https://orcid.org/0000-0002-7785-3357
-
PDF
- Split View
-
Views
-
Cite
Cite
Anuja Majmundar, Jon-Patrick Allem, Tess Boley Cruz, Jennifer B Unger, Mary Ann Pentz, Twitter Surveillance at the Intersection of the Triangulum, Nicotine & Tobacco Research, Volume 24, Issue 1, January 2022, Pages 118–124, https://doi.org/10.1093/ntr/ntab085
Close - Share Icon Share
Abstract
A holistic public health surveillance approach can help capture the public’s tobacco and marijuana-related attitudes and behaviors. Using publicly available data from Twitter, this is one of the first studies to describe key topics of discussions related to each intersection (e-cigarette, combustible tobacco, and marijuana) of the Triangulum framework.
Twitter posts (n = 999 447) containing marijuana, e-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 were defined as an intersection (e-cigarettes and combustible tobacco, combustible tobacco and marijuana, e-cigarettes and marijuana, and marijuana, e-cigarettes and combustible tobacco). Text classifiers and unsupervised machine learning were used to identify predominant topics in posts.
Product Features and Cartridges were commonly referenced at the intersection of e-cigarette and marijuana-related conversations. Blunts and Cigars and Drugs and Alcohol were commonly referenced at the intersection of combustible tobacco and marijuana-related discussions. Flavors and Health Risks were discussed at the intersection of e-cigarette and combustible-related conversations, while discussions about Illicit products and Health risks were key topics of discussion when e-cigarettes, combustible tobacco, and marijuana were referenced all together in a single post.
By examining intersections of marijuana and tobacco products, this study offers inputs for designing comprehensive FDA regulations including regulating product features associated with appeal, improving enforcement to curb sales of illicit products, and informing the FDA’s product review and standards procedures for tobacco products that can be used with marijuana.
This study is the first to leverage the Triangulum framework and Twitter data to describe key topics of discussions at the intersection of e-cigarette, combustible tobacco, and marijuana. Real-time health communication interventions can identify Twitter users posting in the context of e-cigarettes, combustible tobacco, and marijuana by automated methods and deliver tailored messages. This study also demonstrates the utility of Twitter data for surveillance of complex and evolving health behaviors.
Introduction
Rising electronic cigarette (e-cigarette) and marijuana use in the United States1–3 in combination with ongoing legalization of medical and adult recreational marijuana, and evolving tobacco regulations,4,5 have raised concerns over evolving substance use linked to youth progression from experimentation to nicotine and marijuana dependence.6–9Specifically, dual-use (use of two substances, administered at the same time or not), poly-use (use of more than two substances, administered at the same time or not), and co-use (concurrent use and administration of two or more substances) of marijuana, tobacco, and e-cigarettes, collectively referred to as polysubstance use from here on, are also shown to contribute to abuse liability, earlier onset of smoking initiation, and progression to addictive substances among youths.10–14
Evolving polysubstance use accompanied by the proliferation of new devices has also transformed their use, administration, and appeal.15 Electronic Nicotine Delivery Systems (ENDS) have the ability to deliver a variety of substances to the user, including nicotine, flavored aldehydes, and marijuana (term inclusive of cannabis, tetrahydrocannabinol [THC] and other related terms in this article). For instance, hookah pens deliver aerosolized flavored aldehydes, with or without nicotine; electronic versions of flavored little cigars may contain nicotine; vape pens are used for marijuana vaping; open-system pod mods aerosolize liquid THC and nicotine-containing e-liquids.16–19 Evidence also suggests that youth find novel ways of administering marijuana and nicotine particularly appealing.20
Timely public health surveillance of these polysubstance use behaviors can reveal public health concerns to be addressed in future health interventions. Social media surveillance, in particular, offers opportunities to investigate emerging tobacco and marijuana use behaviors. For example, surveillance of marijuana-related posts on Twitter revealed that mentions of high potency cannabis products, unsubstantiated health claims about cannabis products, and the co-use of cannabis with legal and illicit substances were predominant on Twitter during 2018.21 Another study revealed that vaping-related discussions predominantly involved references to polysubstance use.22 These studies offer unique insights to the public’s recent attitudes and behaviors related to tobacco and marijuana. however, these studies examined tobacco, marijuana, and vaporizers in isolation. A siloed surveillance of tobacco control and marijuana research may only offer limited understanding of polysubstance behaviors.16 As such, a holistic public health surveillance approach, which takes into consideration the intersections of substance use, can help capture the public’s recent attitudes and behaviors pertaining to tobacco and marijuana.
