The reasons for using electronic nicotine delivery systems (ENDS) are poorly understood and are primarily documented by expensive cross-sectional surveys that use preconceived close-ended response options rather than allowing respondents to use their own words. We passively identify the reasons for using ENDS longitudinally from a content analysis of public postings on Twitter. All English language public tweets including several ENDS terms (e.g., “e-cigarette” or “vape”) were captured from the Twitter data stream during 2012 and 2015. After excluding spam, advertisements, and retweets, posts indicating a rationale for vaping were retained. The specific reasons for vaping were then inferred based on a supervised content analysis using annotators from Amazon’s Mechanical Turk. During 2012 quitting combustibles was the most cited reason for using ENDS with 43% (95%CI 39–48) of all reason-related tweets cited quitting combustibles, e.g., “I couldn’t quit till I tried ecigs,” eclipsing the second most cited reason by more than double. Other frequently cited reasons in 2012 included ENDS’s social image (21%; 95%CI 18–25), use indoors (14%; 95%CI 11–17), flavors (14%; 95%CI 11–17), safety relative to combustibles (9%; 95%CI 7–11), cost (3%; 95%CI 2–5) and favorable odor (2%; 95%CI 1–3). By 2015 the reasons for using ENDS cited on Twitter had shifted. Both quitting combustibles and use indoors significantly declined in mentions to 29% (95%CI 24–33) and 12% (95%CI 9–16), respectively. At the same time, social image increased to 37% (95%CI 32–43) and lack of odor increased to 5% (95%CI 2–5), the former leading all cited reasons in 2015. Our data suggest the reasons people vape are shifting away from cessation and toward social image. The data also show how the ENDS market is responsive to a changing policy landscape. For instance, smoking indoors was less frequently cited in 2015 as indoor smoking restrictions became more common. Because the data and analytic approach are scalable, adoption of our strategies in the field can inform follow-up survey-based surveillance (so the right questions are asked), interventions, and policies for ENDS.
Citation: Ayers JW, Leas EC, Allem J-P, Benton A, Dredze M, Althouse BM, et al. (2017) Why do people use electronic nicotine delivery systems (electronic cigarettes)? A content analysis of Twitter, 2012-2015. PLoS ONE 12(3): e0170702. https://doi.org/10.1371/journal.pone.0170702
Editor: Donald R. Olson, New York City Department of Health and Mental Hygiene, UNITED STATES
Received: July 18, 2016; Accepted: January 9, 2017; Published: March 1, 2017
Copyright: © 2017 Ayers et al. This is an open access article distributed under the terms of the Creative Commons Attribution License, which permits unrestricted use, distribution, and reproduction in any medium, provided the original author and source are credited.
Data Availability: All data are publicly available. Twitter postings can accessed using the public API (dev.twitter.com/streaming/overview). Please contact the corresponding author if further assistance is required.
Funding: Research reported in this publication was supported by the National Cancer Institute and the FDA Center for Tobacco Products P50CA180905. The funders had no role in the design, conduct, or interpretation of the study; nor the preparation, review, or approval of the manuscript. The content is solely the responsibility of the authors and does not necessarily represent the official views of the NIH or FDA. Bloomberg LP provided support in the form of salaries for authors MD, but did not have any additional role in the study design, data collection and analysis, decision to publish, or preparation of the manuscript.
Competing interests: MD is commercially employed by Bloomberg LP. There are no other relevant competing interests. This does not alter our adherence to PLOS ONE policies on sharing data and materials.
Despite the popularity of electronic cigarettes or electronic nicotine delivery systems (ENDS) [1,2], there is surprisingly little actionable intelligence on why people vape . The most cited reasons include curiosity, enjoyment, and other idiopathic reasons [4–6] that do not inform any particular intervention or policy response. Fewer studies point to actionable reasons that may guide specific control measures to curtail use, such as use indoors  which can be mitigated by banning vaping indoors.
