
open-system pod mods that allow users to aerosolize liquid THC and e-liquids (Budney et al., 2015;
Kostygina et al., 2016; McDonald et al., 2016). E-cigarette use with substances other than tobacco, at
varying levels of nicotine, potentially complicate the valid assessment of whether e-cigarette devices
contribute to susceptibility to smoking and increased harm to public health.
E-cigarette device use, in combination with the ongoing marijuana legislative reform (Stiles,
2019; Wenk, 2019), and proposed national and state tobacco regulations, have implications for changing
patterns of tobacco behaviors and product appeal (Baggio et al., 2014; Dutta-Bergman, 2004; Kostygina
et al., 2016; Majmundar, Kirkpatrick, et al., 2019). Use of two or more products or substances separately
within a specific time period or at the same time or in the same product, are emerging substance use
behaviors that may contribute to substantial health risks (Schauer, Rosenberry, & Peters, 2017). Dual-use
(Budney et al., 2015; El-Toukhy & Choi, 2016; Evans-Polce, Lanza, & Maggs, 2016; Schauer & Peters,
2018; Unger, Soto, & Leventhal, 2016), and poly use of marijuana and nicotine products (El-Toukhy &
Choi, 2016; Haardörfer et al., 2016; Huh & Leventhal, 2016) raise several concerns related to abuse
liability, earlier age of smoking initiation, and addiction among vulnerable populations. Prior work
demonstrates that positive perceptions and peer approval is associated with marijuana and e-cigarette use
(Berg et al., 2015; Frohe et al., 2018). Particularly notable is the high appeal associated with co-use of
these substances. Evidence suggests that novel ways of administering marijuana and nicotine is
particularly appealing among youth (Eggers et al., 2017). Long-term health risks of evolving patterns of
40
co-use behaviors are inconclusive (Budney et al., 2015; Rabin & George, 2015). Public health
surveillance in this area can offer crucial points of departure for investigating emerging public experience.
Researchers have noted the need for a more holistic approach to understand the changing
landscape of marijuana, combustible tobacco and e-cigarettes (McDonald et al., 2016). The Triangulum
(Latin for triangle) offers a useful framework to investigate the intersection of e-cigarettes, marijuana and
combustible tobacco (T.R.D.R.P., 2016). This framework is directly relevant to public health surveillance
(e.g., poly and dual use) (Chu, Majmundar, et al., 2019), policy (e.g., implications of legalization of
marijuana, smoke-free policies) (Henriksen, Schleicher, Ababseh, Johnson, & Fortmann, 2018), treatment
(e.g., implications of dual or poly use on cessation) (Rigotti et al., 2018), and education (e.g., need for
health communication countering pro-marijuana media content) (Napper, Froidevaux, & LaBrie, 2016).
Social media surveillance offers opportunities to investigate emerging tobacco-use behaviors. For
example, Allem and colleagues examined early public conversations around Juul and found that Juul’s
discreetness may have facilitated stealth vaping in locations such as school classrooms (Allem,
Dharmapuri, Unger, et al., 2018). Majmundar and colleagues identified commonly referenced locations in
vaping-related discussions on Twitter (Majmundar, Allem, Cruz, & Unger, 2019). Ayers and colleagues
also used social media surveillance to highlight key reasons for vaping cited in online discourse, which
included favorable social image, and flavors (Ayers et al., 2017). Another study revealed that vaping
discussions predominantly involved references to dual and poly substance use (Allem, Escobedo, Chu,
Boley Cruz, & Unger, 2017).
This study examines key topics of discussions related to each intersection of the Triangulum (e-
cigarettes AND combustible tobacco, combustible tobacco AND marijuana, e-cigarettes AND marijuana)
using Twitter data. The central hypothesis of the study is that appeal (i.e., the attractive and rewarding
product characteristics such as taste and appearance that enhance product liking and encourage use
(Henningfield, Hatsukami, Zeller, & Peters, 2011) will be the predominant topic of conversation among
posts with co-occurring mentions of e-cigarettes, tobacco and marijuana. Appeal, in this context, is
41
defined as attractive and rewarding product characteristics (e.g., taste, appearance) that enhance product
liking and encourage use (Henningfield et al., 2011). Findings from this study should inform FDA’s
overall health risks assessment of tobacco products and prevention efforts.
Methods
Twitter (https://Twitter.com/) posts containing tobacco- and marijuana-related terms were
obtained from January 1, 2018 to December 23, 2019 using Twitter’s Streaming Application Program
Interface (API). The initial corpus consisted of 154,274,919 posts to Twitter. For a complete list of
keywords, see Table 6. Retweets, non-English Twitter posts, Twitter posts from bot accounts (Allem &
Ferrara, 2016), duplicate posts that were not retweets, spam posts and promotional posts were excluded
from this sample. This resulted in an initial sample of 999,447 posts to Twitter. Next, the text in the
sample was prepared for analysis based on basic normalization (e.g., lower case all text, remove special
characters), stop word removal (e.g., words such as “an,” “the,” etc.), normalization of Twitter user
mentions (e.g., “@jimjohn” is converted to “@person”), lemmatization (e.g. “vaper”, “vaper’s” and
“vapers’” are all converted to “vaper”), and non-printable character removal (e.g., emojis and symbols
from other languages).
Table 6. List of tobacco, marijuana, and e-cigarette search keywords
Combustible cigarette-
related keywords
Marijuana-related
keywords
E-cigarette-related
keywords
'cigarette', 'cigarettes',
'marlboro', 'pall mall',
'pallmall', 'cigarillo',
'cigarillos', 'cigar', 'cigars',
'swisher', 'camel crush bold',
'camelcrushbold'
'bong', 'budder', 'cannabis' ,
'cbd' , 'ganja' , 'hash' , 'hemp'
, 'indica', 'kush' , 'marijuana' ,
'marihuana' , 'reefer' , 'sativa'
, 'thc' , 'weed', 'blunt',
'blunts'
'ecig', 'ecigs', 'ecigarette',
'ecigarettes', 'e-cigarette', 'e-
cigarettes', 'e-liquid', 'e-
liquids', 'eliquid', 'eliquids',
'ejuice', 'ejuices', 'e-juice', 'e-
juices', 'vaping', 'vapes',
'vape', 'vaper', 'Juul', 'juuling'
The above initial sample was divided into four analytic subsets. The sample size for each subsets
was: (a) combustible tobacco AND e-cigarette subset with co-occurring mentions of combustible
tobacco-related and e-cigarette-related keywords (N= 456,178; 2018 = 200,914, 2019 = 255,264), (b)
42
combustible tobacco AND marijuana subset with co-occurring mentions of combustible tobacco- and
marijuana-related keywords (N = 188,562; 2018 = 97,678; 2019 = 90,884), (c) e-cigarette AND
marijuana subset with co-occurring mentions of e-cigarette-related and marijuana-related keywords (N =
334,409; 2018 = 121,904; 2019 = 212,505), and (d) combustible tobacco-marijuana- e-cigarette subset
with co-occurring mentions of combustible tobacco-, e-cigarette- and marijuana-related keywords (N =
20,298; 2018 = 6,097; 2019 = 14,201).
To identify key topics of conversation, the following procedure was carried out for each of the
four subsets. Tweets were analyzed using word frequencies (of single words (one-grams) and double-
word (bi-gram) combinations). To illustrate one-grams and bi-grams, a tweet ‘friends like vaping’
consists of 3 one-grams (e.g., friends, like, vaping) and 2 bi-grams (e.g., friends-like, like-vaping). Each
subset was also visualized using word clouds for visual inspection of emerging topics. Based on the visual
and word-count assessment, an initial list of topics was identified.
Next, GloVe (Pennington, Socher, & Manning, 2014), an unsupervised learning algorithm
developed by the Stanford NLP group, was used to identify words similar to the one-grams and bi-grams
identified for each topic (e.g., ‘chill’, ‘party’, hangout’ can be classified under one category of
‘socialization’). GloVe offers a pre-trained vector-representations for words. It calculates the Euclidean
distance (or cosine similarity) between two-word vectors to measure the linguistic or semantic similarity
of the corresponding words (e.g., ‘frog’ and ‘toad’). It also calculates the vector difference between the
two-word vectors to capture meaning specified by the juxtaposition of two words. The advantage of such
calculation helps identify ‘man’ and ‘woman’ as similar words (human beings), which may be otherwise
thought of as keywords representing opposite genders. Manual checks of data were conducted to expand
the list of synonyms (e.g., ‘pot’ may be commonly used in the context of marijuana use). Final
classification of posts to one or more topics was implemented by using rule-based classifiers written in
Python to check for the presence of topic-specific bi-grams or one-grams in a tweet. For each subset, the
overlap of key topics was summarized in a matrix. In other words, number of posts applicable to any two
43
topics would be found at the intersection of the matrix for these two topics. The value of each cell
represents the percentage of the subset and number of Twitter posts. The number of final topics differed
by subset. Please see Table 7 for definitions of key topics at each intersection of the Triangulum.
All analyses relied on public, anonymized data, adhered to the terms and conditions, terms of use,
and privacy policies of Twitter, and will be performed under Institutional Review Board approval from
the authors’ university. To protect privacy, no tweets will be reported verbatim.
44
Table 7. Definitions of key topics at each intersection of the Triangulum
Combustible
tobacco
-
Marijuana
Definition
Electronic
cigarette
-
Marijuana
Definition
Electronic
cigarette
-
Combustible
tobacco
Definition
Electronic
cigarette
-
Combustible
tobacco
-
Marijuana
Definition
Person
tagging
@Person
Person
tagging
@Person
Person
tagging
@Person
Person
tagging
@Person
Cigars and
blunts
Mentions
of cigars
and blunts
Product
features
Descriptor
s of
products
(e.g.,
odorless,
purified)
Flavors
Mentions
of flavors
Illicit
Mentions
of products
that are
illegal or
sold
illegally
Drugs and
alcohol
Mentions
of drugs
and
alcohol
Carts
Mentions
of THC
cartridges
and dank
vapes
Health risks
Mentions
of adverse
health
outcomes
and
injuries
Health risks
Mentions
of adverse
health
outcomes
Legal
Mentions
of tobacco
laws and
regulation
s
Illicit
Mentions
of
products
that are
illegal or
sold
illegally
Product bans
Mentions
of vaping
bans or
policies,
restriction
s on sale
of tobacco
products
and
minimum
age to
legally
purchase
tobacco
products
Public
agencies
Mentions
of federal
agencies
such as
FDA, CDC
Smell
Descriptor
s of
tobacco
smell
Promotions
Mentions
of sales,
discounts,
coupon
codes
Underage
Mentions
of
children
and
adolescent
s
Appeal
Descriptors
of appeal
(e.g., love,
delicious)
Quitting
Mentions
of quitting
combustib
le tobacco
New
products
Mentions
of new
products
(e.g.,
marijuana
tinctures
(defined
as plant
extracts
dissolved
in
ethanol)
and
suppleme
nts)
Buy/Sell
Mentions
of product
purchase
and sales
such as
price,
delivery,
refund,
retail and
modes of
payment
including
venmo
Anti-
regulation
discussions
related to
unfavorable
appraisal of
tobacco
policies
and anti-
tobacco
policy
advocacy
Death
Mentions
of death to
Stoner
imagery
Reference
s to
Quitting
Mentions
of
Addiction
Mentions
of craving,
45
communic
ate the
adverse
health
effects
‘staying
high’ and
dabbing
rituals
referred to
as
‘shatterda
y’
cessation
and
switching
to vape
products
and
dependency
Addressing
audience
Generally
addressing
the
Twitter
audience
(e.g.,
‘People’,
‘Y’all’)
Health
risks
Mentions
of adverse
health
outcomes
such as
death and
poisoning
Nicotine
Mentions
of
nicotine
Alternative
medicine
Mentions
of natural,
herbal and
other
forms of
medicine
Health
Claims
Mentions
of
alleviation
or
managem
ent of
health
issues
such as
anxiety
and pain
Epidemic
Mentions
of
epidemic
of e-
cigarette
use
Transit
Mentions
of forms
of
transportat
ion
Flavors
Mentions
of flavors
Addiction
Mentions
of
craving,
and
dependenc
y
Cigar
marketing
Descriptor
s in cigar
marketing
(e.g.,‘ciga
rotica’,
‘girlswith
cigars’)
Disposable
products
Mentions
of
disposable
products
such as
Puff Bar
School
Mentions
of school
and
sections
of schools
such as
classroom
s
New delivery
devices
Mentions
of devices
such as
the volute,
stogie)
Starter kits
Mentions
of e-
cigarette
starter kits
IQOS
Mentions
of the
product
‘IQOS’, a
heat-not-
burn
tobacco
product
46
Results
Topics at the intersection of combustible tobacco and marijuana are illustrated in Figure 6. The
total coverage of the 12 topics identified constituted 73.10% of all tweets in the subset. The remaining
tweets were too varied to be classified into one topic with meaningful coverage (i.e., meaningful coverage
defined as less than 1% of total tweets). Person tagging (31.72%) was the most predominant topic
followed by discussions about Cigars and Blunts (26.44%), references to Drugs and Alcohol (12.25%),
and Legal aspects of marijuana and combustible use was about 10.84% of the sample. Alternative
Medicine (3.33%) was particularly notable. The least common topics included New delivery devices
(1.93%) such as volute, stogie, and waffle pipes, and descriptors used in Cigar Marketing 1.93%) such as
‘girlswithcigars’, and ‘cigarfam’.