The Triangulum (Latin for triangle), developed by the California Tobacco-Related Disparities Research Program, offers a useful framework to investigate the complexities of intersecting e-cigarettes, marijuana, and combustible tobacco use15 in the context of public health surveillance (eg, polysubstance use among opinion leaders and their followers on social media),23 policy (eg, health impact of marijuana legalizations and smoke-free policies),24 treatment (eg, implications of dual or poly use on cessation),25 and education (eg, health communication initiatives countering pro-marijuana media content).26 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; e-cigarettes, combustible tobacco, and marijuana) using 2018 and 2019 Twitter data. Findings from this study should inform the FDA’s product review and standards procedures for tobacco products that can be used with marijuana.
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 terms, see Table 1. Similar to prior research, retweets, non-English Twitter posts, Twitter posts from bot accounts,27 duplicate posts that were not retweets, spam posts, and promotional posts were excluded from this sample.21,28,29 This resulted in a sample of 999 447 posts to Twitter. Twitter posts from bot accounts were identified with using Botometer (threshold = 0.75), which assigns a score indicating how likely a specific account is to be a bot based on several account characteristics (eg, number of followers, posting frequency).27 Next, the text in the sample was prepared for analysis based on basic normalization (eg, lower case all text, remove special characters), stop word removal (eg, words such as “an,” “the” ), normalization of Twitter user mentions (eg, “@jimjohn” is converted to “@person”), lemmatization (eg, “vaper,” “vaper’s,” and “vapers’” are all converted to “vaper”), and non-printable character removal (eg, emojis and symbols from other languages).21,28,29
| 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” |
| 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” |
| 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” |
| 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 sample was then divided into four analytic subsets based on the Triangulum framework. The breakdown for each subsets was as follows: (1) co-occurring mentions of combustible tobacco-related and e-cigarette-related keywords (N = 456 178; n = 200 914 in 2018 and n = 255 264 in 2019), (2) co-occurring mentions of combustible tobacco- and marijuana-related keywords (n = 188 562), (3) co-occurring mentions of e-cigarette- and marijuana-related keywords (n = 334 409), and (4) co-occurring mentions of combustible tobacco-, e-cigarette- and marijuana-related keywords (n = 20 298).
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 (eg, friends, like, and vaping) and 2 bi-grams (eg, friends-like and 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 (approximately 20 topics per subset) was identified.
Next, GloVe,30 an unsupervised learning algorithm developed by the Stanford Natural Language Processing group, was used to identify words similar to the one-grams and bi-grams identified for each topic (eg, “chill,” “party,” “hangout” can be classified under one category of “socialization”). GloVe offers a pre-trained vector-representations for words (2B tweets, 27B tokens, 1.2M vocab, uncased, 200d vectors). It calculates the Euclidean distance (or cosine similarity) between two-word vectors to measure the linguistic or semantic similarity of the corresponding words (eg, “frog” and “toad”). It also calculates the vector difference between the two-word vectors to capture the 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 be thought of as keywords representing opposite genders. Manual checks of data were conducted to expand the list of synonyms (eg, “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 (Supplementary Figures S1–S4). In other words, the number of posts applicable to any two 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 (12 topics for the combustible tobacco and marijuana subset; 11 topics for the e-cigarette and marijuana subset; 13 topics for the combustible tobacco and e-cigarette subset; 7 topics for the combustible tobacco, e-cigarette, and marijuana subset). Please see Table 1 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 were performed under Institutional Review Board approval from the authors’ university. To protect privacy, no tweets will be reported verbatim.
Results
Table 2 highlights predominant topics of discussion at each intersection of the Triangulum. The following sections highlight findings for each subset.