This knowledge gap, in part, represents the limitations of tobacco control surveillance. All of the known reasons for vaping were derived from cross-sectional surveys. For example, the International Tobacco Control Four-Country Survey asked vapers about why they use ENDS with four yes/no response items  Surveys require substantial time and resources to implement, meaning many reasons are never asked. Moreover, where studies rely on closed-ended response options (rather than allowing participants to use their own words) some reasons are potentially still unknown.
Supplemental approaches are needed to inexpensively and rapidly discover the reasons for using ENDS to inform the science around ENDS (such as follow-up survey-based surveillance), inform the development of policy-based control measures, or aid the design of health interventions. Herein we demonstrate the feasibility of a data-driven protocol that allows the public to describe why they vape in their own words by passively monitoring public tweets—a promising  but underutilized approach in public health [10–16]. Doing so lays the groundwork for a new perspective that in practice can help capture the reasons people vape quickly, informing follow-up surveillance and public health practice.
Materials and methods
The data consisted of 3.3 million public tweets from 2012 and 2015 that was collected from the Twitter API by searching for the following ENDS-related keywords: electronic cigarette(s), electronic cig(s), e cig(s), e-cig(s), eking(s), e cigarette(s), e-cigarette(s), ecigarette(s), vape(s), vaper(s), and vaping. This data collection therefore includes all tweets about ENDS as long as they included the aforementioned terms. We then used a two-stage strategy to identify a subsample for analysis by (a) selecting organic English-language tweets that referenced ENDS use and then (b) using supervised content analysis to discover the reasons for using ENDS from these tweets.
In the first stage, irrelevant tweets were excluded by purging non-English language tweets (using the Lui method ) tweets with URLS (which were almost exclusively advertisements), spam, and retweets (so each tweet counted once). We identified spam tweets using a statistical machine learning classifier developed using a set of 10,157 e-cigarette tweets (60% train, 20% dev, 20% test) that were identified by Amazon Mechanical Turk annotators (mturk.com/mturk/welcome) as either spam or not spam. We used a logistic regression model to predict labels, with a set of n-gram features augmented by 300-dimensional word2vec embeddings using the Mikolov strategy . Drawing from this refined database we selected tweets that indicated use ENDS by the tweeter or another person using human coders from Amazon’s Mechanical Turk . Coders searched for retained tweets about ENDS use until a target sample of 2,900 for each year was achieved (thereby yielding a margin of error less than 0.02 in subsequent analysis) . In practice this collective strategy retained tweets such as “I have an electronic cig and it’s helping me quit,” and excluded tweets such as “Closing sale! #vapeporn #eking #vaping” “just read this eking article [link],” or “I just saw someone vaping” because they are advertising, included a URL, or do not indicate use, respectively. The inter-rater agreement between the Mturk coders identifying tweets as being about ENDS use was Cohen’s Kappa = 0.56, percentage agreement = 0.83. Tweets with conflicting labels were assumed to not be about ENDS use and discarded from further analysis.
In the second stage, we identified any “reasons” for use . The investigators (JWA, ECL, AB, and MD) reviewed tweets and discussed reasons for vaping that emerged in the data simultaneously developing a framework and codebook for annotating the tweets. An open-ended framework was selected that allowed each tweet to have zero or multiple associated reasons; e.g., “I like ecigs because they’re cheap and taste great” would indicate lower price and flavor as reasons (Table 1). Implied reasons were also considered, e.g., “vaping in the club” would indicate vaping indoors. We then developed a protocol for the most commonly cited reasons, avoiding rare reasons because they are not high priorities for intervention or reasons focused on personal emotions because they may not inform precise interventions. The investigators then tested this protocol until a final document was agreed on. Ultimately, our data-driven pilot protocol defined 7 categories: cost, flavors, odor, safety relative to combustibles, social image, quitting combustibles, and use indoors. This protocol was then applied using Amazon’s Mechanical Turk (Cohen’s Kappa = 0.54 averaged over all 7 reasons; percentage agreement = 0.92). The results for the 7 reasons as the percent of reason-related tweets by year were described using bootstrapped 95% confidence intervals (to assess confidence in the point estimates) and non-parametric chi-squared tests (to compare differences in reasons across years).