47
Figure 6. Topics at the intersection of combustible tobacco and marijuana
Topics at the intersection of vaping and marijuana-related discussions are illustrated in Figure 7.
The total coverage of the 11 topics identified constituted 70.62% of all tweets in the subset. The
remaining tweets were too varied to be classified into one topic with meaningful coverage. Person tagging
(32.32%) was the most predominant topic followed by discussions about Product features (13.19%) and
mentions of Carts (or cartridges) (12.73%). Other topics included mentions of Illicit products (12.02%),
Promotions (10.42%), New products (9.62%), and Stoner imagery (9.52%). Harm perceptions (9.21%),
Health claims (8.95%), and Disposable products were the least predominant in the sample (1.12%).
Person tagging
(55698) /
(31.72%)
Cigars & Blunts
(10651) /
(6.07%)
(46419) /
(26.44%)
Drugs & Alcohol
(10071) /
(5.74%)
(1408) /
(0.80%)
(21513) /
(12.25%)
Legal
(8697) /
(4.95%)
(1456) /
(0.83%)
(5919) /
(3.37%)
(19035) /
(10.84%)
Smell
(2868) /
(1.63%)
(1368) /
(0.78%)
(1161) /
(0.66%)
(819) /
(0.47%)
(12169) /
(6.93%)
Quitting
(2775) /
(1.58%)
(1205) /
(0.69%)
(1128) /
(0.64%)
(605) /
(0.34%)
(425) /
(0.24%)
(8903) /
(5.07%)
Death
(3699) /
(2.11%)
(1198) /
(0.68%)
(2288) /
(1.30%)
(2843) /
(1.62%)
(351) /
(0.20%)
(360) /
(0.21%)
(8874) /
(5.05%)
Addressing
audience
(2607) /
(1.48%)
(1350) /
(0.77%)
(1517) /
(0.86%)
(1249) /
(0.71%)
(706) /
(0.40%)
(642) /
(0.37%)
(1099) /
(0.63%)
(8303) /
(4.73%)
Alternative
medicine
(1773) /
(1.01%)
(846) /
(0.48%)
(972) /
(0.55%)
(1067) /
(0.61%)
(89) /
(0.05%)
(204) /
(0.12%)
(404) /
(0.23%)
(205) /
(0.12%)
(5850) /
(3.33%)
Transit
(966) /
(0.55%)
(778) /
(0.44%)
(344) /
(0.20%)
(290) /
(0.17%)
(806) /
(0.46%)
(126) /
(0.07%)
(139) /
(0.08%)
(155) /
(0.09%)
(35) /
(0.02%)
(3423) /
(1.95%)
Cigar Marketing
(73) /
(0.04%)
(2946) /
(1.68%)
(17) /
(0.01%)
(3) /
(0.00%)
(54) /
(0.03%)
(9) /
(0.01%)
(3) /
(0.00%)
(4) /
(0.00%)
(6) / (0.00%)
(4) /
(0.00%)
(3394) /
(1.93%)
New delivery
Devices
(73) /
(0.04%)
(2946) /
(1.68%)
(17) /
(0.01%)
(3) /
(0.00%)
(54) /
(0.03%)
(9) /
(0.01%)
(3) /
(0.00%)
(4) /
(0.00%)
(6) / (0.00%)
(4) /
(0.00%)
(3394) /
(1.93%)
(3394) /
(1.93%)
(Number) / (%
of total)
Person
tagging
Cigars &
Blunts
Drugs &
Alcohol
Legal Smell Quitting Death
Addressing
audience
Alternative
medicine
Transit
Cigar
Marketing
New
delivery
Devices
Total number of tweets: (128346) \ (73.10%)
48
Figure 7. Topics at the intersection of e-cigarettes and marijuana
Topics at the intersection of e-cigarettes and combustible tobacco are illustrated in Figure 8. The
total coverage of the 13 topics identified constituted 74.98% of all tweets in the subset. The remaining
tweets were too varied to be classified into one topic with meaningful coverage. Person tagging (32.32%)
was the most predominant topic. Flavors were the second-most predominant topic (18.56%) followed by
discussions about recent vaping Bans (17.65%). Health Risks (16.12%), Underage (14.64%), Buy/Sell
(9.29%), Quitting (8.65%), followed by references to Nicotine (7.79%). References to the vaping
Epidemic (5.09%), Addiction (4.76%), School (3.22%), and vape Starter Kits (2.74%). IQOS, was the
least predominant topic (0.66%).
ATPerson
(108092) /
(32.32%)
Product
features
(11566) /
(3.46%)
(44106) /
(13.19%)
Carts
(15001) /
(4.49%)
(9198) /
(2.75%)
(42561) /
(12.73%)
Illicit
(22235) /
(6.65%)
(5785) /
(1.73%)
(8718) /
(2.61%)
(40182) /
(12.02%)
Promotions
(7981) /
(2.39%)
(5335) /
(1.60%)
(6176) /
(1.85%)
(2926) /
(0.87%)
(34843) /
(10.42%)
New
products
(8456) /
(2.53%)
(7805) /
(2.33%)
(2329) /
(0.70%)
(1362) /
(0.41%)
(4350) /
(1.30%)
(32158) /
(9.62%)
Stoner
Imagery
(5854) /
(1.75%)
(6860) /
(2.05%)
(7944) /
(2.38%)
(1241) /
(0.37%)
(7813) /
(2.34%)
(3859) /
(1.15%)
(31823) /
(9.52%)
Harm
perceptions
(15793) /
(4.72%)
(4178) /
(1.25%)
(5766) /
(1.72%)
(12089) /
(3.62%)
(1515) /
(0.45%)
(753) /
(0.23%)
(346) /
(0.10%)
(30804) /
(9.21%)
Health
Claims
(7158) /
(2.14%)
(10112) /
(3.02%)
(1690) /
(0.51%)
(892) /
(0.27%)
(3841) /
(1.15%)
(10047) /
(3.00%)
(4449) /
(1.33%)
(1315) /
(0.39%)
(29941) /
(8.95%)
Flavors
(10573) /
(3.16%)
(4928) /
(1.47%)
(4327) /
(1.29%)
(5751) /
(1.72%)
(3363) /
(1.01%)
(4026) /
(1.20%)
(1929) /
(0.58%)
(3263) /
(0.98%)
(2028) /
(0.61%)
(27739) /
(8.29%)
Disposables
(1052) /
(0.31%)
(802) /
(0.24%)
(636) /
(0.19%)
(195) /
(0.06%)
(479) /
(0.14%)
(235) /
(0.07%)
(316) /
(0.09%)
(42) /
(0.01%)
(219) /
(0.07%)
(550) /
(0.16%)
(3757) /
(1.12%)
(Number) /
(% of total)
Person
tagging
Product
features
Carts Illicit Promotions
New
products
Stoner
Imagery
Harm
perceptions
Health
Claims
Flavors Disposables
Total number of tweets: (236147) \ (70.62%)
49
Figure 8. Topics at the intersection of e-cigarette and combustible tobacco
Topics at the intersection of e-cigarettes, marijuana, and combustible tobacco are illustrated in
Figure 9. The total coverage of the 7 topics identified constituted 67.08% of all tweets in the corpus.
Similar to the above intersections, Person tagging was the most predominant topic (42.16%), followed by
discussions about Illicit products on the market (23.39%), Health Risks (18.12%), references to Public
Agencies (9.46%), Appeal (9.12%), Anti-regulation (6.64%), and references to Addiction (2.74%).
Person
tagging
(142230) /
(31.18%)
Flavors
(31138) /
(6.83%)
(84652) /
(18.56%)
Ban
(24173) /
(5.30%)
(27151) /
(5.95%)
(80501) /
(17.65%)
Health
Risks
(23997) /
(5.26%)
(13433) /
(2.94%)
(11245) /
(2.47%)
(73531) /
(16.12%)
Underage
(23462) /
(5.14%)
(19338) /
(4.24%)
(17811) /
(3.90%)
(10244) /
(2.25%)
(66770) /
(14.64%)
Buy/Sell
(11698) /
(2.56%)
(13851) /
(3.04%)
(13749) /
(3.01%)
(4174) /
(0.91%)
(8273) /
(1.81%)
(42362) /
(9.29%)
Quitting
(16375) /
(3.59%)
(7600) /
(1.67%)
(3501) /
(0.77%)
(6162) /
(1.35%)
(4468) /
(0.98%)
(2480) /
(0.54%)
(39472) /
(8.65%)
Nicotine
(14618) /
(3.20%)
(10073) /
(2.21%)
(4152) /
(0.91%)
(6518) /
(1.43%)
(7697) /
(1.69%)
(2473) /
(0.54%)
(5502) /
(1.21%)
(35527) /
(7.79%)
Epidemic
(7639) /
(1.67%)
(5593) /
(1.23%)
(7484) /
(1.64%)
(5565) /
(1.22%)
(12459) /
(2.73%)
(2372) /
(0.52%)
(916) /
(0.20%)
(2090) /
(0.46%)
(23217) /
(5.09%)
Addiction
(8138) /
(1.78%)
(5267) /
(1.15%)
(3080) /
(0.68%)
(4044) /
(0.89%)
(7696) /
(1.69%)
(1546) /
(0.34%)
(3167) /
(0.69%)
(9272) /
(2.03%)
(1647) /
(0.36%)
(21704) /
(4.76%)
School
(4563) /
(1.00%)
(2710) /
(0.59%)
(2039) /
(0.45%)
(2333) /
(0.51%)
(4990) /
(1.09%)
(769) /
(0.17%)
(501) /
(0.11%)
(1167) /
(0.26%)
(2105) /
(0.46%)
(1176) /
(0.26%)
(14673) /
(3.22%)
Starter kits
(1255) /
(0.28%)
(304) /
(0.07%)
(25) /
(0.01%)
(82) /
(0.02%)
(29) /
(0.01%)
(526) /
(0.12%)
(215) /
(0.05%)
(109) /
(0.02%)
(18) /
(0.00%)
(24) /
(0.01%)
(4) /
(0.00%)
(12502) /
(2.74%)
IQOS
(980) /
(0.21%)
(1249) /
(0.27%)
(345) /
(0.08%)
(572) /
(0.13%)
(97) /
(0.02%)
(253) /
(0.06%)
(199) /
(0.04%)
(258) /
(0.06%)
(51) /
(0.01%)
(52) /
(0.01%)
(19) /
(0.00%)
(62) /
(0.01%)
(2997) /
(0.66%)
(Number) /
(% of total)
Person
tagging
Flavors Ban
Health
Risks
Underage Buy/Sell Quitting Nicotine Epidemic Addiction School Starter kits IQOS
Total number of tweets: (342035) \ (74.98%)
50
Figure 9. Topics at the intersection of e-cigarettes, marijuana, and combustible tobacco
Table 8 presents a summary of topics for each intersection of the Triangulum. Person tagging was
the most predominant topic for all intersections. Flavors were one of the most predominant topics of
discussion at the intersection of e-cigarette and combustible tobacco, and one of the least predominant at
the intersection of e-cigarette and marijuana. Health risks were common topics at all intersections except
for combustible tobacco and marijuana. Addiction was common to e-cigarette and combustible tobacco,
and e-cigarette, combustible tobacco and marijuana-related conversations.