| Combustible tobacco and marijuana . | . | . | E-cigarette and marijuana . | . | . | E-cigarette and combustible tobacco . | . | . | E-cigarette, combustible tobacco, and marijuana . | . | . |
|---|---|---|---|---|---|---|---|---|---|---|---|
| Topic . | Definition . | N (%) . | Topic . | Definition . | N (%) . | Topic . | Definition . | N (%) . | Topic . | Definition . | N (%) . |
| Person tagging | Mention someone’s Twitter account @Person | 55 698 (31.72%) | Person tagging | Mention someone’s Twitter account @Person | 108 092 (32.32%) | Person tagging | Mention someone’s Twitter account @Person | 142 230 (31.18%) | Person tagging | Mention someone’s Twitter account @Person | 8558 (42.16%) |
| Cigars and blunts | Mentions of cigars and blunts | 46 419 (26.44)% | Product features | Descriptors of products (eg, odorless, purified) | 44 106 (13.19%) | Flavors | Mentions of flavors | 84 652 (18.56%) | Illicit | Mentions of products that are illegal or sold illegally | 4747 (23.39%) |
| Drugs and alcohol | Mentions of drugs and alcohol | 21 513 (12.25%) | Carts | Mentions of THC cartridges and dank vapes | 42 561 (12.73%) | Health risks | Mentions of adverse health outcomes and injuries | 73 531 (16.21%) | Health risks | Mentions of adverse health outcomes | 3677 (18.12%) |
| Legal | Mentions of tobacco laws and regulations (eg, Tobacco 21) | 19 035 (10.84%) | Illicit | Mentions of products that are illegal or sold illegally | 40 182 (12.02%) | Product bans | Mentions of vaping bans or policies, restrictions on sale of tobacco products and minimum age to legally purchase tobacco products | 80 501 (17.65%) | Public agencies | Mentions of federal agencies such as FDA, Centers for Disease Control and Prevention | 1920 (9.46%) |
| Smell | Descriptors of tobacco smell | 12 169 (6.93%) | Promotions | Mentions of sales, discounts, coupon codes | 34 843 (10.42%) | Underage | Mentions of children and adolescents | 73 531 (16.21%) | Appeal | Descriptors of appeal (eg, love, delicious) | 1852 (9.12%) |
| Quitting | Mentions of quitting combustible tobacco | 8903 (5.07%) | New products | Mentions of new products (eg, marijuana tinctures [defined as plant extracts dissolved in ethanol] and supplements) | 32 158 (9.62%) | Buy or Sell | Mentions of product purchase and sales such as price, delivery, refund, retail and modes of payment including venmo | 42 362 (9.29%) | Anti-regulation | Discussions related to unfavorable appraisal of tobacco policies and antitobacco policy advocacy | 1348 (6.64%) |
| Death | Mentions of death to communicate the adverse health effects | 8874 (5.05%) | Stoner imagery | References to “staying high” and dabbing rituals referred to as “shatterday” | 31 823 (9.52%) | Quitting | Mentions of cessation and switching to vape products | 39 472 (8.65%) | Addiction | Mentions of craving, and dependency | 557 (2.74%) |
| Addressing audience | Generally addressing the Twitter audience (eg, “People,” “Y,” “all’) | 8303 (4.73%) | Health risks | Mentions of adverse health outcomes such as death and poisoning | 30 804 (9.21%) | Nicotine | Mentions of nicotine | 35 527 (7.79%) | |||
| Alternative medicine | Mentions of natural, herbal and other forms of medicine | 5850 (3.33%) | Health claims | Mentions of alleviation or management of health issues such as anxiety and pain | 29 941 (8.95%) | Epidemic | Mentions of epidemic of e-cigarette use | 23 217 (5.09%) | |||
| Transit | Mentions of forms of transportation | 3423 (1.95%) | Flavors | Mentions of flavors | 27 739 (8.29%) | Addiction | Mentions of craving, and dependency | 21 704 (4.76%) | |||
| Cigar marketing | Descriptors in cigar marketing (eg,“cigarotica,” “girlswithcigars”) | 3394 (1.93%) | Disposable products | Mentions of disposable products such as Puff Bar | 3757 (1.12%) | School | Mentions of school and sections of schools such as classrooms | 14 673 (3.22%) | |||
