This is an open-access article distributed under the terms of the Creative Commons Attribution License (https://creativecommons.org/licenses/by/4.0/), which permits unrestricted use, distribution, and reproduction in any medium, provided the original work, first published in the Journal of Medical Internet Research, is properly cited. The complete bibliographic information, a link to the original publication on https://www.jmir.org/, as well as this copyright and license information must be included.
Examining public perception of tobacco products is critical for effective tobacco policy making and public education outreach. While the link between traditional tobacco products and lung cancer is well established, it is not known how the public perceives the association between electronic cigarettes (e-cigarettes) and lung cancer. In addition, it is unclear how members of the public interact with official messages during cancer campaigns on tobacco consumption and lung cancer.
In this study, we aimed to analyze e-cigarette and smoking tweets in the context of lung cancer during National Cancer Prevention Month in 2018 and examine how e-cigarette and traditional tobacco product discussions relate to implementation of tobacco control policies across different states in the United States.
We mined tweets that contained the term “lung cancer” on Twitter from February to March 2018. The data set contained 13,946 publicly available tweets that occurred during National Cancer Prevention Month (February 2018), and 10,153 tweets that occurred during March 2018. E-cigarette–related and smoking-related tweets were retrieved, using topic modeling and geospatial analysis.
Debates on harmfulness (454/915, 49.7%), personal experiences (316/915, 34.5%), and e-cigarette risks (145/915, 15.8%) were the major themes of e-cigarette tweets related to lung cancer. Policy discussions (2251/3870, 58.1%), smoking risks (843/3870, 21.8%), and personal experiences (776/3870, 20.1%) were the major themes of smoking tweets related to lung cancer. Geospatial analysis showed that discussion on e-cigarette risks was positively correlated with the number of state-level smoke-free policies enacted for e-cigarettes. In particular, the number of indoor and on campus smoke-free policies was related to the number of tweets on e-cigarette risks (smoke-free indoor,
Though people recognized the importance of traditional tobacco control policies in reducing lung cancer incidences, their views on e-cigarette risks were divided, and discussions on the importance of e-cigarette policy control were missing from public discourse. Findings suggest the need for health organizations to continuously engage the public in discussions on the potential health risks of e-cigarettes and raise awareness of the insidious lobbying efforts from the tobacco industry.
Tobacco control has been identified as a global public health priority by the World Health Organization [
Particularly, health education campaigns and control policies should pay attention to tobacco consumption trends. In terms of traditional tobacco use, the CDC reported that smoking among adults had declined from 20.9% in 2005 to 13.7% in 2018, and the proportion of smokers who reported quitting had increased [
The overwhelming popularity of e-cigarettes may be due to conflicting messages in the public. Some argue that e-cigarettes could help with smoking cessation as they appear to pose fewer health risks than traditional cigarettes [
Survey studies have investigated the perceived associations of smoking traditional tobacco products and e-cigarettes with health diseases such as lung cancer. Smoking is recognized as a major risk factor for lung cancer by the public [
Furthermore, to the best of our knowledge, none of the existing social media studies has investigated spatial patterns of tobacco conversations on social media in relation to implementation of health policies. For example, it has been found that the number of obesity-related policies in certain geographic regions were associated with an increase in obesity prevention discussions on Twitter within the same area [
While the majority of social media studies largely examine public discourse of e-cigarette and smoking in general, this study examined tobacco discourse in the context of US National Cancer Prevention Month, which is an annual campaign led by the American Institute for Cancer Research in the month of February that aims to foster cancer knowledge and promote cancer prevention practices [
Past research has shown that public discourse during cancer campaigns may be different from that during other months. Cancer campaigns raise public awareness by promoting cancer conversations about risk factors and preventions on Twitter [
First, we aimed to examine if National Cancer Prevention Month plays a role in promoting conversations on the link of e-cigarettes and smoking with lung cancer. Second, we aimed to examine and compare public discourse in the United States on smoking and e-cigarette in the context of lung cancer on Twitter during National Cancer Prevention Month. Third, we aimed to examine if there were spatial patterns of smoking and e-cigarette’s themes in the United States during National Cancer Prevention Month. Fourth, we explored the relationship between e-cigarette and smoking discussions on Twitter during National Cancer Prevention Month with implementation of tobacco control policies. As such, we put forth 4 research questions: (1) Does national cancer prevention month promote e-cigarette and smoking conversation related to lung cancer? (2) What are the key themes in e-cigarette and smoking tweets within the broader context of lung cancer discussion during National Cancer Prevention Month? (3) Are there geospatial differences in how e-cigarette and smoking tweets were distributed across the United States during National Cancer Prevention Month? (4) What is the relationship between the number of tobacco control policies in states and themes of e-cigarette and smoking tweets?
