Spanish-Language Tobacco-Related Posts on Twitter: Content Analysis

Abstract

Introduction

Twitter data have been used to surveil public sentiment about tobacco products; however, most tobacco-related Twitter research has been conducted with English-language posts. There is a gap in the literature on tobacco-related discussions on Twitter in languages other than English. This study summarized tobacco-related discussions in Spanish on Twitter.

Methods

A set of Spanish terms reflecting electronic cigarettes (eg, "cigarillos electrónicos"), cigarettes (eg, “pitillo”), and cigars (eg, “cigaro”) were identified. A content analysis of tweets (n = 1352) drawn from 2021 was performed to examine themes and sentiment. An initial codebook was developed in English then translated to Spanish and then translated back to English by a bilingual (Spanish and English) member of the research team. Two bilingual members of the research team coded the tweets into themes and sentiment.

Results

Themes in the tweets included (1) product promotion (n = 168, 12.4%), (2) health warnings (n = 161, 11.9%), (3) tobacco use (n = 136, 10.1%), (4) health benefits of vaping (n = 58, 4.3%), (5) cannabis use (n = 50, 3.7%), (6) cessation (n = 47, 3.5%), (7) addiction (n = 33, 2.4%), (8) policy (n = 27, 2.0%), and (9) polysubstance use (n = 12, 0.9%). Neutral (n = 955, 70.6%) was the most common category of sentiment observed in the data.

Conclusions

Tobacco products are discussed in multiple languages on Twitter and can be summarized by bilingual research teams. Future research should determine if Spanish-speaking individuals are frequently exposed to pro-tobacco content on social media and if such exposure increases susceptibility to use tobacco among never users or sustained use among current users.

Implications

Spanish-language pro-tobacco content exists on Twitter, which has implications for Spanish-speaking individuals who may be exposed to this content. Spanish-language pro-tobacco-related posts may help normalize tobacco use among Spanish-speaking populations. As a result, anti-tobacco tweets in Spanish may be necessary to counter areas of the online environment that can be considered pro-tobacco.

Introduction

Public health researchers have used Twitter data to monitor the tobacco-related attitudes and behaviors of the public in near real time.^1,2 Twitter data, used in this way, can be considered a massive focus group whereby Twitter users organically share their tobacco-related attitudes and behaviors in their own words without the prompt of a researcher or instrument bias. Past literature has largely relied on examinations of English-language tweets,³ with studies purposefully removing, or forgoing analysis of, tweets in other languages. As a result, there is a gap in the literature on the content of tobacco-related discussions on Twitter in languages other than English, with Spanish as a prime example.

Past research on English-language tobacco-related tweets has generated insights about user preferences for flavored tobacco,⁴ difficulty quitting e-cigarettes and other novel tobacco products,⁵ locations of youth tobacco use (eg, at school),⁶ unsubstantiated health claims,⁷ among other topics. However, it is unclear whether Spanish-language tobacco-related tweets express the same tobacco-related attitudes and behaviors as those in English. This gap in the literature is problematic for multiple reasons. Because many anti-tobacco messages across social media are in English, especially in the United States, it is possible that Spanish speakers have not benefited equally from tobacco prevention and cessation messages and/or may believe misinformation about the health effects of tobacco use. Additionally, those with limited English proficiency face barriers to accessing and utilizing traditional resources/programs (eg, smoking quitlines).⁸ What is more, in the United States, Spanish speakers (ie, Hispanics/Latinos) are more likely to receive, read, and share misinformation online compared to the general population.⁹ As a result, Spanish speakers may be at higher risk for harm from social media, depending on the nature of the posts available on the platforms.