| New delivery devices | Mentions of devices such as the volute, stogie | 3394 (1.93%) | Starter kits | Mentions of e-cigarette starter kits | 12 502 (2.74%) | ||||||
| IQOS | Mentions of the product “IQOS,” a heat-not-burn tobacco product | 2997 (0.66%) | |||||||||
| Total | 128 346 (73.10%) | 236 147 (70.62%) | 342 035 (74.98%) | 13 615 (67.08%) |
| Combustible tobacco and marijuana . | . | . | E-cigarette and marijuana . | . | . | E-cigarette and combustible tobacco . | . | . | E-cigarette, combustible tobacco, and marijuana . | . | . |
|---|---|---|---|---|---|---|---|---|---|---|---|
| Topic . | Definition . | N (%) . | Topic . | Definition . | N (%) . | Topic . | Definition . | N (%) . | Topic . | Definition . | N (%) . |
| Person tagging | Mention someone’s Twitter account @Person | 55 698 (31.72%) | Person tagging | Mention someone’s Twitter account @Person | 108 092 (32.32%) | Person tagging | Mention someone’s Twitter account @Person | 142 230 (31.18%) | Person tagging | Mention someone’s Twitter account @Person | 8558 (42.16%) |
| Cigars and blunts | Mentions of cigars and blunts | 46 419 (26.44)% | Product features | Descriptors of products (eg, odorless, purified) | 44 106 (13.19%) | Flavors | Mentions of flavors | 84 652 (18.56%) | Illicit | Mentions of products that are illegal or sold illegally | 4747 (23.39%) |
| Drugs and alcohol | Mentions of drugs and alcohol | 21 513 (12.25%) | Carts | Mentions of THC cartridges and dank vapes | 42 561 (12.73%) | Health risks | Mentions of adverse health outcomes and injuries | 73 531 (16.21%) | Health risks | Mentions of adverse health outcomes | 3677 (18.12%) |
| Legal | Mentions of tobacco laws and regulations (eg, Tobacco 21) | 19 035 (10.84%) | Illicit | Mentions of products that are illegal or sold illegally | 40 182 (12.02%) | Product bans | Mentions of vaping bans or policies, restrictions on sale of tobacco products and minimum age to legally purchase tobacco products | 80 501 (17.65%) | Public agencies | Mentions of federal agencies such as FDA, Centers for Disease Control and Prevention | 1920 (9.46%) |
| Smell | Descriptors of tobacco smell | 12 169 (6.93%) | Promotions | Mentions of sales, discounts, coupon codes | 34 843 (10.42%) | Underage | Mentions of children and adolescents | 73 531 (16.21%) | Appeal | Descriptors of appeal (eg, love, delicious) | 1852 (9.12%) |
| Quitting | Mentions of quitting combustible tobacco | 8903 (5.07%) | New products | Mentions of new products (eg, marijuana tinctures [defined as plant extracts dissolved in ethanol] and supplements) | 32 158 (9.62%) | Buy or Sell | Mentions of product purchase and sales such as price, delivery, refund, retail and modes of payment including venmo | 42 362 (9.29%) | Anti-regulation | Discussions related to unfavorable appraisal of tobacco policies and antitobacco policy advocacy | 1348 (6.64%) |
| Death | Mentions of death to communicate the adverse health effects | 8874 (5.05%) | Stoner imagery | References to “staying high” and dabbing rituals referred to as “shatterday” | 31 823 (9.52%) | Quitting | Mentions of cessation and switching to vape products | 39 472 (8.65%) | Addiction | Mentions of craving, and dependency | 557 (2.74%) |
| Addressing audience | Generally addressing the Twitter audience (eg, “People,” “Y,” “all’) | 8303 (4.73%) | Health risks | Mentions of adverse health outcomes such as death and poisoning | 30 804 (9.21%) | Nicotine | Mentions of nicotine | 35 527 (7.79%) | |||
| Alternative medicine | Mentions of natural, herbal and other forms of medicine | 5850 (3.33%) | Health claims | Mentions of alleviation or management of health issues such as anxiety and pain | 29 941 (8.95%) | Epidemic | Mentions of epidemic of e-cigarette use | 23 217 (5.09%) | |||
| Transit | Mentions of forms of transportation | 3423 (1.95%) | Flavors | Mentions of flavors | 27 739 (8.29%) | Addiction | Mentions of craving, and dependency | 21 704 (4.76%) | |||