Data were retrieved from an existing data set of US English-based lung cancer tweets that contained the term “lung cancer” purchased from Twitter. A list of 28 keywords, such as “e-cigarette,” “vape,” and “juul” [
Search keywords for data collection.
Topic | Keywords |
E-cigarette | electronic cigarette; vap*; e-cig*; ecig*; e cig; e-pen; epen; e pen; e-juice; ejuice; e juice; e-liquid; eliquid; e liquid; esmoke; e-smoke; e smoke; e-hookah; ehookah; e hookah; e-pipe; epipe; e pipe; atomizer; juul; njoy; v2 cig; joye510 |
Smoking | cig*; tobacco; waterpipe; water pipe; hooka; smok*; chew; nicotine; shisha; sheesha; bidi; beedi; kretek; narghile; argileh; cheroot; snuff; snus; betel; gutkha; toombak |
R statistical software (The R Project) was used for textual analyses. Data were preprocessed and cleaned before advanced textual analysis. Texts were formatted to lower case. Different forms of phrases that had the same meaning were transformed into a common format to facilitate future text processing, such that “e cig,” “e-cig,” “ecig,” and “electronic cigarette” were reformatted into “ecigarette.” Common English stop words, such as “the” and “of,” special characters, and punctuations were removed. The remaining texts were tokenized and lemmatized to further avoid inflected words. The word “lung cancer” was also removed from the analysis.
Topic modeling, using latent Dirichlet allocation, was employed to understand the differences between themes of e-cigarette and smoking in the context of lung cancer discourse. Latent Dirichlet allocation is a popular and widely used algorithm for topic modeling, by which documents are modeled as mixtures over topics and a topic is characterized as a distribution of words [
The latent Dirichlet allocation algorithm requires a predefined number,
To evaluate the prevalence of topics, we used methods described in [
We compared the temporal distribution of e-cigarette and smoking tweets using chi-square analysis.
The geolocation of a tweet was determined by the self-reported location in the profile of the relevant Twitter user. We imported the location strings into the Google Maps geocoordinates application programming interface (API) to obtain the geocoordinates and the corresponding states in the United States. Then, we manually checked to ensure the state information is correct for each tweet. Tweets that did not have a user-reported location or whose reported location string did not return any results were excluded from the geospatial analysis. To further understand the spatial distribution of themes, ratio values were calculated for each state by dividing the number of tweets in each theme by the total number of tweets in each state. We plotted the ratio values by state to visualize the spatial distribution patterns.
To further understand how state policies might affect twitter discussions of e-cigarette and smoking, we compared the number of tweets with the number of state policies (existing, introduced, or recently enacted in the first quarter of 2018) related to e-cigarettes and smoking. State policies were obtained from the tobacco use data portal from the US CDC [
The data set had 13,946 publicly available tweets obtained during National Cancer Prevention Month (ie, February) and 10,153 tweets obtained during March in 2018 in the United States. The keyword queries returned 1061 e-cigarette tweets and 4019 smoking tweets during National Cancer Prevention Month, and 171 e-cigarette tweets and 1919 smoking tweets during March (
Study flowchart.
There was a significant difference between the temporal distributions for e-cigarette and smoking tweets (χ2=256.85,
During National Cancer Prevention Month, 3 major e-cigarette themes—comprising 5 topics—emerged (
E-cigarette themes from tweets (based on latent Dirichlet allocation algorithm).
Theme | Words | Examples | Tweets (n= 915), n (%) | |
|
|
|
454 (49.7) | |
|
Affective reasoning | Juul, hit, kid, people, rumor, cause, shit, really, report, untrue | Drop your juuls like deadass. Have heard of three people my age who have been diagnosed with lung cancer from juuls. Feel like we all saw this coming | 158 (17.3) |
|
Cognitive reasoning | Juul, year, cause, cancer, develop, know, take, use, lung, say | When people think that juuls give you lung cancer but lung cancer takes years to develop | 127 (13.9) |
|
Sarcasm | Juul, everyone, day, February, simultaneously, friend, today, people, college, think | February 7, 2018. The day everyone who juuls simultaneously got lung cancer. | 169 (18.5) |
Personal experiences | Friend, cousin, stop, girl, good, sister, neighbor, immediately, son, sorority | STOP JUULING IMMEDIATELY‚ My best friends neighbors girl friend’s sorority sister’s cousin’s step son got lung cancer from a single hit of juul. Drop these cancer sticks. | 316 (34.5) | |
E-cigarette risks | Juul, link, lung, kid, disease, ecigarette, severe, flavoring, hit, addict | ecigarette Flavorings linked to Severe LUNG disease https://t.co/2VhStSkI0s #lungdisease #lungcancer #ecigarettes #cancer #ecigaretteflavoring #severlungdisease #howbadareecigarettes | 145 (15.8) |
Unlike themes expressed in e-cigarette tweets, which showed that users were divided over the association of e-cigarettes with lung cancer, those expressed in smoking tweets (
Smoking themes (based on latent Dirichlet allocation) from tweets.