During 2012 quitting combustibles was the most cited reason for using ENDS. Forty-three percent 43% (95%CI 39–48) of reason-related-tweets mentioned quitting combustibles, e.g., “I couldn’t quit till I tried ecigs”, eclipsing all other reasons by more than double (Fig 1).
Each bar shows the percentage of reason-related tweets for the focal reason by year with bootstrapped 95% confidence intervals and non-parametric chi-squared test results between years shown where statistically significant.
The remaining reasons fell into 3 strata by rank. Ranking second, third, and fourth, and significantly higher than all remaining reasons was social image which was cited in 21% (95%CI 18–25) of reason-related-tweets, e.g., “I want one of those e-cigs, it'll make me look cool.” use indoors, e.g., “vaping in the club” which was cited in 17% (95%CI 14–20) of reason-related tweets, and flavor choices, e.g., “…the best part is the taste,” which was cited in 14% (95%CI 11–17) of reason-related tweets. Ranking fifth and significantly higher than all remaining reasons, safety relative to combustibles, e.g., “…and vaping is safe,” was cited in 9% (95%CI 7–11) of reason-related tweets. The final strata included ENDS’ favorable odor and low cost ranking sixth and seventh with 3% (95%CI 2–5) and 2% (95%CI 1–3), respectively.
By 2015 both quitting combustibles and use indoors significantly declined to 29% (95%CI 24–33) and 12% (95%CI 9–16) falling from first to second in the rankings and third to fourth in the rankings. At the same time, social image increased to 37% (95%CI 32–43) and favorable odor increased to 5% (95%CI 2–5), the former eclipsing all reasons in 2015 and the latter rising from seventh to sixth in the ranking of reasons reported on Twitter. Other reasons remained stable from 2012 to 2015.
Without any priming or direct costs associated with data collection, public health can use social media surveillance to understand why people vape, yielding actionable intelligence for decision making regarding ENDS now and a pathway forward for additional intelligence using our novel strategy in the future.
Our findings regarding 2012 confirmed with traditional studies; including using ENDS to quit combustibles and use indoors [22,23]. But by 2015 the reasons for vaping as reported on Twitter shifted, with both quitting combustibles and vaping indoors declining in mentions, a finding that has not been reported elsewhere. Taken together with other more recent findings and anecdotal evidence, the shift in reported reasons for vaping appears face valid. Google searches for ENDS for quitting smoking have been on the decline  supporting our conclusion that cessation is declining as a reason for vaping. Nearly 500 legislative bodies now ban vaping where smoking is prohibited  meaning ENDS cannot be used to avoid clean air laws as before. Similarly ENDS marketing has substantially grown with a focus on social image [25–27] consequently this may be why positive social image dominated all reported reasons on Twitter in 2015 .
At the same time, our findings suggest the reasons for using ENDS is in part to circumvent existing policy regulations for controlling combustible tobacco use. Three of the 7 most cited reasons for vaping focused on evading policies, such as novel flavors (e.g., cherry) that have been banned in cigarettes . The appeal of ENDS can potentially be curtailed by targeting these reasons with public policies that ban the use of non-tobacco flavors in ENDS, apply clean indoor air laws to ENDS, and tax ENDS like combustible tobacco products. Moreover, given these reasons were cited in about 30% of tweets during 2015 a change in public policy could have substantial impact on the appeal of ENDS.
The most important implication is the long-term value of our strategy for open-ended and real-time surveillance, and how it resembles a massive and passive focus group. For example, assuming 2% of adults use ENDS and a sample of 1,000 participants using random-digit dialing costs roughly $70,000, 50,000 interviews would need to be completed to have a single comparable snapshot—a fiftyfold increase in cost representing about $3.5 million. This does not mean our approach replaces surveys, but it can inform their design so that surveys are asking “the right” questions about reasons that are known to resonate with the public and public health leaders have an agile and replicable early surveillance system to guide debate.