| Cigar marketing | Descriptors in cigar marketing (eg,“cigarotica,” “girlswithcigars”) | 3394 (1.93%) | Disposable products | Mentions of disposable products such as Puff Bar | 3757 (1.12%) | School | Mentions of school and sections of schools such as classrooms | 14 673 (3.22%) | |||
| New delivery devices | Mentions of devices such as the volute, stogie | 3394 (1.93%) | Starter kits | Mentions of e-cigarette starter kits | 12 502 (2.74%) | ||||||
| IQOS | Mentions of the product “IQOS,” a heat-not-burn tobacco product | 2997 (0.66%) | |||||||||
| Total | 128 346 (73.10%) | 236 147 (70.62%) | 342 035 (74.98%) | 13 615 (67.08%) |
| Combustible tobacco and marijuana . | . | . | E-cigarette and marijuana . | . | . | E-cigarette and combustible tobacco . | . | . | E-cigarette, combustible tobacco, and marijuana . | . | . |
|---|---|---|---|---|---|---|---|---|---|---|---|
| Topic . | Definition . | N (%) . | Topic . | Definition . | N (%) . | Topic . | Definition . | N (%) . | Topic . | Definition . | N (%) . |
| Person tagging | Mention someone’s Twitter account @Person | 55 698 (31.72%) | Person tagging | Mention someone’s Twitter account @Person | 108 092 (32.32%) | Person tagging | Mention someone’s Twitter account @Person | 142 230 (31.18%) | Person tagging | Mention someone’s Twitter account @Person | 8558 (42.16%) |
| Cigars and blunts | Mentions of cigars and blunts | 46 419 (26.44)% | Product features | Descriptors of products (eg, odorless, purified) | 44 106 (13.19%) | Flavors | Mentions of flavors | 84 652 (18.56%) | Illicit | Mentions of products that are illegal or sold illegally | 4747 (23.39%) |
| Drugs and alcohol | Mentions of drugs and alcohol | 21 513 (12.25%) | Carts | Mentions of THC cartridges and dank vapes | 42 561 (12.73%) | Health risks | Mentions of adverse health outcomes and injuries | 73 531 (16.21%) | Health risks | Mentions of adverse health outcomes | 3677 (18.12%) |
| Legal | Mentions of tobacco laws and regulations (eg, Tobacco 21) | 19 035 (10.84%) | Illicit | Mentions of products that are illegal or sold illegally | 40 182 (12.02%) | Product bans | Mentions of vaping bans or policies, restrictions on sale of tobacco products and minimum age to legally purchase tobacco products | 80 501 (17.65%) | Public agencies | Mentions of federal agencies such as FDA, Centers for Disease Control and Prevention | 1920 (9.46%) |
| Smell | Descriptors of tobacco smell | 12 169 (6.93%) | Promotions | Mentions of sales, discounts, coupon codes | 34 843 (10.42%) | Underage | Mentions of children and adolescents | 73 531 (16.21%) | Appeal | Descriptors of appeal (eg, love, delicious) | 1852 (9.12%) |
| Quitting | Mentions of quitting combustible tobacco | 8903 (5.07%) | New products | Mentions of new products (eg, marijuana tinctures [defined as plant extracts dissolved in ethanol] and supplements) | 32 158 (9.62%) | Buy or Sell | Mentions of product purchase and sales such as price, delivery, refund, retail and modes of payment including venmo | 42 362 (9.29%) | Anti-regulation | Discussions related to unfavorable appraisal of tobacco policies and antitobacco policy advocacy | 1348 (6.64%) |
| Death | Mentions of death to communicate the adverse health effects | 8874 (5.05%) | Stoner imagery | References to “staying high” and dabbing rituals referred to as “shatterday” | 31 823 (9.52%) | Quitting | Mentions of cessation and switching to vape products | 39 472 (8.65%) | Addiction | Mentions of craving, and dependency | 557 (2.74%) |
| Addressing audience | Generally addressing the Twitter audience (eg, “People,” “Y,” “all’) | 8303 (4.73%) | Health risks | Mentions of adverse health outcomes such as death and poisoning | 30 804 (9.21%) | Nicotine | Mentions of nicotine | 35 527 (7.79%) | |||
| Alternative medicine | Mentions of natural, herbal and other forms of medicine | 5850 (3.33%) | Health claims | Mentions of alleviation or management of health issues such as anxiety and pain | 29 941 (8.95%) | Epidemic | Mentions of epidemic of e-cigarette use | 23 217 (5.09%) | |||