Person
tagging
(8558) /
(42.16%)
Illicit
(3091) /
(15.23%)
(4747) /
(23.39%)
Health Risks
(2002) /
(9.86%)
(1468) /
(7.23%)
(3677) /
(18.12%)
Pubic agencies
(1073) /
(5.29%)
(870) /
(4.29%)
(618) /
(3.04%)
(1920) /
(9.46%)
Appeal
(688) /
(3.39%)
(410) /
(2.02%)
(359) /
(1.77%)
(265) /
(1.31%)
(1852) /
(9.12%)
Anti-
regulation
(593) /
(2.92%)
(365) /
(1.80%)
(152) /
(0.75%)
(94) /
(0.46%)
(35) /
(0.17%)
(1348) /
(6.64%)
Addiction
(367) /
(1.81%)
(216) /
(1.06%)
(238) /
(1.17%)
(174) /
(0.86%)
(170) /
(0.84%)
(7) /
(0.03%)
(557) /
(2.74%)
(Number) /
(% of total)
Person
tagging
Illicit Health Risks
Pubic
agencies
Appeal
Anti-
regulation
Addiction
Total number of tweets: (13615) \ (67.08%)
51
Table 8. Predominant topics at each intersection of the Triangulum
Combustible
tobacco
-
Marijuana
N
(%)
Electronic
cigarette
-
Marijuana
N
(%)
Electronic
cigarette
-
Combustible
tobacco
N
(%)
Electronic
cigarette
-
Combustible
tobacco
-
Marijuana
N
(%)
Person
tagging
55698
(31.72%)
Person
tagging
108092
(32.32%)
Person
tagging
142230
(31.18%)
Person
tagging
8558
(42.16%)
Cigars and
blunts
46419
(26.44%
Product
features
44106
(13.19%)
Flavors
84652
(18.56%)
Illicit
4747
(23.39%)
Drugs and
alcohol
21513
(12.25%)
Carts
42561
(12.73%)
Health risks
73531
(16.21%)
Health risks
3677
(18.12%)
Legal
19035
(10.84%)
Illicit
40182
(12.02%)
Product bans
80501
(17.65%)
Public
agencies
1920
(9.46%)
Smell
12169
(6.93%)
Promotions
34843
(10.42%)
Underage
73531
(16.21%)
Appeal
1852
(9.12%)
Quitting
8903
(5.07%)
New
products
32158
(9.62%)
Buy/Sell
42362
(9.29%)
Anti-
regulation
1348
(6.64%)
Death
8874
(5.05%)
Stoner
imagery
31823
(9.52%)
Quitting
39472
(8.65%)
Addiction
557
(2.74%)
Addressing
audience
8303
(4.73%)
Health
risks
30804
(9.21%)
Nicotine
35527
(7.79%)
Alternative
medicine
5850
(3.33%
Health
Claims
29941
(8.95%)
Epidemic
23217
(5.09%)
Transit
3423
(1.95%)
Flavors
27739
(8.29%)
Addiction
21704
(4.76%)
Cigar
marketing
3394
(1.93%)
Disposable
products
3757
(1.12%)
School
14673
(3.22%)
New
delivery
devices
3394
(1.93%)
Starter kits
12502
(2.74%)
IQOS
2997
(0.66%)
128346
(73.10%)
236147
(70.62%)
342035
(74.98%)
13615
(67.08%)
Discussion
This study is one of the largest Twitter studies focused on distinct aspects of the Triangulum,
describing 999,447 Twitter posts over a span of two years. A number of different topics were identified at
each intersection of the Triangulum, ranging from Flavors, Nicotine, Health claims, New products, to
Anti-Regulation. Person tagging was the most predominant topic across all intersections, while
discussions about Illicit products was one of the key topics at in the intersection of e-cigarettes,
combustible tobacco, and marijuana, Blunts and Cigars at the intersection of combustible tobacco and
marijuana, Product features among e-cigarette and marijuana-related conversations, and Flavors among e-
52
cigarette and combustible-related discussions. A number of other topics were also common across at least
two of the four intersections, including Flavors, Addiction and Health risks.
Person tagging is a common theme in tobacco-related conversations on social media (Allem,
Dharmapuri, Leventhal, Unger, & Cruz, 2018; Allem, Dharmapuri, Unger, et al., 2018). As discussed in
prior work (Allem, Escobedo, & Dharmapuri, 2020), person tagging in online messages involves others
into conversations about attitudes and behaviors about topics such as tobacco and marijuana. Such
involvement may have implications for tobacco and marijuana use. Real-time health communication
interventions can identify Twitter users posting about tobacco by automated methods and deliver tailored
messages pertaining to relevant intersections of the Triangulum on social media platforms. Such nuanced
understanding of key topics of conversations at each intersection has the potential to inform more tailored
health communication campaigns and public health interventions. Recent efforts in this area show
promising potential of machine learning-based interventions on social media. Examples of such efforts
include, use of recommendation algorithms to deliver tailored health messages (Kim et al., 2019), and
real-time, targeted delivery of tailored health messages (Deb et al., 2018). Such interventions can enhance
public awareness about the unique health risks associated with relevant intersections of the Triangulum.
Future work may investigate the extent to which Person-tagged social media conversations pertain to
questions, interest, or knowledge-sharing about products at the intersection of the Triangulum.
Cigars and blunts was the predominant topic of conversation related to combustible tobacco and
marijuana. This finding is in line with previous evidence, highlighting correlates of cigar use with
marijuana (Cohn, Johnson, Ehlke, & Villanti, 2016). Blunt use, defined as hollowing out a cigar to refill it
with marijuana for smoking, exposes individuals to nicotine, that is present in the blunt wrapper
(Fairman & Anthony, 2017). To address this trend, past work has highlighted the need for coordinated
tobacco control and substance abuse prevention efforts (Soldz, Huyser, & Dorsey, 2003), including
making it more difficult to modify cigar wrappers for blunt use. Alcohol and Drugs was another common
topic. Past work suggests emerging trends in alcohol and blunt use (Montgomery, Zapolski, Banks, &
53
Floyd, 2019). Future tobacco prevention interventions addressing the use of other substances with tobacco
products, can limit the progression of vulnerable populations to more addictive or harmful health
behaviors.
From a health communication perspective, leveraging discussions about some of the existing
topics found in this study - the undesirable Smell of products, and adverse health outcomes such as Death,
may play a crucial role in changing norms about cigars, the third most popular tobacco product among
adolescents (Jamal et al., 2017). Additionally, communication efforts addressing cessation attempts
(Majmundar, Cerrada, Fang, & Huh, 2020) and potential misinformation related to beneficial health
effects, referred to as Alternative Medicine, from tobacco and marijuana use, are also warranted to protect
vulnerable populations.
Product features was one of the most predominant topics at the intersection of e-cigarettes and
marijuana-related public conversations. Past work surveilling Twitter conversations about vaping support
this finding (Basáñez et al., 2019; Basanez, Majmundar, Cruz, & Unger, 2018). These studies suggest that
vaping is marketed as a health-enhancing behavior (Basanez et al., 2018) capable of delivering vitamins
while vape products are in use (Basáñez et al., 2019). Current findings add to the literature by
highlighting that such conversations are particularly predominant at the specific intersection of e-
cigarettes and marijuana-related conversations on Twitter. Exposure to messages about product features
highlighting health benefits may create favorable social norms about vaping marijuana. Studies
examining downstream effects of exposure to such features such as attitudinal ambivalence are needed to
assess their impact and appeal, and research on potential messaging strategies involving use of two-sided
messages to counter such effects may also be explored (Cornelis, Heuvinck, & Majmundar, 2020).
Regulatory efforts may consider regulating product features associated with appeal. Discussions about
vape Cartridges and mentions of illicit substances are situated in the context of expanding illicit products,
marijuana legalization, widespread marijuana use, and availability of new generation products such as
open-system pod mods (Galstyan et al., 2018; King, Jones, Baldwin, & Briss, 2020). Findings highlight
54
the need for public awareness initiatives to limit the adverse health effects of inferior quality or
adulterated products. Subjecting marijuana-containing products that can be vaped, to safety and quality
testing and appropriate labelling may also address these concerns (Smith & Goniewicz, 2020).
Flavors, one of the predominant topics at the intersection of e-cigarettes and combustible tobacco,
has been previously linked to e-cigarette-use initiation, continued use, greater satisfaction, greater
perceived addiction, and higher nicotine consumption compared to those that vape non-flavored products,
especially among young adults (Cwalina, Majmundar, Unger, Barrington-Trimis, & Pentz, 2020; Landry
et al., 2019; St Helen, Dempsey, Havel, Jacob, & Benowitz, 2017). The recent introduction of state-level
flavor bans that now include certain types of e-cigarettes may be crucial in curbing the appeal of these
products among younger populations. Other predominant topics including Health risks and recent Product
bans and legal restrictions on sale of tobacco products and introduction of a federal minimum age to
legally purchase tobacco products, highlight the changing regulatory landscape. It is possible to conduct
large-scale pre- and post-surveillance of public sentiment toward regulatory changes using social media
data to inform future regulatory communication and outreach efforts of the FDA.
Conversations mentioning e-cigarettes, combustible tobacco and marijuana mostly centered
around topics highlighting Illicit products, Health risks and Addiction. Future education campaigns may
consider informing the public about the scientific facts related to the above topics to address this
prevailing misinformation. Enhanced regulatory enforcement efforts to regulate sales of unauthorized
products are also warranted. On the other side of the spectrum were product Appeal and Anti- tobacco
regulation-related topics that may create favorable norms for initiation or continued use of tobacco and/or
marijuana. Insights from social media surveillance, in this context, have highlighted ways in which
evolving tobacco-use behaviors drive appeal. For instance, YouTube videos are predominantly pro-
marijuana-vaping and pro-vaping (Yang et al., 2018), devices such as open-system pod mods are
promoted extensively on Instagram (Majmundar, Kirkpatrick, et al., 2019), and individuals engage
extensively in sharing their product use experiences by sharing pictures of vaporizers and non-traditional
55
forms of administering marijuana (Cavazos-Rehg, Krauss, Sowles, & Bierut, 2016). While regulation of
marijuana is beyond the purview of the FDA, findings may warrant consideration as part of FDA’s
product review and standard procedures for e-cigarette and combustible tobacco products that can be co-
used with marijuana.
Limitations
This study drew data from Twitter and findings may not generalize to other social media
platforms. Findings may also not represent data from individuals with private Twitter accounts. The time
range of the data is January 2018 to December 2019, and findings may not generalize to other years. Data
captured in this research relied specified keywords of interest (marijuana, combustible tobacco, e-
cigarettes). While every effort was made to create a comprehensive set of keywords for data collection,
data may not be completely representative of the entire area of interest. Currently, there are regional
variations in tobacco and marijuana regulations. States such as California support tobacco restrictive but
lax marijuana policies compared to other states. Other states may not have as restrictive tobacco
regulations by, for instance, allowing e-cigarette use in public spaces and also support legalization of
marijuana use. Such variations may influence the prevalence of themes on Twitter and findings may not
be representative of such state-wide differences.