Big data has already played a central role in ENDS surveillance, [2,12,30,31] even first identifying their popularity explosion , and has similarly improved tobacco control surveillance generally [32–38]. Yet, most studies using Twitter are superficial, describing a general trend for keyword searches like “cholera”  or “quit smoking” , as detailed in several recent critiques [41–44]. Herein we deviate from the existing literature and demonstrate a protocol that renders specific data for a specific, but poorly addressed, high priority research question that takes advantage of the richness in Twitter data by going beyond keyword-based analyses. Moreover, our strategy has implications beyond ENDS or tobacco control, holding value for studying the appeal of medical devices, public policies, illicit drugs, and commercial products.
In our demonstrative study, we constrained our focus to the most common reasons and analyzed all tweets in aggregate, but in the future this procedure can be refined to overcome these limitations. For example, our strategy can discover additional reasons or variations within reasons (e.g., what are the most popular flavors?). Moreover, reasons can be described across demographic traits that can be inferred from a Twitter profile, such as gender, ethnicity, and location, yielding demographically specific insights akin to the traditional survey. This potential along with the empirical insights herein suggests our protocol holds great value for ENDS, tobacco control, and public health surveillance going forward.
Research reported in this publication was supported by the National Cancer Institute and the FDA Center for Tobacco Products P50CA180905. The funders had no role in the design, conduct, or interpretation of the study; nor the preparation, review, or approval of the manuscript. The content is solely the responsibility of the authors and does not necessarily represent the official views of the NIH or FDA. This work was previously shared with James Fowler (UCSD), Seth Noar (UNC), and Lee Westmaas (ACS) and improved by presentations at the University of California San Diego and San Diego State University.
- Conceptualization: JWA.
- Formal analysis: JWA AB MD BMA.
- Methodology: JWA ECL AB MD.
- Software: JWA AB MD.
- Writing – original draft: JWA ECL JPA AB MD.
- Writing – review & editing: JWA ECL JPA AB MD BMA TBC JBU.
- 1. Ayers JW, Ribisl KM, and Brownstein JS. Tracking the rise in popularity of electronic nicotine delivery systems (electronic cigarettes) using search query surveillance. Am J Prev Med. Netherlands; 2011;40(4):448–53.
- 2. Ayers JW, Althouse BM, Allem JP, Leas EC, Dredze M, and Williams RS. Revisiting the Rise of Electronic Nicotine Delivery Systems Using Search Query Surveillance. Am J Prev Med. 2016;50:e173–e181. pmid:26876772
- 3. Grana R, Benowitz N, Glantz SA. E-cigarettes: a scientific review. Circulation. 2014;129:1972–86. pmid:24821826
- 4. Pepper JK, Ribisl KM, Emery SL, and Brewer NT. Reasons for starting and stopping electronic cigarette use. Int J Environ Res Public Health. 2014;11(10):10345–61. pmid:25286168
- 5. Saddleson ML, Kozlowski LT, Giovino GA, Goniewicz ML, Mahoney MC, Homish GG, and Arora A. Enjoyment and other reasons for electronic cigarette use: Results from college students in New York. Addict Behav. 2016;54:33–9. pmid:26704429
- 6. Soule EK, Rosas SR, and Nasim A. Reasons for electronic cigarette use beyond cigarette smoking cessation: A concept mapping approach. Addict Behav. 2016;56:41–50. pmid:26803400
- 7. Czogala J, Goniewicz ML, Fidelus B, Zielinska-Danch W, Travers MJ, and Sobczak A. Secondhand exposure to vapors from electronic cigarettes. Nicotine Tob Res. 2014;16(6):655–62. pmid:24336346
- 8. Adkison SE, O'Connor RJ, Bansal-Travers M, Hyland A, Borland R, Yong HH, Cummings KM, McNeill A, Thrasher JF, Hammond D, and Fong GT. Electronic nicotine delivery systems: international tobacco control four-country survey. Am J Prev Med. 2013;44(3):207–15. pmid:23415116
- 9. Ayers JW, Althouse BM, and Dredze M. Could Behavioral Medicine Lead the Web Data Revolution? JAMA. 2014;311(14):1399–400. pmid:24577162
- 10. Ayers JW, Westmaas JL, Leas EC, Benton A, Chen Y, Dredze M, and Althouse BM. Leveraging Big Data to Improve Health Awareness Campaigns: A Novel Evaluation of the Great American Smokeout. JMIR Public Health and Surveillance. 2016;2(1):e16. pmid:27227151
- 11. Kavuluru R, and Sabbir AK. Toward automated e-cigarette surveillance: Spotting e-cigarette proponents on Twitter. J Biomed Inform. 2016;61:19–26. pmid:26975599
- 12. Allem J-P, Escobedo P, Chu K-H, Soto DW, Cruz TB, and Unger JB. Campaigns and counter campaigns: reactions on Twitter to e-cigarette education. Tobacco control. 2016.