| Transit | Mentions of forms of transportation | 3423 (1.95%) | Flavors | Mentions of flavors | 27 739 (8.29%) | Addiction | Mentions of craving, and dependency | 21 704 (4.76%) | |||
| Cigar marketing | Descriptors in cigar marketing (eg,“cigarotica,” “girlswithcigars”) | 3394 (1.93%) | Disposable products | Mentions of disposable products such as Puff Bar | 3757 (1.12%) | School | Mentions of school and sections of schools such as classrooms | 14 673 (3.22%) | |||
| New delivery devices | Mentions of devices such as the volute, stogie | 3394 (1.93%) | Starter kits | Mentions of e-cigarette starter kits | 12 502 (2.74%) | ||||||
| IQOS | Mentions of the product “IQOS,” a heat-not-burn tobacco product | 2997 (0.66%) | |||||||||
| Total | 128 346 (73.10%) | 236 147 (70.62%) | 342 035 (74.98%) | 13 615 (67.08%) |
| Combustible tobacco and marijuana . | . | . | E-cigarette and marijuana . | . | . | E-cigarette and combustible tobacco . | . | . | E-cigarette, combustible tobacco, and marijuana . | . | . |
|---|---|---|---|---|---|---|---|---|---|---|---|
| Topic . | Definition . | N (%) . | Topic . | Definition . | N (%) . | Topic . | Definition . | N (%) . | Topic . | Definition . | N (%) . |
| Person tagging | Mention someone’s Twitter account @Person | 55 698 (31.72%) | Person tagging | Mention someone’s Twitter account @Person | 108 092 (32.32%) | Person tagging | Mention someone’s Twitter account @Person | 142 230 (31.18%) | Person tagging | Mention someone’s Twitter account @Person | 8558 (42.16%) |
| Cigars and blunts | Mentions of cigars and blunts | 46 419 (26.44)% | Product features | Descriptors of products (eg, odorless, purified) | 44 106 (13.19%) | Flavors | Mentions of flavors | 84 652 (18.56%) | Illicit | Mentions of products that are illegal or sold illegally | 4747 (23.39%) |
| Drugs and alcohol | Mentions of drugs and alcohol | 21 513 (12.25%) | Carts | Mentions of THC cartridges and dank vapes | 42 561 (12.73%) | Health risks | Mentions of adverse health outcomes and injuries | 73 531 (16.21%) | Health risks | Mentions of adverse health outcomes | 3677 (18.12%) |
| Legal | Mentions of tobacco laws and regulations (eg, Tobacco 21) | 19 035 (10.84%) | Illicit | Mentions of products that are illegal or sold illegally | 40 182 (12.02%) | Product bans | Mentions of vaping bans or policies, restrictions on sale of tobacco products and minimum age to legally purchase tobacco products | 80 501 (17.65%) | Public agencies | Mentions of federal agencies such as FDA, Centers for Disease Control and Prevention | 1920 (9.46%) |
| Smell | Descriptors of tobacco smell | 12 169 (6.93%) | Promotions | Mentions of sales, discounts, coupon codes | 34 843 (10.42%) | Underage | Mentions of children and adolescents | 73 531 (16.21%) | Appeal | Descriptors of appeal (eg, love, delicious) | 1852 (9.12%) |
| Quitting | Mentions of quitting combustible tobacco | 8903 (5.07%) | New products | Mentions of new products (eg, marijuana tinctures [defined as plant extracts dissolved in ethanol] and supplements) | 32 158 (9.62%) | Buy or Sell | Mentions of product purchase and sales such as price, delivery, refund, retail and modes of payment including venmo | 42 362 (9.29%) | Anti-regulation | Discussions related to unfavorable appraisal of tobacco policies and antitobacco policy advocacy | 1348 (6.64%) |
| Death | Mentions of death to communicate the adverse health effects | 8874 (5.05%) | Stoner imagery | References to “staying high” and dabbing rituals referred to as “shatterday” | 31 823 (9.52%) | Quitting | Mentions of cessation and switching to vape products | 39 472 (8.65%) | Addiction | Mentions of craving, and dependency | 557 (2.74%) |
| Addressing audience | Generally addressing the Twitter audience (eg, “People,” “Y,” “all’) | 8303 (4.73%) | Health risks | Mentions of adverse health outcomes such as death and poisoning | 30 804 (9.21%) | Nicotine | Mentions of nicotine | 35 527 (7.79%) | |||