Conclusion
Person tagging, discussions about illicit products in the market was one of the key topics at in the
intersection of e-cigarettes, combustible tobacco, and marijuana, Blunts and Cigars at the intersection of
combustible tobacco and marijuana, Product features among e-cigarette and marijuana-related
conversations, and Flavors among e-cigarette and combustible-related discussions. By examining
intersections of marijuana and other tobacco products, this study offers inputs for designing
comprehensive FDA regulations and public health interventions for tobacco products. This study also
demonstrates the utility of Twitter data for surveillance of complex and evolving health behaviors.
56
Concluding Remarks
The primary goal of this dissertation is to characterize and examine relationships between
commonly referenced health effects of e-cigarettes, subsequent stance toward e-cigarettes, and evolving
e-cigarette-use behaviors by leveraging two years of organic, public conversations on Twitter.
Study 1 findings indicate that references to Neurological, Death, Mental Health, and Respiratory,
were some of the most predominant categories in the past two years. Compared to 2018, Death, Mental
Health, and Respiratory health effects were more prevalent in 2019. While Death, Mental Health, and
Neurological health effects were one of the most predominant categories during 2018 and 2019, the topic
of respiratory health effects emerged as part of the top five categories in 2019. Future campaigns can
enhance awareness of adverse outcomes related to the predominant categories of health effects: Death,
Mental Health, and Neurological. Future health communication campaigns can enhance awareness of
adverse outcomes related to the predominant categories of health effects: Death, Mental Health, and
Neurological. This study may also inform specific categories of health effects that may be addressed in a
patient-provider setting. Categories of health effects such as Cancer, Cardiovascular diseases that were
not as predominant may also warrant adequate consideration, while discussing the potential long-term
health risks of vaping. Findings from Study 1 also offer insights on commonly discussed and/or
experienced health effects of e-cigarettes that may be addressed in the PMTA applications.
Study 2 findings highlight that the majority of posts were classified as Pro-Stance toward E-
Cigarettes followed by Anti-Stance, and Neutral-Stance. Neurological was commonly associated with a
Pro-Stance followed Mental Health, and Respiratory. Contrary to the hypothesis, Pain was significantly
less likely to be pro-e-cigarette compared Mental Health Effects. This study describes the varying
positions toward e-cigarette use as they relate to perceived health effects from use. While some e-cigarette
users discuss a health issue in a negative frame towards e-cigarettes, these attitudes were not universal.
Interventions could be designed around discussions of posts to Twitter however findings may only be
57
limited to those with negative stances. Further research is needed to identify ways to communicate the
health risks of e-cigarette use.
Study 3 explored key topics of discussion at the intersection of the Triangulum. Person tagging
was the most predominant topic across all intersections. Discussions about Illicit products in the
marketplace was one of the key topics at in the intersection of e-cigarettes, combustible tobacco, and
marijuana. Enhanced regulatory enforcement efforts to regulate sales of unauthorized products are also
warranted. Blunts and Cigars was a predominant topic at the intersection of combustible tobacco and
marijuana. Future tobacco prevention interventions addressing the use of other substances with tobacco
products, can limit the progression of vulnerable populations to more addictive or harmful health
behaviors. Product features was a key topic among e-cigarette and marijuana-related conversations.
FDA’s regulatory efforts may consider regulating appealing product features highlighting health benefits.
Exposure to messages about product features highlighting health benefits may create favorable social
norms about vaping marijuana. As discussed previously, modular and compatible designs of e-cigarette
devices may facilitate transition to other types of devices (e.g. open-system) that are potentially
compatible with marijuana and raise implications for abuse liability(Allem, Majmundar, Dharmapuri,
Unger, & Cruz, 2019). Studies examining the downstream effects of exposure to such features are needed
to assess their impact and appeal. Lastly, Flavors, a key topic among e-cigarette and combustible-related
discussions, offer support to previous findings linking flavors with e-cigarette-use progression.
58
References
Alexander, J. P., Williams, P., & Lee, Y. O. (2019). Youth who use e-cigarettes regularly: A qualitative
study of behavior, attitudes, and familial norms. Preventive Medicine Reports, 13, 93-97.
doi:10.1016/j.pmedr.2018.11.011
Allem, J.-P., Dharmapuri, L., Leventhal, A. M., Unger, J. B., & Cruz, T. B. (2018). Hookah-related posts
to Twitter from 2017 to 2018: Thematic analysis. Journal of Medical Internet Research, 20(11),
1-1. doi: 10.2196/11669
Allem, J.-P., Dharmapuri, L., Unger, J. lB., & Cruz, T. B. (2018). Characterizing Juul-related posts on
Twitter. Drug and Alcohol Dependence, 190, 1-5. doi:10.1016/j.drugalcdep.2018.05.018
Allem, J.-P., Escobedo, P., Chu, K. H., Boley Cruz, T., & Unger, J. B. (2017). Images of little cigars and
cigarillos on Instagram identified by the hashtag #swisher: Thematic analysis. Journal of Medical
Internet Research, 19(7), e255. doi:10.2196/jmir.7634
Allem, J.-P., Escobedo, P., Chu, K. H., Soto, D. W., Cruz, T. B., & Unger, J. B. (2016). Campaigns and
counter campaigns: Reactions on Twitter to e-cigarette education. Tobacco Control.
doi:10.1136/tobaccocontrol-2015-052757
Allem, J.-P., Escobedo, P., & Dharmapuri, L. (2020). Cannabis surveillance with Twitter data: Emerging
topics and social bots. American Journal of Public Health, 110(3), 357-362.
doi:10.2105/AJPH.2019.305461
Allem, J.-P., & Ferrara, E. (2016). The importance of debiasing social media data to better understand e-
cigarette-related attitudes and behaviors. Journal of Medical Internet Research, 18(8), e219.
doi:10.2196/jmir.6185
Allem, J.-P., Ferrara, E., Uppu, S. P., Cruz, T. B., & Unger, J. B. (2017). E-cigarette surveillance with
social media data: Social bots, emerging topics, and trends. JMIR Public Health Surveillance,
3(4), e98. doi:10.2196/publichealth.8641
Allem, J.-P., Majmundar, A., Dharmapuri, L., Cruz, T. B., & Unger, J. B. (2019). E-liquid-related posts to
Twitter in 2018: Thematic analysis. Addictive Behaviors Reports, 10, 100196.
doi:10.1016/j.abrep.2019.100196
Allem, J. P., Cruz, T. B., Unger, J. B., Toruno, R., Herrera, J., & Kirkpatrick, M. G. (2018). Return of
cartoon to market e-cigarette-related products. Tobacco Control. doi:10.1136/tobaccocontrol-
2018-054437
Allem, J. P., Majmundar, A., Dharmapuri, L., Unger, J. B., & Cruz, T. B. (2019). Insights on electronic
cigarette products from reviews on the Reddit forum. Tobacco Prevention Cessation, 5, 29.
doi:10.18332/tpc/111533
Allem, J. P., Ramanujam, J., Lerman, K., Chu, K. H., Boley Cruz, T., & Unger, J. B. (2017). Identifying
sentiment of hookah-related posts on Twitter. JMIR Public Health Surveillance, 3(4), e74.
doi:10.2196/publichealth.8133
Anderson, M., & Jiang, J. (2019). Teens, social media & technology. Retrieved May 9, 2019
https://www.pewinternet.org/2018/05/31/teens-social-media-technology-2018/
Ayers, J. W., Leas, E. C., Allem, J.-P., Benton, A., Dredze, M., Althouse, B. M., . . . Unger, J. B. (2017).
Why do people use electronic nicotine delivery systems (electronic cigarettes)? A content
analysis of Twitter, 2012-2015. PloS One, 12(3), e0170702. doi:10.1371/journal.pone.0170702
Ayers, J. W., Ribisl, K. M., & Brownstein, J. S. (2011). Tracking the rise in popularity of electronic
nicotine delivery systems (electronic cigarettes) using search query surveillance. American
Journal of Preventive Medicine, 40(4), 448-453. doi:10.1016/j.amepre.2010.12.007
Baggio, S., Deline, S., Studer, J., Mohler-Kuo, M., Daeppen, J.-B., & Gmel, G. (2014). Routes of
administration of cannabis used for nonmedical purposes and associations with patterns of drug
use. Journal of Adolescent Health, 54(2), 235-240. doi:10.1016/j.jadohealth.2013.08.013
59
Barnett, T. E., Shensa, A., Kim, K. H., Cook, R. L., Nuzzo, E., & Primack, B. A. (2013). The predictive
utility of attitudes toward hookah tobacco smoking. American Journal of Health Behavior, 37(4),
433-439. doi:10.5993/AJHB.37.4.1
Barrington-Trimis, J. L., & Leventhal, A. M. (2018). Adolescents' use of “pod mod" e-cigarettes - urgent
concerns. New England Journal of Medicine, 379(12), 1099-1102. doi:10.1056/NEJMp1805758
Basáñez, T., Majmundar, A., Cruz, T. B., Allem, J.-P., & Unger, J. B. (2019). E-cigarettes are being
marketed as “vitamin delivery” devices. American Journal of Public Health, 109(2), 194-196.
doi:10.2105/ajph.2018.304804
Basanez, T., Majmundar, A., Cruz, T. B., & Unger, J. B. (2018). Vaping associated with healthy food
words: A content analysis of Twitter. Addictive Behaviors Reports, 8, 147-153.
doi:10.1016/j.abrep.2018.09.007
Bello, M. S., Khoddam, R., Stone, M. D., Cho, J., Yoon, Y., Lee, J. O., & Leventhal, A. M. (2019). Poly-
product drug use disparities in adolescents of lower socioeconomic status: Emerging trends in
nicotine products, marijuana products, and prescription drugs. Behaviour Research and Therapy,
115, 103-110. doi:10.1016/j.brat.2018.11.014
Benowitz, N. L., & Goniewicz, M. L. (2013). The regulatory challenge of electronic cigarettes. JAMA,
310(7), 685-686. doi:10.1001/jama.2013.109501
Berg, C. J., Stratton, E., Schauer, G. L., Lewis, M., Wang, Y., Windle, M., & Kegler, M. (2015).
Perceived harm, addictiveness, and social acceptability of tobacco products and marijuana among
young adults: Marijuana, hookah, and electronic cigarettes win. Substance Use and Misuse,
50(1), 79-89. doi:10.3109/10826084.2014.958857
Berry, K. M., Reynolds, L. M., Collins, J. M., Siegel, M. B., Fetterman, J. L., Hamburg, N. M., . . .
Stokes, A. (2019). E-cigarette initiation and associated changes in smoking cessation and
reduction: The population assessment of tobacco and health study, 2013-2015. Tobacco Control,
28(1), 42-49. doi:10.1136/tobaccocontrol-2017-054108
Blood, M. (2019). Fake vape gear is selling on l.A. Streets despite an emerging public health crisis. Time.
Retrieved from https://time.com/5678954/fake-vape-gear-health-crisis-vaping/
Blundell, M. S., Dargan, P. I., & Wood, D. M. (2018). The dark cloud of recreational drugs and vaping.