Theme | Words | Examples | Tweets (n= 3870), n (%) | ||||
|
|
|
2251 (58.1) | ||||
|
Tobacco lobbying | Tobacco, kid, one, lose, prevent, explain, love, lobbyist, invite, this | This is like inviting tobacco lobbyists to explain to kids who have lost loved ones to lung cancer how we can prevent smoking deaths. Lobbyists don’t deserve a seat at this table. | 1831 (47.3) | |||
|
Smoking control policy | Smoker, cigarette, gun, smoke, cause, blame, tobacco, gum, chew, death | Higher prices / taxes on cigarettes=less deaths due to lung cancer. Seat belts / stricter regulations on vehicle safety=less auto deaths | 420 (10.8) | |||
Smoking risks | Smoke, cancer, smoker, lung, non, die, people, risk, quit, cigarette | 433 Americans die daily from #lungcancer. The majority of people living with lung cancer r nonsmokers or have quit smoking. Anyone, smoker or nonsmoker, can get lung cancer. While smoking greatly increases the risk of #lungcancer , NO ONE DESERVES CANCER. @theNCI #LCSM https://t.co/9CtilnJLzm | 843 (21.8) | ||||
Personal experiences | Smoke, get, cigarette, people, die, tobacco, cause, someone, kill, make | Great! My big brother was smoking for the last fifty years Lung cancer finally killed him. Small pain under the arm one sunny morning. Two years later a 3cm tumor killed him. | 776 (20.1) |
The geospatial analysis included 96.7% (885/915) of the e-cigarette tweets and 89.3% (3455/3870) of the smoking tweets during National Cancer Prevention Month. Overall, discussions of e-cigarettes and smoking in relation to lung cancer occurred mostly in the coastal areas and the eastern part of the country (
The spatial distribution of e-cigarette and smoking tweets mentioning lung cancer during US national cancer prevention month (February 2018).
For e-cigarette tweets, California, Arizona, Pennsylvania, Illinois, and Virginia had more tweets debating e-cigarettes than tweets containing the other two themes. Oregon, Texas, Tennessee, and North Carolina had more tweets asking people to stop vaping based on personal stories. Hawaii and the Washington state had more tweets on scientific evidence showing the link between e-cigarettes and lung diseases.
For smoking tweets, most states predominantly had tweets about policy discussions regarding smoke control and how tobacco control policies were important to reign in tobacco companies, equivalent to how gun control policies would restrict gun lobbyists. Nevada and Kentucky had more tweets about the scientific evidence of smoking risks than those about the other two themes.
The number of tweets under the theme
We examined the prevailing topics and distributions of discussions in Twitter about e-cigarettes and traditional tobacco consumption during the National Cancer Prevention Month in 2018 within the broader context of lung cancer to offer key insights on how the public perceives health risks of both e-cigarettes and smoking and potentially help public health organizations to be more strategic in their messaging and tobacco control efforts by targeting different tobacco products.