| Alternative medicine | Mentions of natural, herbal and other forms of medicine | 5850 (3.33%) | Health claims | Mentions of alleviation or management of health issues such as anxiety and pain | 29 941 (8.95%) | Epidemic | Mentions of epidemic of e-cigarette use | 23 217 (5.09%) | |||
| Transit | Mentions of forms of transportation | 3423 (1.95%) | Flavors | Mentions of flavors | 27 739 (8.29%) | Addiction | Mentions of craving, and dependency | 21 704 (4.76%) | |||
| Cigar marketing | Descriptors in cigar marketing (eg,“cigarotica,” “girlswithcigars”) | 3394 (1.93%) | Disposable products | Mentions of disposable products such as Puff Bar | 3757 (1.12%) | School | Mentions of school and sections of schools such as classrooms | 14 673 (3.22%) | |||
| New delivery devices | Mentions of devices such as the volute, stogie | 3394 (1.93%) | Starter kits | Mentions of e-cigarette starter kits | 12 502 (2.74%) | ||||||
| IQOS | Mentions of the product “IQOS,” a heat-not-burn tobacco product | 2997 (0.66%) | |||||||||
| Total | 128 346 (73.10%) | 236 147 (70.62%) | 342 035 (74.98%) | 13 615 (67.08%) |
Combustible Tobacco and Marijuana
The total coverage of the 12 topics identified at this intersection constituted 73.1% of all tweets in the subset. The remaining tweets (26.9%) were too varied to be classified into one topic with meaningful coverage (ie, meaningful coverage defined as <1% of total tweets). Person Tagging (31.7%) was the most predominant topic followed by discussions about Cigars and Blunts (26.4%), references to Drugs and Alcohol (12.2%), and Legal aspects of marijuana and combustible use was about 10.8% of the sample. Smell of products (6.9%), Death (5.0%), and Alternative Medicine (3.3%) were also notable. The least common topics included New delivery devices (1.9%) such as volute, stogie, and waffle pipes, and descriptors used in Cigar Marketing (1.9%) such as “girlswithcigars,” and “cigarfam.” Please refer to Supplementary Figure S1 for more details on prevalence of overlapping topics of discussion.
E-cigarette and Marijuana
The total coverage of the 11 topics identified at this intersection constituted 70.6% of all tweets in the subset. Person Tagging (mentioning someone’s Twitter account in a Twitter post; 32.3%) was the most predominant topic followed by discussions about Product Features (13.2%) and mentions of Carts (or cartridges; 12.7%). Other topics included mentions of Illicit products (12.0%), Promotions (10.4%), New Products (9.6%), and Stoner Imagery (9.5%). Harm Perceptions (9.2%), Health Claims (8.9%), and Disposable products were the least predominant in the sample (1.1%). Please refer to Supplementary Figure S2 for more details on prevalence of overlapping topics of discussion.
E-cigarettes and Combustible Tobacco
A total of 13 topics at the intersection of e-cigarettes and combustible tobacco constituted 75% of all tweets in the subset. Person Tagging (32.3%) was the most predominant topic. Flavors were the second-most predominant topic (18.6%) followed by discussions about recent vaping Bans (17.6%). Health Risks (16.1%), Underage (14.6%), Buy or Sell (9.3%), Quitting (8.6%), followed by references to Nicotine (7.8%). References to the vaping Epidemic (5.1%), Addiction (4.8%), School (3.2%), and vape Starter Kits (2.7%). IQOS was the least predominant topic (0.7%). Please refer to Supplementary Figure S3 for more details on prevalence of overlapping topics of discussion.
E-cigarettes, Marijuana, and Combustible Tobacco
A total of 7 topics at this intersection constituted 67.1% of all tweets in the subset. Similar to the above intersections, Person Tagging was the most predominant topic (42.2%), followed by discussions about Illicit products on the market (23.4%), Health Risks (18.1%), references to Public Agencies (9.5%), Appeal (9.1%), Anti-Regulation (6.6%), and references to Addiction (2.7%). Please refer to Supplementary Figure S4 for more details on prevalence of overlapping topics of discussion.