QJM: An International Journal of Medicine, 111(3), 145-148. doi:10.1093/qjmed/hcx049
Bodenreider, O. (2004). The Unified Medical Language System (UMLS): Integrating biomedical
terminology. Nucleic Acids Research, 32, D267-D270. doi:10.1093/nar/gkh061
Budney, A. J., Sargent, J. D., & Lee, D. C. (2015). Vaping cannabis (marijuana): Parallel concerns to e-
cigs? Addiction, 110(11), 1699-1704. doi:10.1111/add.13036
Callahan-Lyon, P. (2014). Electronic cigarettes: Human health effects. Tobacco Control, 23 Suppl 2, ii36-
40. doi:10.1136/tobaccocontrol-2013-051470
Cavazos-Rehg, P. A., Krauss, M. J., Sowles, S. J., & Bierut, L. J. (2016). Marijuana-related posts on
Instagram. Prevention Science, 17(6), 710-720. doi:10.1007/s11121-016-0669-9
Charles, M. (2003). ‘This mystery…’: A corpus-based study of the use of nouns to construct stance in
theses from two contrasting disciplines. Journal of English for Academic Purposes, 2(4), 313-326.
doi:10.1016/S1475-1585(03)00048-1
Choi, K., & Forster, J. (2013). Characteristics associated with awareness, perceptions, and use of
electronic nicotine delivery systems among young us midwestern adults. American Journal of
Public Health, 103(3), 556-561. doi:10.2105/AJPH.2012.300947
Chu, K.-H., Allem, J.-P., Unger, J. B., Cruz, T. B., Akbarpour, M., & Kirkpatrick, M. G. (2019).
Strategies to find audience segments on Twitter for e-cigarette education campaigns. Addictive
Behaviors, 91, 222-226. doi:10.1016/j.addbeh.2018.11.015
Chu, K.-H., Allem, J. P., Cruz, T. B., & Unger, J. B. (2016). Vaping on Instagram: Cloud chasing, hand
checks and product placement. Tobacco Control, 26(5), 575-578. doi:10.1136/tobaccocontrol-
2016-053052
60
Chu, K.-H., Majmundar, A., Allem, J.-P., Soto, D. W., Cruz, T. B., & Unger, J. B. (2019). Tobacco use
behaviors, attitudes, and demographic characteristics of tobacco opinion leaders and their
followers: Twitter analysis. Journal of Medical Internet Research, 21(6), e12676.
doi:10.2196/12676
Chu, K. H., Colditz, J. B., Primack, B. A., Shensa, A., Allem, J. P., Miller, E., . . . Cruz, T. B. (2018).
Juul: Spreading online and offline. Journal of Adolescent Health, 63(5), 582-586.
doi:10.1016/j.jadohealth.2018.08.002
Chu, K. H., Unger, J. B., Allem, J. P., Pattarroyo, M., Soto, D., Cruz, T. B., . . . Yang, C. C. (2015).
Diffusion of messages from an electronic cigarette brand to potential users through Twitter. PloS
One, 10(12), e0145387. doi:10.1371/journal.pone.0145387
Cohn, A., Johnson, A., Ehlke, S., & Villanti, A. C. (2016). Characterizing substance use and mental
health profiles of cigar, blunt, and non-blunt marijuana users from the National Survey of Drug
Use and Health. Drug and Alcohol Dependence, 160, 105-111.
doi:10.1016/j.drugalcdep.2015.12.017
Coleman, B. N., Johnson, S. E., Tessman, G. K., Tworek, C., Alexander, J., Dickinson, D. M., . . . Green,
K. M. (2016). "It's not smoke. It's not tar. It's not 4000 chemicals. Case closed": Exploring
attitudes, beliefs, and perceived social norms of e-cigarette use among adult users. Drug and
Alcohol Dependence, 159, 80-85. doi:10.1016/j.drugalcdep.2015.11.028
Cooper, M., Loukas, A., Harrell, M. B., & Perry, C. L. (2017). College students' perceptions of risk and
addictiveness of e-cigarettes and cigarettes. Journal of American College Health, 65(2), 103-111.
doi:10.1080/07448481.2016.1254638
Cornelis, E., Heuvinck, N., & Majmundar, A. (2020). The ambivalence story: Using refutation to counter
the negative effects of ambivalence in two-sided messages. International Journal of Advertising,
39(3), 410-432. doi:10.1080/02650487.2019.1624348
Correa, J. B., Brandon, K. O., Meltzer, L. R., Hoehn, H. J., Piñeiro, B., Brandon, T. H., & Simmons, V.
N. (2018). Electronic cigarette use among patients with cancer: Reasons for use, beliefs, and
patient-provider communication. Psycho-Oncology, 27(7), 1757-1764. doi:10.1002/pon.4721
Cwalina, S. N., Majmundar, A., Unger, J. B., Barrington-Trimis, J. L., & Pentz, M. A. (2020). Adolescent
menthol cigarette use and risk of nicotine dependence: Findings from the national population
assessment on tobacco and health (path) study. Drug and Alcohol Dependence, 206, 107715.
doi:10.1016/j.drugalcdep.2019.107715
Dai, H., Catley, D., Richter, K. P., Goggin, K., & Ellerbeck, E. F. (2018). Electronic cigarettes and future
marijuana use: A longitudinal study. Pediatrics, 141(5), e20173787. doi:10.1542/peds.2017-3787
Deb, A., Majmundar, A., Seo, S., Matsui, A., Tandon, R., Yan, S., . . . Ferrara, E. (2018, 28-31 Aug.
2018). Social bots for online public health interventions. Paper presented at the 2018 IEEE/ACM
International Conference on Advances in Social Networks Analysis and Mining (ASONAM).
Dutta-Bergman, M. J. (2004). Primary sources of health information: Comparisons in the domain of
health attitudes, health cognitions, and health behaviors. Health Communication, 16(3), 273-288.
doi:10.1207/S15327027HC1603_1
Eggers, M. E., Lee, Y. O., Jackson, K., Wiley, J. L., Porter, L., & Nonnemaker, J. M. (2017). Youth use
of electronic vapor products and blunts for administering cannabis. Addictive Behaviors, 70, 79-
82. doi:10.1016/j.addbeh.2017.02.020
El-Toukhy, S., & Choi, K. (2016). A risk-continuum categorization of product use among us youth
tobacco users. Nicotine & Tobacco Research, 18(7), 1596-1605. doi:10.1093/ntr/ntw008
England, L. J., Anderson, B. L., Tong, V. T. K., Mahoney, J., Coleman-Cowger, V. H., Melstrom, P., &
Schulkin, J. (2014). Screening practices and attitudes of obstetricians-gynecologists toward new
and emerging tobacco products. American Journal of Obstetrics and Gynecology, 211(6),
695.e691-695.e6957. doi:10.1016/j.ajog.2014.05.041
Etter, J. F., Zather, E., & Svensson, S. (2013). Analysis of refill liquids for electronic cigarettes.
Addiction, 108(9), 1671-1679. doi:10.1111/add.12235
61
Evans-Polce, R., Lanza, S., & Maggs, J. (2016). Heterogeneity of alcohol, tobacco, and other substance
use behaviors in U.S. College students: A latent class analysis. Addictive Behaviors, 53, 80-85.
doi:10.1016/j.addbeh.2015.10.010
F.D.A. (2019). Premarket Tobacco Product Applications for Electronic Nicotine Delivery Systems.
Retrieved from https://www.fda.gov/media/127853/download
F.D.A. (2020). The Real Cost campaign. Retrieved from https://www.fda.gov/tobacco-products/public-
health-education/real-cost-campaign
Fadus, M. C., Smith, T. T., & Squeglia, L. M. (2019). The rise of e-cigarettes, pod mod devices, and Juul
among youth: Factors influencing use, health implications, and downstream effects. Drug and
Alcohol Dependence, 201, 85-93. doi:10.1016/j.drugalcdep.2019.04.011
Fairman, B. J., & Anthony, J. C. (2017). Does starting to smoke cigars trigger onset of cannabis blunt
smoking? Nicotine & Tobacco Research, 20(3), 355-361. doi:10.1093/ntr/ntx015
Fishbein, M., & Cappella, J. N. (2006). The role of theory in developing effective health communications.
Journal of Communication, 56(s1), S1-S17. doi:10.1111/j.1460-2466.2006.00280.x
Fishbein, M., & Yzer, M. C. (2003). Using theory to design effective health behavior interventions.
Communication Theory, 13(2), 164-183. doi:10.1111/j.1468-2885.2003.tb00287.x
Fox, S. (2011). The social life of health information 2011.
Frohe, T., Leeman, R. F., Patock-Peckham, J., Ecker, A., Kraus, S., & Foster, D. W. (2018). Correlates of
cannabis vape-pen use and knowledge among U.S. College students. Addictive Behaviors
Reports, 7, 32-39. doi:10.1016/j.abrep.2017.11.004
Galstyan, E., Galimov, A., & Sussman, S. (2018). Commentary: The emergence of pod mods at vape
shops. Evaluation and the Health Professions, 0(0), 0163278718812976.
doi:10.1177/0163278718812976
Gibson, L. A., Creamer, M. R., Breland, A. B., Giachello, A. L., Kaufman, A., Kong, G., . . . Halpern-
Felsher, B. (2018). Measuring perceptions related to e-cigarettes: Important principles and next
steps to enhance study validity. Addictive Behaviors, 79, 219-225.
doi:10.1016/j.addbeh.2017.11.017
Glantz, S. A., & Bareham, D. W. (2018). E-cigarettes: Use, effects on smoking, risks, and policy
implications. Annual Review of Public Health, 39(1), 215-235. doi:10.1146/annurev-publhealth-
040617-013757
Glasser, A. M., Collins, L., Pearson, J. L., Abudayyeh, H., Niaura, R. S., Abrams, D. B., & Villanti, A. C.
(2017). Overview of Electronic Nicotine Delivery Systems: A systematic review. American
Journal of Preventive Medicine, 52(2), e33-e66. doi:10.1016/j.amepre.2016.10.036
Gorzkowski, J. A., Whitmore, R. M., Kaseeska, K. R., Brishke, J. K., & Klein, J. D. (2016). Pediatrician
knowledge, attitudes, and practice related to electronic cigarettes. Journal of Adolescent Health,
59(1), 81-86. doi:10.1016/j.jadohealth.2016.03.036
Grana, R., Benowitz, N., & Glantz, S. A. (2014). E-cigarettes: A scientific review. Circulation, 129(19),
1972-1986. doi:10.1161/CIRCULATIONAHA.114.007667
Haardörfer, R., Berg, C. J., Lewis, M., Payne, J., Pillai, D., McDonald, B., & Windle, M. (2016).
Polytobacco, marijuana, and alcohol use patterns in college students: A latent class analysis.
Addictive Behaviors, 59, 58-64. doi:10.1016/j.addbeh.2016.03.034
Hammig, B., Daniel-Dobbs, P., & Blunt-Vinti, H. (2017). Electronic cigarette initiation among minority
youth in the United States. The American Journal of Drug and Alcohol Abuse, 43(3), 306-310.
doi:10.1080/00952990.2016.1203926
Harris, J. K., Moreland-Russell, S., Choucair, B., Mansour, R., Staub, M., & Simmons, K. (2014).
Tweeting for and against public health policy: Response to the Chicago Department of Public
Health's electronic cigarette Twitter campaign. Journal of Medical Internet Research, 16(10),
e238. doi:10.2196/jmir.3622
Hart, J. L., Walker, K. L., Sears, C. G., Lee, A. S., Ridner, S. L., & Keith, R. J. (2018). E-cigarette use
and perceived health change: Better health through vaping? Tobacco Induced Diseases,
16(October). doi:10.18332/tid/95218
62
Hart, J. L., Walker, K. L., Sears, C. G., Lee, A. S., Smith, C., Siu, A., . . . Ridner, S. L. (2016). Vape shop
employees: Public health advocates? Tob Prev Cessat, 2(Suppl). doi:10.18332/tpc/67800
Hartnett, K. P., Kite-Powell, A., Patel, M. T., Haag, B. L., Sheppard, M. J., Dias, T. P., . . . Adjemian, J.
(2019). Syndromic surveillance for e-cigarette, or vaping, product use-associated lung injury.
New England Journal of Medicine. doi:10.1056/NEJMsr1915313
Haugtvedt, C. P., & Petty, R. E. (1989). Need for cognition and attitude persistence. Advances in
Consumer Research, 16, 33-36.