Dating back to 1980 in the United States, Spanish speakers have been targeted in tobacco advertising by the major tobacco companies.¹⁰ For example, Winston and Camel brands created marketing materials designed to appeal to Hispanics/Latinos by incorporating messages/imagery speaking to their cultural values (eg, strong family ties), and affiliated with culturally significant activities/events.¹⁰ Although the prevalence of tobacco use is slightly lower among Hispanics/Latinos than among non-Hispanic Whites, Hispanic/Latino youth are at a higher risk of experimenting with novel tobacco products such as e-cigarettes.⁸ Spanish-speaking Hispanic/Latino youth in the United States may also be less likely to benefit from prevention programs offered through mass media (eg, The Real Cost), or school-based programs, that are primarily available in English. Therefore, understanding the online discussions about tobacco use among Hispanics/Latinos is an important public health priority. Because exposure to pro-tobacco messages in social media has been associated with tobacco use,^11–13 it is important to understand the content of tobacco-related social media posts in multiple languages to develop comprehensive public health campaigns.¹⁴

Data are currently unavailable on the themes of Spanish-language tobacco-related tweets. While the content of tobacco-related discussions on Twitter in Spanish may be like those in English, Spanish posts may offer unique insights that have gone undetected in the tobacco-related social media literature, which could lead to a missed opportunity to provide anti-tobacco health communications to Spanish-speaking communities. This study identified and summarized Spanish-language tobacco-related posts on Twitter.

Methods

Data Collection

The research team has been collecting tobacco-related posts from Twitter’s streaming application programming interface since 2015, with 191 203 075 tweets collected in 2021. This corpus of Twitter posts was searched for Spanish-language tobacco-related posts. To accomplish this, a set of Spanish terms reflecting three popular classes of tobacco products were identified, including electronic cigarettes (ie, "cigarillos electrónicos,” “vaporizador,” “vaporizadores,” “pluma vape,” “vapear,” “vapeador,” “vapeadores,”), cigarettes (ie, “cigarrillo,” “pitillo,” “cigarillos,” “pitillos,”), and cigars (ie, “cigaro,” “puro,” “cigaros,” “puros”). The total number of Spanish-language tobacco-related posts from 2021 was (N = 63 036). These Spanish terms were selected based on feedback from members of the research team who are native Spanish speakers from Colombian/Colombian American and Mexican/Mexican American backgrounds. As such, each team member offered suggestions for included terms, including colloquialisms. This process included a primary focus on technical terms for vaping and other tobacco products, for example, vapear for vaping, then moved on to consider possible slang terms used in different Spanish-speaking populations, for example, pitillos for cigarettes or puro for cigar.

To arrive at a sample of tweets to content analyze, retweets were removed (n = 32 093),⁶ ensuring that each tweet could be treated as an independent observation. Tweets were then selected proportionally by term and by week to arrive at an analytical sample (n = 1480). However, as the coding process took place some tweets (n = 128) were no longer available (ie, the tweet was made unavailable by the content creator due to changes in privacy settings), resulting in a final analytic sample of (n = 1352).

Coding Procedure

The research team familiarized themselves with the data and drafted an initial codebook in English. The purpose of this process was to summarize the underlying themes evident in the tweets. Saturation was determined to be reached with nine themes. The initial codebook was then translated into Spanish by a bilingual (Spanish and English) member of the research team. To verify translation accuracy, the Spanish version of the codebook was translated back to English by a second member of the research team to compare the two versions of the codebook—the original version and the final version in English. All authors agreed that the two English-language versions of the codebook were similar in meaning and expression. The Spanish codebook and its English translation can be found in Online Appendix A.

Themes included (1) tobacco use (ie, mentioned self-reported tobacco use, including e-cigarettes, little cigars, or hookah), (2) cessation (ie, mentioned quitting nicotine or giving up e-cigarette use or other forms of tobacco use), (3) health warnings (ie, mentioned the health effects or adverse effects of e-cigarette, or other tobacco product use), (4) health benefits of vaping (ie, mentioned expected public health benefits from vaping, such as reductions in smoking rates, lives saved, and/or reducing the burden of tobacco-related disease), (5) policy (ie, mentioned a tobacco control policy), (6) product promotion (ie, mentioned price, specific product lines, or contact information including email, physical address, or phone number), (7) addiction (ie, mentioned addiction to nicotine or e-cigarettes such as JUUL, little cigars, hookah, or other tobacco products), (8) polysubstance use (ie, mentioned using e-cigarettes, little cigars, or other tobacco products in combination with substances such as alcohol, marijuana, or other drugs), and (9) cannabis (ie, mentioned self-reported cannabis use, including joints, blunts, pipes, bongs, or edibles).