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 2 years. A number of different topics were identified at each intersection of the Triangulum, with Person Tagging (mentioning a Twitter user’s account name) found most commonly at each intersection. Other predominant topics ranged from Flavors, Nicotine, Health Claims, New Products, to Anti-Regulation. 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; Flavors among e-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.28,31 As discussed in prior work,21 Person Tagging in online messages brings 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 tagging peers and posting about tobacco by automated methods and deliver tailored messages pertaining to relevant intersections of the Triangulum on social media platforms. Recent example efforts in this area include the use of recommendation algorithms to deliver tailored health messages,32 and real-time, targeted delivery of tailored health messages.33 Such interventions can enhance public awareness about the unique health risks associated with each intersection of the Triangulum. For instance, health messages may highlight emerging findings suggesting that co-use of marijuana and tobacco poses additive risk of toxicant exposure.34
Cigars and Blunts were the predominant topic of conversation related to combustible tobacco and marijuana. This finding is in line with previous research, highlighting correlates of cigar use with marijuana.35 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.36 Coordinated tobacco control and substance abuse prevention efforts may help address this trend, including making it more difficult to modify cigar wrappers for blunt use. References to Alcohol and Drugs raise concerns about polysubstance use. Past work suggests that dual alcohol and blunt use are on the rise, especially among African American youth.37
Leveraging discussions about the undesirable Smell of products, and adverse health outcomes such as Death in health interventions may play a crucial role in changing norms about cigars, the third most popular tobacco product among adolescents.38 Additionally, communication efforts addressing cessation attempts 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 suggest that vaping is sometimes marketed as a health-enhancing behavior39 capable of delivering vitamins while vape products are in use.40 Current findings add to the literature by highlighting that conversations about product characteristic are generally 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 to determine optimal messaging strategies such as two-sided messages to counter such effects.41 The discussions we observed about vape Cartridges and mentions of Illicit substances are potentially situated in the context of an expanding range of illicit products, marijuana legalization, widespread marijuana use, and availability of new generation products such as open-system pod mods.42,43 Marijuana-containing products that can be vaped need to be subjected to safety and quality testing and appropriate labeling to address these concerns.44
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 with those that vape non-flavored products, especially among young adults.45,46 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.47 Other predominant topics including Health Risks, recent Product Bans, and Underage or legal restrictions on the sale of tobacco products and introduction of a federal minimum age to legally purchase tobacco products, highlight public attention to 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 Food and Drug Administration (FDA).
Conversations mentioning e-cigarettes, combustible tobacco and marijuana mostly centered around public concerns related to Illicit products, Health Risks and Addiction. Appeal and Anti-tobacco Regulation-related topics may create favorable norms for initiation or continued polysubstance use. Past work has highlighted ways in which appeal manifests on social media platforms. For instance, YouTube videos are predominantly pro-marijuana-vaping and pro-vaping,48 devices such as open-system pod mods are promoted extensively on Instagram,19 and individuals engage extensively in sharing their product use experiences by sharing pictures of vaporizers and nontraditional forms of administering marijuana.49 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 upon specified keywords of interest (marijuana, combustible tobacco, e-cigarettes). Rule-based classifiers also relied on a representative set of keywords. While every effort was made to create a comprehensive set of keywords for data collection and data classification, data may not be completely representative of the entire area of interest or each of the topics. Currently, there are regional variations in tobacco and marijuana regulations.50 States such as California support tobacco restrictions that are often stricter than other states, but legal recreational and medical marijuana policies that permit types of sales and use not found in all states. Such variations may influence the prevalence of themes on Twitter and findings may not be representative of such state-wide differences.
Conclusion
Person Tagging and discussions about illicit products in the market were key topics at the intersection of e-cigarettes, combustible tobacco, and marijuana. Discussions about Product Features and Cartridges were predominant when both e-cigarette and marijuana were referenced in a single Twitter post. Blunts and Cigars and Drugs and Alcohol were commonly referenced when both combustible tobacco and marijuana were referenced in a single post, suggesting concerns about polysubstance use among Twitter users. Flavors and Health Risks were predominant among posts referencing both e-cigarette and combustible-related conversations. Discussions about Illicit products and Health Risks were predominant when e-cigarettes, combustible tobacco, and marijuana were referenced all together. By examining intersections of marijuana and other tobacco products, this study offers insights of emerging public experience at each intersection of the Triangulum. Real-time health communication interventions can identify Twitter users posting in the context of e-cigarettes, combustible tobacco, and marijuana by automated methods and deliver tailored messages. This study also demonstrates the utility of Twitter data for surveillance of complex and evolving health behaviors.
Supplementary Material
A Contributorship Form detailing each author’s specific involvement with this content, as well as any supplementary data, are available online at https://academic.oup.com/ntr.
Funding
This work was supported by the National Institutes of Health, grant U54CA180905.
Declaration of Interests
None declared.
Acknowledgements
The authors would like to thank Pablo Barberá, Department of Political Science and International Relations at the University of Southern California, for his valuable inputs.
The first author performed this research as part of her doctoral dissertation research51 at the Department of Preventive Medicine, Keck School of Medicine, University of Southern California.
Comments