Hausmann, J. S., Touloumtzis, C., White, M. T., Colbert, J. A., & Gooding, H. C. (2017). Adolescent and
young adult use of social media for health and its implications. Journal of Adolescent Health,
60(6), 714-719. doi:10.1016/j.jadohealth.2016.12.025
Hébert, E. T., Case, K. R., Kelder, S. H., Delk, J., Perry, C. L., & Harrell, M. B. (2017). Exposure and
engagement with tobacco- and e-cigarette-related social media. Journal of Adolescent Health,
61(3), 371-377. doi:10.1016/j.jadohealth.2017.04.003
Heinz, A. J., Giedgowd, G. E., Crane, N. A., Veilleux, J. C., Conrad, M., Braun, A. R., . . . Kassel, J. D.
(2013). A comprehensive examination of hookah smoking in college students: Use patterns and
contexts, social norms and attitudes, harm perception, psychological correlates and co-occurring
substance use. Addictive Behaviors, 38(11), 2751-2760. doi:10.1016/j.addbeh.2013.07.009
Henningfield, J. E., Hatsukami, D. K., Zeller, M., & Peters, E. (2011). Conference on abuse liability and
appeal of tobacco products: Conclusions and recommendations. Drug and Alcohol Dependence,
116(1-3), 1-7. doi:10.1016/j.drugalcdep.2010.12.009
Henriksen, L., Schleicher, N. C., Ababseh, K., Johnson, T. O., & Fortmann, S. P. (2018). Marijuana as a
‘concept’ flavour for cigar products: Availability and price near california schools. Tobacco
Control, 27(5), 585. doi:10.1136/tobaccocontrol-2017-053961
Hoffman, L., Delahanty, J., Johnson, S. E., & Zhao, X. (2018). Sexual and gender minority cigarette
smoking disparities: An analysis of 2016 Behavioral Risk Factor Surveillance System data.
Preventive Medicine, 113, 109-115. doi:10.1016/j.ypmed.2018.05.014
Hua, M., Alfi, M., & Talbot, P. (2013a). Health-related effects reported by electronic cigarette users in
online forums. Journal of Medical Internet Research, 15(4), e59. doi:10.2196/jmir.2324
Huang, G. C., Soto, D., Fujimoto, K., & Valente, T. W. (2014). The interplay of friendship networks and
social networking sites: Longitudinal analysis of selection and influence effects on adolescent
smoking and alcohol use. American Journal of Public Health, 104(8), e51-59.
doi:10.2105/AJPH.2014.302038
Huang, J., Feng, B., Weaver, S. R., Pechacek, T. F., Slovic, P., & Eriksen, M. P. (2019). Changing
perceptions of harm of e-cigarette vs cigarette use among adults in 2 U.S. national surveys from
2012 to 2017. JAMA Network Open, 2(3), e191047-e191047.
doi:10.1001/jamanetworkopen.2019.1047
Huang, J., Kornfield, R., Szczypka, G., & Emery, S. L. (2014). A cross-sectional examination of
marketing of electronic cigarettes on Twitter. Tobacco Control, 23 Suppl 3, iii26-30.
doi:10.1136/tobaccocontrol-2014-051551
Huh, J., & Leventhal, A. M. (2016). Progression of poly-tobacco product use patterns in adolescents.
American Journal of Preventive Medicine, 51(4), 513-517. doi:10.1016/j.amepre.2016.04.004
Husten, C. G., & Deyton, L. R. (2013). Understanding the tobacco control act: Efforts by the U.S. Food
and Drug Administration to make tobacco-related morbidity and mortality part of the USA's past,
not its future. Lancet, 381(9877), 1570-1580. doi:10.1016/S0140-6736(13)60735-7
Jamal, A., Gentzke, A., Hu, S. S., Cullen, K. A., Apelberg, B. J., Homa, D. M., & King, B. A. (2017).
Tobacco use among middle and high school students - United States, 2011-2016. MMWR.
Morbidity and mortality weekly report, 66(23), 597-603. doi:10.15585/mmwr.mm6623a1
Jamison, A. M., Broniatowski, D. A., & Quinn, S. C. (2019). Malicious actors on Twitter: A guide for
public health researchers. American Journal of Public Health, 109(5), 688-692.
doi:10.2105/AJPH.2019.304969
63
Kavuluru, R., Han, S., & Hahn, E. J. (2019). On the popularity of the USB flash drive-shaped electronic
cigarette Juul. Tobacco Control, 28(1), 110-112. doi:10.1136/tobaccocontrol-2018-054259
Kim, H. S., Yang, S., Kim, M., Hemenway, B., Ungar, L., & Cappella, J. N. (2019). An experimental
study of recommendation algorithms for tailored health communication. Computational
Communication Research, 1(1), 103-129.
King, B. A., Jones, C. M., Baldwin, G. T., & Briss, P. A. (2020). The evali and youth vaping epidemics -
implications for public health. New England Journal of Medicine, 382(8), 689-691.
doi:10.1056/NEJMp1916171
Kirham, C. (2018). Special report: Juul copycats flood e-cig market, despite FDA rule. Reuters. Retrieved
from https://www.reuters.com/article/us-vaping-regulation-Juul-specialreport/special-report-Juul-
copycats-flood-e-cig-market-despite-fda-rule-idUSKCN1M418W
Kirkpatrick, M. G., Cruz, T. B., Unger, J. B., Herrera, J., Schiff, S., & Allem, J. P. (2019). Cartoon-based
e-cigarette marketing: Associations with susceptibility to use and perceived expectations of use.
Drug and Alcohol Dependence, 201, 109-114. doi:10.1016/j.drugalcdep.2019.04.018
Klein, E. G., Berman, M., Hemmerich, N., Carlson, C., Htut, S., & Slater, M. (2016). Online e-cigarette
marketing claims: A systematic content and legal analysis. Tobacco Regulatroy Science, 2(3),
252-262. doi:10.18001/TRS.2.3.5
Kollath-Cattano, C., Dorman, T., Albano, A. W., Jr., Jindal, M., Strayer, S. M., & Thrasher, J. F. (2019).
E-cigarettes and the clinical encounter: Physician perspectives on e-cigarette safety, effectiveness,
and patient educational needs. Journal of Evaluation in Clinical Practice. doi:10.1111/jep.13111
Kostygina, G., Glantz, S. A., & Ling, P. M. (2016). Tobacco industry use of flavours to recruit new users
of little cigars and cigarillos. Tobacco Control, 25(1), 66. doi:10.1136/tobaccocontrol-2014-
051830
Krauss, M. J., Sowles, S. J., Mylvaganam, S., Zewdie, K., Bierut, L. J., & Cavazos-Rehg, P. A. (2015).
Displays of dabbing marijuana extracts on youtube. Drug and Alcohol Dependence, 155, 45-51.
doi:10.1016/j.drugalcdep.2015.08.020
Krejzl, P., Hourová, B., & Steinberger, J. (2017). Stance detection in online discussions. arXiv preprint
arXiv:1701.00504.
Kreuter, M. W., & Wray, R. J. (2003). Tailored and targeted health communication: Strategies for
enhancing information relevance. American Journal of Health Behavior, 27 Suppl 3, S227-232.
Landry, R. L., Groom, A. L., Vu, T.-H. T., Stokes, A. C., Berry, K. M., Kesh, A., . . . Payne, T. J. (2019).
The role of flavors in vaping initiation and satisfaction among u.S. Adults. Addictive Behaviors,
99, 106077. doi:10.1016/j.addbeh.2019.106077
Lazard, A. J., Wilcox, G. B., Tuttle, H. M., Glowacki, E. M., & Pikowski, J. (2017). Public reactions to e-
cigarette regulations on Twitter: A text mining analysis. Tobacco Control, 26(e2), e112-e116.
doi:10.1136/tobaccocontrol-2016-053295
Lee, D. C., Crosier, B. S., Borodovsky, J. T., Sargent, J. D., & Budney, A. J. (2016). Online survey
characterizing vaporizer use among cannabis users. Drug and Alcohol Dependence, 159, 227-
233. doi:10.1016/j.drugalcdep.2015.12.020
Leventhal, A. M., Strong, D. R., Kirkpatrick, M. G., Unger, J. B., Sussman, S., Riggs, N. R., . . . Audrain-
McGovern, J. (2015). Association of electronic cigarette use with initiation of combustible
tobacco product smoking in early adolescence. JAMA, 314(7), 700-707.
doi:10.1001/jama.2015.8950
Link, A. R., Cawkwell, P. B., Shelley, D. R., & Sherman, S. E. (2015). An exploration of online
behaviors and social media use among hookah and electronic-cigarette users. Addictive
Behaviors Reports, 2, 37-40. doi:10.1016/j.abrep.2015.05.006
Love, B., Himelboim, I., Holton, A., & Stewart, K. (2013). Twitter as a source of vaccination
information: Content drivers and what they are saying. American Journal of Infection Control,
41(6), 568-570. doi:10.1016/j.ajic.2012.10.016
64
Majeed, B. A., Weaver, S. R., Gregory, K. R., Whitney, C. F., Slovic, P., Pechacek, T. F., & Eriksen, M.
P. (2017). Changing perceptions of harm of e-cigarettes among U.S. Adults, 2012-2015.
American Journal of Preventive Medicine, 52(3), 331-338. doi:10.1016/j.amepre.2016.08.039
Majmundar, A., Allem, J.-P., Boley Cruz, T., & Unger, J. B. (2018). The why we retweet scale. PloS
One, 13(10), e0206076. doi:10.1371/journal.pone.0206076
Majmundar, A., Allem, J. P., Cruz, T. B., & Unger, J. B. (2019). Where do people vape? Insights from
Twitter data. International Journal of Environmental Research and Public Health, 16(17).
doi:10.3390/ijerph16173056
Majmundar, A., Cerrada, C., Fang, W., & Huh, J. (2020). In lieu of smoking: Are smokers more likely to
enact certain types of implementation intention plans during a jit smoking cessation? International
Journal of Behavioral Medicine, 27(5), 556-564. doi:10.1007/s12529-020-09893-6
Majmundar, A., Chou, C.-P., Cruz, T. B., & Unger, J. B. (2018). Relationship between social media
engagement and e-cigarette policy support. Addictive Behaviors Reports, 100155.
doi:10.1016/j.abrep.2018.100155
Majmundar, A., Kirkpatrick, M., Cruz, T. B., Unger, J. B., & Allem, J.-P. (2019). Characterising
KandyPens-related posts to Instagram: Implications for nicotine and cannabis use. Tobacco
Control, tobaccocontrol-2019-055006. doi:10.1136/tobaccocontrol-2019-055006
Martinez, L. S., Hughes, S., Walsh-Buhi, E. R., & Tsou, M.-H. (2018). “Okay, we get it. You vape”: An
analysis of geocoded content, context, and sentiment regarding e-cigarettes on Twitter. Journal of
Health Communication, 23(6), 550-562. doi:10.1080/10810730.2018.1493057
McDonald, E. A., Popova, L., & Ling, P. M. (2016). Traversing the triangulum: The intersection of
tobacco, legalised marijuana and electronic vaporisers in denver, colorado. Tobacco Control,
25(Suppl 1), i96. doi:10.1136/tobaccocontrol-2016-053091
McMillen, R., Klein, J. D., Wilson, K., Winickoff, J. P., & Tanski, S. (2019). E-cigarette use and future
cigarette initiation among never smokers and relapse among former smokers in the path study.