- 13. Chu KH, Unger JB, Allem JP, Pattarroyo M, Soto D, Cruz TB, Yang H, Jiang L, and Yang CC. Diffusion of Messages from an Electronic Cigarette Brand to Potential Users through Twitter. PLoS One. United States; 2015;10(12):e0145387.
- 14. Huang J, Kornfield R, Szczypka G, and Emery SL. A cross-sectional examination of marketing of electronic cigarettes on Twitter. Tob Control. England; 2014;23 Suppl 3:iii26–30.
- 15. Myslín M, Zhu SH, Chapman W, and Conway M. Using twitter to examine smoking behavior and perceptions of emerging tobacco products. J Med Internet Res. Canada; 2013;15(8):e174.
- 16. Paul MJ, and Dredze M. A model for mining public health topics from Twitter. HEALTH. 2012;11:16–6.
- 17. Lui M, and Baldwin T. langid. py: An off-the-shelf language identification tool. Proceedings of the ACL 2012 System Demonstrations. 2012:25–30.
- 18. Mikolov T, Sutskever I, Chen K, Corrado GS, and Dean J. Distributed representations of words and phrases and their compositionality. Advances in neural information processing systems. 2013:3111–3119.
- 19. Callison-Burch C, and Dredze M. Creating speech and language data with Amazon's Mechanical Turk. Proceedings of the NAACL HLT 2010 Workshop on Creating Speech and Language Data with Amazon's Mechanical Turk. 2010:1–12.
- 20. Kotrlik J, and Higgins C. Organizational research: Determining appropriate sample size in survey research appropriate sample size in survey research. Information technology, learning, and performance journal. 2001;19(1):43.
- 21. Krippendorff K. Content analysis: An introduction to its methodology. Content analysis: An introduction to its methodology. Sage; 2012.
- 22. Siegel MB, Tanwar KL, and Wood KS. Electronic cigarettes as a smoking-cessation: tool results from an online survey. Am J Prev Med. 2011;40(4):472–5. pmid:21406283
- 23. Grana RA, and Ling PM. "Smoking revolution": a content analysis of electronic cigarette retail websites. Am J Prev Med. 2014;46(4):395–403. pmid:24650842
- 24. American NonSmokes’ Rights Foundation. States and municipalities with laws regulating use of electronic cigarettes. 2016. http://www.no-smoke.org/pdf/ecigslaws.pdf
- 25. de Andrade M, Hastings G, and Angus K. Promotion of electronic cigarettes: tobacco marketing reinvented? BMJ. 2013;347:f7473. pmid:24361526
- 26. Kim AE, Arnold KY, and Makarenko O. E-cigarette advertising expenditures in the U.S., 2011–2012. Am J Prev Med. 2014;46(4):409–12. pmid:24650844
- 27. Chu KH, Allem JP, Cruz TB, Unger JB Vaping on Instagram: cloud chasing, hand checks and product placement. Tob Control. 2016.