First, the findings of temporal distributions of e-cigarette and smoking tweets suggest that National Cancer Prevention Month promoted both e-cigarette and smoking conversations related to lung cancer. What we found notable was that National Cancer Prevention Month promoted e-cigarette conversations more than smoking conversations in the context of lung cancer. Without the cancer campaigns, lung cancer discourse on Twitter were rarely about e-cigarettes (76/10,153, 0.75%) and mostly revolved around the harms of smoking (1824/10,153, 18.0%). This is likely for a few reasons. In February 2018, the American Cancer Society [
Second, the findings of our thematic analysis suggest that Twitter users were aware of the risks of lung cancer from smoking but were split over the potential health effects of vaping. While some of the Twitter users evaluated the link between e-cigarettes and lung cancer based on personal experiences or anecdotal stories they have heard, others processed e-cigarette information in a more cerebral manner and were convinced of the health risks of e-cigarettes. This split in attitude toward e-cigarettes may be the result of mixed communication messaging from public health organizations. For instance, while the CDC has acknowledged the risks of e-cigarettes, particularly for young people due to the presence of nicotine, the long-term health effects of e-cigarettes have been debated [
Another significant finding was that themes of political lobbying and policy making were absent from e-cigarette tweets, but not from those about traditional tobacco consumption, during the cancer campaign. When discussing smoking, Twitter users were mindful of the political lobbying by tobacco industries (and equated it to that of gun lobbyists), but this particular theme was missing from e-cigarette tweets. This is crucial, as it suggests that the political lobbying efforts by e-cigarette companies may not be as visible or prominent as those of the traditional tobacco industry. This is a cause for concern. After all, the tobacco industry is very much involved in the e-cigarette industry, as shown by the acquisition of Juul by Altria (formerly known as Philip Morris) for US $12.8 billion in 2018 [
Third, this study demonstrated geospatial differences in e-cigarette and smoking discussions on Twitter during National Cancer Prevention Month. In terms of discussing e-cigarette risks, results showed that only 2 states—Hawaii and Washington—had more discussions than those of the others. In the state of Washington, 30% of 12th grade students used e-cigarettes [
There was a positive correlation between discussion of e-cigarette risks and the number of smoke-free policies at the state level. While the findings cannot be used to make any causal claims, it is worth nothing that there may be a reciprocal relationship between public awareness of e-cigarette risks and the passing of smoke-free policies. In other words, when the public becomes aware of the risks of e-cigarettes, they may encourage local representatives to push for more smoke-free policies. At the same time, the passing of smoke-free policies may further increase awareness of e-cigarette risks in the general public.
Though discussion of tobacco risks and the number of smoke-free policies were not correlated, as discussed, people still mentioned the importance of smoking control policies in their tweets. The data suggest that policy engagement and public awareness and discussion of tobacco risks are symbiotic. When the risk awareness of a tobacco product is low, especially for emerging tobacco products such as e-cigarettes, public policy engagement motivated by the community leaders or public health organizations may heighten risk awareness. Once the public are adequately educated on the health risks of a tobacco product (eg, combustible cigarettes), this risk awareness may, in turn, fuel discussions on the need for stringent tobacco control policies, as well as strategically address tactics of political lobbying and messaging by the tobacco industry. In other words, the findings of our study suggest that public health organizations should focus on both improving risk awareness of tobacco products, as well as engaging and educating the public on the importance of tobacco control policies, because these strategies complement and reinforce one another.
We believe that our findings will be useful to help health communication scholars understand public perception and attitudes toward e-cigarettes and smoking. Future studies should (1) test potential reciprocal relationship between policy engagement and risk awareness of tobacco products; (2) investigate the underlying mechanisms, specifically examine how National Cancer Prevention Month or other cancer awareness months could promote e-cigarette discussions with randomized controlled trials, and identify the best strategies in educating the public about the harms of vaping; and (3) replicate our study by examining how cancer awareness months drive conversations about other cancers (eg, breast cancer, prostate cancer) compared with other noncancer awareness months and how various health policies (eg, health insurance) across different states are associated with cancer discussion.
First, while social media sources, such as Twitter, can be used to gauge public opinion and sentiments toward smoking and e-cigarette, we are mindful that they may not be representative, and as such, there are constraints on the generalizability of the results. For example, our data came from publicly available posts, and thus, we were not able to capture themes and sentiments toward smoking and e-cigarettes in private posts. Also, because not all users reported their locations in their profiles, there may be potential selection biases in the geospatial analysis. In addition, we are cognizant that there potentially could be a spill-over effect because our data were collected from consecutive months in February and March. However, we are confident that this was not a major issue of concern given that the number of e-cigarette tweets in February (n=1061) during National Cancer Prevention Month was much greater than the number of e-cigarette tweets in March (n=171). Finally, we excluded tweets with both sets of keywords that might introduce bias and ran the same analyses; we found that the results did not substantially differ; therefore, we are confident that our results are robust.
The public is aware of smoking and lung cancer risks, but people were generally divided over the risks of e-cigarettes in relation to lung cancer. Public health organizations should invest in strategic messaging efforts over social media to address any misinformation about e-cigarettes because there is a reciprocal relationship between public awareness and discussion on tobacco products on social media and the implementation of tobacco control policies.
The spatial distribution of Twitter themes in relation to lung cancer during US national cancer prevention month (February 2018).
Correlation between the tweet number of themes and the number of state-level tobacco policies.
Centers for Disease Control and Prevention
electronic cigarette
This project was supported by a research grant from Nanyang Technological University (grant M020060110).
JL and EL contributed to the design of the study. JL performed the analyses. JL and EL drafted the manuscript. Both authors contributed to revision of the manuscript and approved the final version for submission.
None declared.