Public Health Reports, 134(5), 528-536. doi:10.1177/0033354919864369
Montgomery, L., Zapolski, T., Banks, D. E., & Floyd, A. (2019). Puff, puff, drink: The association
between blunt and alcohol use among African American adolescents and young adults. The
American journal of orthopsychiatry, 89(5), 609-615. doi:10.1037/ort0000400
Morean, M. E., Kong, G., Camenga, D. R., Cavallo, D. A., & Krishnan-Sarin, S. (2015). High school
students' use of electronic cigarettes to vaporize cannabis. Pediatrics, 136(4), 611-616.
doi:10.1542/peds.2015-1727
Morean, M. E., Lipshie, N., Josephson, M., & Foster, D. (2017). Predictors of adult e-cigarette users
vaporizing cannabis using e-cigarettes and vape-pens. Substance Use and Misuse, 52(8), 974-981.
doi:10.1080/10826084.2016.1268162
Napper, L. E., Froidevaux, N. M., & LaBrie, J. W. (2016). Being blunt about marijuana: Parent
communication about marijuana with their emerging adult children. Prevention Science, 17(7),
882-891. doi:10.1007/s11121-016-0681-0
National Academies of Sciences Engineering and Medicine (U.S.). (2018). Public health consequences of
e-cigarettes. Washington, DC: The National Academies Press.
National Cancer Institute. (2019). NCI dictionary of cancer terms. Retrieved from
https://www.cancer.gov/publications/dictionaries/cancer-terms/def/electronic-cigarette
Oncken, C., Ricci, K. A., Kuo, C.-L., Dornelas, E., Kranzler, H. R., & Sankey, H. Z. (2017). Correlates of
electronic cigarettes use before and during pregnancy. Nicotine & Tobacco Research, 19(5), 585-
590. doi:10.1093/ntr/ntw225
Orr, D., Baram-Tsabari, A., & Landsman, K. (2016). Social media as a platform for health-related public
debates and discussions: The polio vaccine on facebook. Israel Journal of Health Policy Research,
5(1), 34. doi:10.1186/s13584-016-0093-4
Park, A., & Conway, M. (2017). Tracking health related discussions on Reddit for public health
applications. AMIA Annu Symp Proc, 2017, 1362-1371.
65
Pennington, J., Socher, R., & Manning, C. (2014). Glove: Global vectors for word representation. Paper
presented at the Proceedings of the 2014 Conference on Empirical Methods in Natural Language
Processing (EMNLP).
Pepper, J. K., Emery, S. L., Ribisl, K. M., Rini, C. M., & Brewer, N. T. (2015). How risky is it to use e-
cigarettes? Smokers’ beliefs about their health risks from using novel and traditional tobacco
products. Journal of Behavioral Medicine, 38(2), 318-326. doi:10.1007/s10865-014-9605-2
Pepper, J. K., Lee, Y. O., Watson, K. A., Kim, A. E., Nonnemaker, J. M., & Farrelly, M. C. (2017). Risk
factors for youth e-cigarette “vape trick” behavior. Journal of Adolescent Health, 61(5), 599-605.
doi:10.1016/j.jadohealth.2017.05.010
Pepper, J. K., McRee, A.-L., & Gilkey, M. B. (2014). Healthcare providers' beliefs and attitudes about
electronic cigarettes and preventive counseling for adolescent patients. Journal of Adolescent
Health, 54(6), 678-683. doi:10.1016/j.jadohealth.2013.10.001
Perrin, A., & Anderson, M. (2017). Social media fact sheet Retrieved May 9, 2019
http://www.pewinternet.org/fact-sheet/social-media/
Peterson, E. B., Fisher, C. L., & Zhao, X. (2018). Pediatric primary healthcare providers’ preferences,
experiences and perceived barriers to discussing electronic cigarettes with adolescent patients.
Journal of Communication in Healthcare, 11(4), 245-251. doi:10.1080/17538068.2018.1460960
Petty, R. E., & Cacioppo, J. T. (1983). Central and peripheral routes to persuasion: Application to
advertising. Advertising and consumer psychology, 1, 3-23.
Petty, R. E., & Cacioppo, J. T. (1984). The effects of involvement on responses to argument quantity and
quality - central and peripheral routes to persuasion. Journal of Personality and Social
Psychology, 46(1), 69-81. doi:10.1037/0022-3514.46.1.69
Petty, R. E., & Cacioppo, J. T. (1986). The Elaboration Likelihood Model of persuasion. In L. Berkowitz
(Ed.), Advances in experimental social psychology (Vol. 19, pp. 123-205): Academic Press.
Pisinger, C., & Dossing, M. (2014). A systematic review of health effects of electronic cigarettes.
Preventive Medicine, 69, 248-260. doi:10.1016/j.ypmed.2014.10.009
Pokhrel, P., Fagan, P., Herzog, T. A., Chen, Q., Muranaka, N., Kehl, L., & Unger, J. B. (2016). E-
cigarette advertising exposure and implicit attitudes among young adult non-smokers. Drug and
Alcohol Dependence, 163, 134-140. doi:10.1016/j.drugalcdep.2016.04.008
Pratt, S. I., Sargent, J., Daniels, L., Santos, M. M., & Brunette, M. (2016). Appeal of electronic cigarettes
in smokers with serious mental illness. Addictive Behaviors, 59, 30-34.
doi:10.1016/j.addbeh.2016.03.009
Prochaska, J. J. (2018). The public health consequences of e-cigarettes: A review by the national
academies of sciences. A call for more research, a need for regulatory action. Addiction.
doi:10.1111/add.14478
Rabin, R. A., & George, T. P. (2015). A review of co-morbid tobacco and cannabis use disorders:
Possible mechanisms to explain high rates of co-use. The American Journal on Addictions, 24(2),
105-116. doi:10.1111/ajad.12186
Rhoades, D. A., Comiford, A. L., Dvorak, J. D., Ding, K., Hopkins, M., Spicer, P., . . . Doescher, M. P.
(2019). Perceptions of smoking and vaping on weight control among adult american indians who
smoke. Journal of Community Health, 44(6), 1120-1126. doi:10.1007/s10900-019-00694-x
Rigotti, N. A., Chang, Y., Tindle, H. A., Kalkhoran, S. M., Levy, D. E., Regan, S., . . . Singer, D. E.
(2018). Association of e-cigarette use with smoking cessation among smokers who plan to quit
after a hospitalization: A prospective studyassociation of e-cigarette use with smoking cessation
after hospitalization. Annals of Internal Medicine, 168(9), 613-620. doi:10.7326/m17-2048
Rimmer, A., & Iacobucci, G. (2019). Vaping: Cdc investigation continues after patient dies from severe
lung injuries. BMJ, 366, l5320. doi:10.1136/bmj.l5320
Roditis, M., Delucchi, K., Cash, D., & Halpern-Felsher, B. (2016). Adolescents' perceptions of health
risks, social risks, and benefits differ across tobacco products. Journal of Adolescent Health,
58(5), 558-566. doi:10.1016/j.jadohealth.2016.01.012
66
Rosenman, R., Tennekoon, V., & Hill, L. G. (2011). Measuring bias in self-reported data. International
Journal of Behavioural and Healthcare Research, 2(4), 320-332. doi:10.1504/IJBHR.2011.043414
Schauer, G. L., & Peters, E. N. (2018). Correlates and trends in youth co-use of marijuana and tobacco in
the united states, 2005-2014. Drug and Alcohol Dependence, 185, 238-244.
doi:10.1016/j.drugalcdep.2017.12.007
Schauer, G. L., Rosenberry, Z. R., & Peters, E. N. (2017). Marijuana and tobacco co-administration in
blunts, spliffs, and mulled cigarettes: A systematic literature review. Addictive Behaviors, 64,
200-211. doi:10.1016/j.addbeh.2016.09.001
Schroeder, M. J., & Hoffman, A. C. (2014). Electronic cigarettes and nicotine clinical pharmacology.
Tobacco Control, 23 Suppl 2, ii30-35. doi:10.1136/tobaccocontrol-2013-051469
Sharma, R., Wigginton, B., Meurk, C., Ford, P., & Gartner, C. E. (2016). Motivations and limitations
associated with vaping among people with mental illness: A qualitative analysis of Reddit
discussions. International Journal of Environmental Research and Public Health, 14(1), 7.
doi:10.3390/ijerph14010007
Sherman, L. E., Payton, A. A., Hernandez, L. M., Greenfield, P. M., & Dapretto, M. (2016). The power of
the like in adolescence: Effects of peer influence on neural and behavioral responses to social
media. Psychological Science, 27(7), 1027-1035. doi:10.1177/0956797616645673
Sidani, J. E., Colditz, J. B., Barrett, E. L., Shensa, A., Chu, K. H., James, A. E., & Primack, B. A. (2019).
I wake up and hit the Juul: Analyzing Twitter for Juul nicotine effects and dependence. Drug and
Alcohol Dependence, 204, 107500. doi:10.1016/j.drugalcdep.2019.06.005
Singh, B., Hrywna, M., Wackowski, O. A., Delnevo, C. D., Jane Lewis, M., & Steinberg, M. B. (2017).
“Knowledge, recommendation, and beliefs of e-cigarettes among physicians involved in tobacco
cessation: A qualitative study”. Preventive Medicine Reports, 8, 25-29.
doi:10.1016/j.pmedr.2017.07.012
Smith, D. M., & Goniewicz, M. L. (2020). The role of policy in the EVALI outbreak: Solution or
contributor? Lancet Respir Med. doi:10.1016/S2213-2600(20)30065-5
Soldz, S., Huyser, D. J., & Dorsey, E. (2003). The cigar as a drug delivery device: Youth use of blunts.
Addiction, 98(10), 1379-1386. doi:10.1046/j.1360-0443.2003.00492.x
Soneji, S., Barrington-Trimis, J. L., Wills, T. A., Leventhal, A. M., Unger, J. B., Gibson, L. A., . . .
Sargent, J. D. (2017). Association between initial use of e-cigarettes and subsequent cigarette
smoking among adolescents and young adults: A systematic review and meta-analysise-cigarette
use and subsequent cigarette smoking in adolescents and young adultse-cigarette use and
subsequent cigarette smoking in adolescents and young adults. JAMA pediatrics, 171(8), 788-
797. doi:10.1001/jamapediatrics.2017.1488
Soneji, S., Pierce, J. P., Choi, K., Portnoy, D. B., Margolis, K. A., Stanton, C. A., . . . Sargent, J. (2017).
Engagement with online tobacco marketing and associations with tobacco product use among u.S.
Youth. Journal of Adolescent Health, 61(1), 61-69. doi:10.1016/j.jadohealth.2017.01.023
St Helen, G., Dempsey, D. A., Havel, C. M., Jacob, P., 3rd, & Benowitz, N. L. (2017). Impact of e-liquid
flavors on nicotine intake and pharmacology of e-cigarettes. Drug and Alcohol Dependence, 178,
391-398. doi:10.1016/j.drugalcdep.2017.05.042
Stiles, M. (2019). L.A. County expands smoking ban to vaping tobacco and smoking pot in public.
Retrieved from https://www.latimes.com/local/lanow/la-me-smoking-ban-beaches-vape-
cannabis-20190326-story.html
Stratton, K. R., Kwan, L. Y., Eaton, D. L., & National Academies of Sciences Engineering and Medicine
(U.S.). Committee on the Review of the Health Effects of Electronic Nicotine Delivery Systems.
(2018). Public health consequences of e-cigarettes. Washington, DC: The National Academies
Press.