- 28. Fairchild AL, Bayer R, and Colgrove J. The renormalization of smoking? E-cigarettes and the tobacco "endgame". N Engl J Med. 2014;370(4):293–5. pmid:24350902
- 29. Family Smoking Prevention and Tobacco Control Act. 21 USC §301 (2009), Pub L No. 111–31, 123 Stat 1776–1858. http://www.gpo.gov.libproxy.usc.edu/fdsys/pkg/PLAW-111publ31/pdf/PLAW-111publ31.pdf (accessed 6 Nov 2014). Family Smoking Prevention and Tobacco Control Act. 21 USC §301 (2009), Pub L No. 111–31, 123 Stat 1776–1858. http://www.gpo.gov.libproxy.usc.edu/fdsys/pkg/PLAW-111publ31/pdf/PLAW-111publ31.pdf (accessed 6 Nov 2014).
- 30. Hua M, Alfi M, Talbot P. Health-related effects reported by electronic cigarette users in online forums. J Med Internet Res. 2013;15:e59. pmid:23567935
- 31. Harris JK, Moreland-Russell S, Choucair B, Mansour R, Staub M, Simmons K. Tweeting for and against public health policy: response to the Chicago Department of Public Health's electronic cigarette twitter campaign. J Med Internet Res. 2014;16:e238. pmid:25320863
- 32. Ayers JW, Althouse BM, Ribisl KM, and Emery S. Digital detection for tobacco control: Online reactions to the United States' 2009 cigarette excise tax increase. Nicotine Tob Res. 2013;16(5):576–83. pmid:24323570
- 33. Ayers JW, Ribisl K, and Brownstein JS. Using search query surveillance to monitor tax avoidance and smoking cessation following the United States' 2009 "SCHIP" cigarette tax increase. PLoS One. 2011;6(3):e16777. pmid:21436883
- 34. Ayers JW, Althouse BM, and Emery S. Changes in Internet searches associated with the "Tips from Former Smokers" campaign. Am J Prev Med. 2015;48(6):e27–e33. pmid:25998929
- 35. Ayers JW, Althouse BM. "Tips From Former Smokers" Can Benefit From Considering All Available Data: Reply to McAfee et al. Am J Prev Med. 2015;49(6):e133–e134. pmid:26456875
- 36. Allem JP, Ayers JW, Althouse BM, Williams R. When a ban really is not a ban: internet loopholes and Djarum flavoured cigarettes in the USA. Tob Control. 2015.
- 37. Ayers JW, Althouse BM, Johnson M, Cohen . Circaseptan (weekly) rhythms in smoking cessation considerations. JAMA Intern Med. 2014;174:146–148. pmid:24166181
- 38. Ayers JW, Althouse BM, Allem JP, Ford DE, Ribisl KM, Cohen JE. A novel evaluation of world no tobacco day in Latin America. J Med Internet Res. 2012;14:e77. pmid:22634568
- 39. Chunara R, Andrews JR, Brownstein JS. Social and news media enable estimation of epidemiological patterns early in the 2010 Haitian cholera outbreak. Am J Trop Med Hyg. 2012;86:39–45. pmid:22232449
- 40. Prochaska JJ, Pechmann C, Kim R, and Leonhardt JM. Twitter = quitter? An analysis of Twitter quit smoking social networks. Tob Control. 2012;21(4):447–9. pmid:21730101
- 41. Lazer D, Kennedy R, King G, and Vespignani A. The parable of Google Flu: Traps in big data analysis. Science. 2014;343:1203–1205. pmid:24626916
- 42. Santillana M, Zhang DW, Althouse BM, and Ayers JW. What can digital disease detection learn from (an external revision to) Google Flu Trends? Am J Prev Med. 2014;47(3):341–347. pmid:24997572
- 43. Olson DR, Konty KJ, Paladini M, Viboud C, Simonsen L. Reassessing Google Flu Trends data for detection of seasonal and pandemic influenza: a comparative epidemiological study at three geographic scales. PLoS Comput Biol. 2013;9:e1003256. pmid:24146603
- 44. Althouse BM, Scarpino SV, Meyers LA, et al. Enhancing disease surveillance with novel data streams: challenges and opportunities. EPJ Data Science. 2015;4:1.