T.R.D.R.P. (2016). The triangulum: Tobacco, marijuana and ecigarettes. Retrieved from
http://www.trdrp.org/files/triangulum/phil-gardiner-slides.pdf
67
Tan, A. S. L., Lee, C.-J., Nagler, R. H., & Bigman, C. A. (2017). To vape or not to vape? Effects of
exposure to conflicting news headlines on beliefs about harms and benefits of electronic cigarette
use: Results from a randomized controlled experiment. Preventive Medicine, 105, 97-103.
doi:10.1016/j.ypmed.2017.08.024
Teng, S., Khong, K. W., & Goh, W. W. (2014). Conceptualizing persuasive messages using ELM in
social media. Journal of Internet Commerce, 13(1), 65-87. doi:10.1080/15332861.2014.910729
Tierney, P. A., Karpinski, C. D., Brown, J. E., Luo, W., & Pankow, J. F. (2016). Flavour chemicals in
electronic cigarette fluids. Tobacco Control, 25(e1), e10-15. doi:10.1136/tobaccocontrol-2014-
052175
Truth Initiative. (2019). Truth confronts Juul and other e-cigarette manufacturers in new campaign,
‘tested on humans’ [Press release]. Retrieved from https://truthinitiative.org/press/press-
release/truth-confronts-Juul-and-other-e-cigarette-manufacturers-new-campaign-tested
Unger, J. B., Soto, D. W., & Leventhal, A. (2016). E-cigarette use and subsequent cigarette and marijuana
use among hispanic young adults. Drug and Alcohol Dependence, 163, 261-264.
doi:10.1016/j.drugalcdep.2016.04.027
Unger, J. B., Urman, R., Cruz, T. B., Majmundar, A., Barrington-Trimis, J., Pentz, M. A., & McConnell,
R. (2018). Talking about tobacco on Twitter is associated with tobacco product use. Preventive
Medicine, 114, 54-56. doi:10.1016/j.ypmed.2018.06.006
US Department of Health Human Services. (2016). E-cigarette use among youth and young adults. A
report of the surgeon general. Atlanta, GA.
Vance, K., Howe, W., & Dellavalle, R. P. (2009). Social internet sites as a source of public health
information. Dermatologic Clinics, 27(2), 133-136. doi:10.1016/j.det.2008.11.010
Villanti, A. C., Feirman, S. P., Niaura, R. S., Pearson, J. L., Glasser, A. M., Collins, L. K., & Abrams, D.
B. (2018). How do we determine the impact of e-cigarettes on cigarette smoking cessation or
reduction? Review and recommendations for answering the research question with scientific
rigor. Addiction, 113(3), 391-404. doi:10.1111/add.14020
Vraga, E. K., & Bode, L. (2018). I do not believe you: How providing a source corrects health
misperceptions across social media platforms. Information, Communication & Society, 21(10),
1337-1353. doi:10.1080/1369118X.2017.1313883
Wackowski, O. A., Bover Manderski, M. T., & Delnevo, C. D. (2015). Smokers' sources of e-cigarette
awareness and risk information. Preventive Medicine Reports, 2, 906-910.
doi:10.1016/j.pmedr.2015.10.006
Wackowski, O. A., & Delnevo, C. D. (2015). Young adults’ risk perceptions of various tobacco products
relative to cigarettes: Results from the national young adult health survey. Health Education and
Behavior, 43(3), 328-336. doi:10.1177/1090198115599988
Wagoner, K. G., Berman, M., Rose, S. W., Song, E., Cornacchione Ross, J., Klein, E. G., . . . Sutfin, E. L.
(2019). Health claims made in vape shops: An observational study and content analysis. Tobacco
Control, tobaccocontrol-2018-054537. doi:10.1136/tobaccocontrol-2018-054537
Walley, S. C., Wilson, K. M., Winickoff, J. P., & Groner, J. (2019). A public health crisis: Electronic
cigarettes, vape, and Juul. Pediatrics, 143(6), e20182741. doi:10.1542/peds.2018-2741
Weinberger, A. H., Platt, J., Copeland, J., & Goodwin, R. D. (2018). Is cannabis use associated with
increased risk of cigarette smoking initiation, persistence, and relapse? Longitudinal data from a
representative sample of us adults. Journal of Clinical Psychiatry, 79(2).
doi:10.4088/JCP.17m11522
Wenk, A. (2019). Atlanta eyes smoking, vaping ban including at airport. Retrieved from
https://www.bizjournals.com/atlanta/news/2019/04/15/atlanta-eyes-smoking-vaping-ban-
including-at.html
Wigginton, B., Gartner, C., & Rowlands, I. J. (2017). Is it safe to vape? Analyzing online forums
discussing e-cigarette use during pregnancy. Women's Health Issues, 27(1), 93-99.
doi:10.1016/j.whi.2016.09.008
68
Wilsey, B., Marcotte, T., Deutsch, R., Gouaux, B., Sakai, S., & Donaghe, H. (2013). Low-dose vaporized
cannabis significantly improves neuropathic pain. Journal of Pain, 14(2), 136-148.
doi:10.1016/j.jpain.2012.10.009
Wilsey, B., Marcotte, T. D., Deutsch, R., Zhao, H., Prasad, H., & Phan, A. (2016). An exploratory human
laboratory experiment evaluating vaporized cannabis in the treatment of neuropathic pain from
spinal cord injury and disease. Journal of Pain, 17(9), 982-1000. doi:10.1016/j.jpain.2016.05.010
Yang, Q., Sangalang, A., Rooney, M., Maloney, E., Emery, S., & Cappella, J. N. (2018). How is
marijuana vaping portrayed on youtube? Content, features, popularity and retransmission of
vaping marijuana youtube videos. Journal of Health Communication, 23(4), 360-369.
doi:10.1080/10810730.2018.1448488
Zeng, Q. T., & Tse, T. (2006). Exploring and developing consumer health vocabularies. Journal of
American Medical Informatics Association, 13(1), 24-29. doi:10.1197/jamia.M1761
Zhang, M., & Yang, C. C. (2014). Classification of online health discussions with text and health feature
sets. Paper presented at the Workshops at the Twenty-Eighth AAAI Conference on Artificial
Intelligence.
Zhu, S. H., Sun, J. Y., Bonnevie, E., Cummins, S. E., Gamst, A., Yin, L., & Lee, M. (2014). Four hundred
and sixty brands of e-cigarettes and counting: Implications for product regulation. Tobacco
Control, 23 Suppl 3, iii3-9. doi:10.1136/tobaccocontrol-2014-051670
Abstract (if available)
Abstract
Electronic-cigarette (e-cigarette) use (vaping) has increased sharply among U.S. youth and adults, signaling the emergence of a new public health epidemic. While emerging findings suggest that vaping may facilitate smoking cessation, it may also result in adverse health effects. Additionally, evolving patterns of vaping behaviors and the introduction of new products compatible with nicotine and marijuana among never and current smokers further complicate the challenge of assessing the health effects of these products, devising effective strategies for tobacco regulation, and communicating scientific findings with the public. Social media surveillance of real-time, naturally occurring public conversations about the health effects of vaping offers an opportunity to capture timely insights that complement and extend findings from traditional research methodologies. This dissertation: (1) monitors Twitter discourse about predominant categories of health effects of e-cigarettes over two years (2018 and 2019), (2) examines associations between categories of health effects of e-cigarettes discussed on Twitter and subsequent stance toward e-cigarettes (pro-, anti-, neutral-), (3) explores key topics of discussions pertaining to co-occurring mentions of marijuana, tobacco, and e-cigarettes based on the Triangulum framework, which has been adopted by the state of California to investigate the interrelated influences of use of these substances. This dissertation found that: (1) In 2019, Death was most commonly referenced followed by Neurological and Respiratory health effects in 2019. In 2018, Neurological was the most common health effect, followed by Mental Health and Death, (2) Majority of posts were pro-e-cigarettes. Posts pertaining to Neurological health effects were predominantly pro-e-cigarettes followed Mental Health, and Respiratory health effects. Pain was significantly less likely to be pro-e-cigarette compared to Mental Health effects, (3) Person tagging was the most predominant topic across all intersections of the Triangulum (marijuana, tobacco, and e-cigarettes). Discussions about Illicit products in the marketplace was one of the key topics at in the intersection of e-cigarettes, combustible tobacco, and marijuana, Blunts and Cigars at the intersection of combustible tobacco and marijuana, Product features and undesirable Smell among e-cigarette and marijuana-related conversations, and Flavors among e-cigarette and combustible-related discussions. Future health communication campaigns may consider enhancing awareness of health risks associated with adverse outcomes related to the predominant categories of health effects: Death, Mental Health, and Neurological. Another point of intervention to curb e-cigarette use could be when e-cigarette users discuss health-related concerns linked to a subsequent Pro-Stance towards e-cigarettes online. By examining the intersections of marijuana and other tobacco products, this study offers inputs for designing comprehensive FDA regulations including regulating product features that may be appealing and improving enforcement efforts to curb sales of illicit products. Future public health interventions for tobacco products may warrant including leveraging discussions about undesirable aspects of tobacco use such as the undesirable smell of smoking and adverse health outcomes.
Linked assets
University of Southern California Dissertations and Theses
Conceptually similar
PDF
Effects of flavorings in electronic cigarettes on the use and appeal of e-cigarettes among youth and adults
PDF
Anxiety symptoms and nicotine use among adolescents and young adults
PDF
The effect of an educational fotonovela on the prevention of secondhand smoke
PDF
Cultural risk and protective factors for tobacco use behaviors and depressive symptoms among American Indian adolescents in California
PDF
Examining the built and online environments to understand cannabis and e-cigarette availability, marketing, and product use in Southern California
PDF
Smoke-free housing policies and secondhand smoke exposure in low income multiunit housing in Los Angeles County
PDF
Social self-control and adolescent substance use
PDF
Evaluating social networks and impact of micro-influencers who promote e-cigarettes on social media
PDF
Role transitions, past life events, and their associations with multiple categories of substance use among emerging adults
PDF
Normative and network influences on electronic cigarette use among adolescents
PDF
The influence of contextual factors on the processes of adoption and implementation of evidence-based substance use prevention and tobacco cessation programs in schools
PDF
The role of social support in the relationship between adverse childhood experiences and addictive behaviors across adolescence and young adulthood
PDF
Energy drink consumption, substance use and attention-deficit/hyperactivity disorder among adolescents
PDF
Emotion regulation and heart rate variability among smokers
PDF
Substance use disparities in adolescents of lower socioeconomic status: emerging trends in electronic cigarettes, alternative tobacco products, marijuana, and prescription drug use
Asset Metadata
Creator
Majmundar, Anuja
(author)
Core Title
Tobacco and marijuana surveillance using Twitter data
School
Keck School of Medicine
Degree
Doctor of Philosophy
Degree Program
Preventive Medicine (Health Behavior Research)
Publication Date
10/26/2020
Defense Date
10/20/2020
Publisher
University of Southern California
(original),
University of Southern California. Libraries
(digital)
Tag
big data,computational social science,data science,e-cigarette,electronic-cigarette,ENDS,machine learning,monitoring,natural language processing,OAI-PMH Harvest,Public Health,social media,substance use,surveillance,tobacco,tobacco regulations,tobacco regulatory science,Twitter,vaping
Language
English
Contributor
Electronically uploaded by the author
(provenance)
Advisor
Pentz, Mary Ann (
committee chair
), Allem, Jon-Patrick (
committee member
), Barberá, Pablo (
committee member
), Cruz, Tess Boley (
committee member
), Unger, Jennifer Beth (
committee member
)
Creator Email
amajmund@usc.edu,majmundar.anuja@gmail.com
Permanent Link (DOI)
https://doi.org/10.25549/usctheses-c89-384936
Unique identifier
UC11666400
Identifier
etd-MajmundarA-9065.pdf (filename),usctheses-c89-384936 (legacy record id)
Legacy Identifier
etd-MajmundarA-9065.pdf
Dmrecord
384936
Document Type
Dissertation
Rights
Majmundar, Anuja
Type
texts
Source
University of Southern California
(contributing entity),
University of Southern California Dissertations and Theses
(collection)
Access Conditions
The author retains rights to his/her dissertation, thesis or other graduate work according to U.S. copyright law. Electronic access is being provided by the USC Libraries in agreement with the a...
Repository Name
University of Southern California Digital Library
Repository Location
USC Digital Library, University of Southern California, University Park Campus MC 2810, 3434 South Grand Avenue, 2nd Floor, Los Angeles, California 90089-2810, USA
Tags
big data
computational social science
data science
e-cigarette
electronic-cigarette
ENDS
machine learning
monitoring
natural language processing
social media
substance use
surveillance
tobacco regulations
tobacco regulatory science
Twitter
vaping