Similar to past research,¹⁵ this study coded for the sentiment of each tobacco-related post. If the post promoted/encouraged tobacco use it was coded as positive. If the post discouraged/deterred tobacco use it was coded as negative. If the post was ambiguous and/or had no clear position toward tobacco use it was coded as neutral. The unit of analysis was the text and any corresponding hashtags in a tweet. The coders were instructed to place a “1” in the cell if a theme was present and if not, to place a “0.” Each post could contain more than one theme and positive and negative sentiment.

To establish interrater reliability, two coders analyzed a subsample of posts (n = 389). Discrepancies between the coders were resolved with the help of a third coder. Coding instructions were clarified to help with discrepancies during this process. All coders were bilingual. The percent agreement ranged from 78.7% to 97.4% across the nine themes, which is considered acceptable.¹⁶ The percent agreement for sentiment was 68.6% for neutral, 70.7% for positive, and 82.5% for negative. Descriptive statistics were reported for all the themes, each category of sentiment, and the cross-tabulation of themes and sentiment. All analyses were performed using Excel version 16.54. All the posts in this dataset were publicly available and anonymized, and all analyses adhered to the terms and conditions, terms of use, and privacy policies of Twitter, and were performed under the University of Southern California Institutional Review Board approval.

Results

The total coverage of the nine themes identified in the data constituted 44.6% of all tweets in the corpus. The remaining 55.4% of tweets were too varied to be classified into a single topic with meaningful coverage (ie, coverage of each subsequent topic would be less than 0.05% of total tweets). The theme that was most frequently observed in the corpus of posts was product promotion (n = 168, 12.4%), followed by health warnings (n = 161, 11.9%; Table 1). Polysubstance use (n = 12, 0.9%) was the least common theme observed in the data. Few posts contained more than one theme (5.6%). Neutral (n = 955, 70.6%) was the most common category of sentiment observed in the data. Cross-tabulations of theme and sentiment showed that health warnings and negative (n = 138, 10.2%) were most frequently observed, followed by product promotion and positive (n = 129, 9.5%; Table 2).

Table 1.

Prevalence of Themes and Sentiment of Spanish Posts on Twitter (N = 1352)

Themes	N	%
Product promotion	168	12.4
Health warnings	161	11.9
Tobacco use	136	10.1
Health benefits of vaping	58	4.3
Cannabis use	50	3.7
Cessation	47	3.5
Addiction	33	2.4
Policy	27	2.0
Polysubstance use	12	0.9
Sentiment
Neutral	955	70.6
Positive	225	16.6
Negative	174	12.8
Total*	1354

^*Two posts contained both positive and negative sentiments (n = 2).

Table 2.

Cross-tabulation of Themes and Sentiment of Spanish Posts on Twitter (N = 1352)

	Sentiment
Themes	Positive	Negative	Neutral
Product promotion	129 (9.5)	1 (0.10)	38 (2.8)
Tobacco use	40 (3.0)	4 (0.30)	92 (6.8)
Health benefits of vaping	24 (1.8)	7 (0.50)	29 (2.1)
Addiction	9 (0.70)	9 (0.70)	15 (1.1)
Cannabis use	8 (0.60)	1 (0.10)	41 (3.0)
Cessation	7 (0.50)	15 (1.1)	26 (1.9)
Polysubstance use	4 (0.30)	1 (0.10)	7 (0.50)
Health warnings	3 (0.20)	138 (10.2)	22 (1.6)
Policy	2 (0.10)	7 (0.50)	18 (1.3)

Values in the cells represent the frequencies (percentages) of Spanish posts on Twitter.

Discussion

This study provided a summary of publicly available Spanish-language tobacco-related posts on Twitter from 2021. Themes in the corpus included pro-tobacco themes, such as product promotion and tobacco use, as well as anti-tobacco themes, such as cessation and health warnings. Most tweets contained a neutral sentiment. Given that a minority of tweets were negative in sentiment, counter-marketing campaigns, programs, and interventions that are delivered in Spanish on Twitter may be needed to help denormalize tobacco use among Spanish-speaking populations, especially in the United States.

Product promotion was the most observed theme in the corpus. While past research relying on Twitter data has shown similar findings,¹⁷ this study demonstrates that tobacco products are promoted in languages aside from English on Twitter. While the tobacco industry has a long history of sponsoring offline community cultural events and/or chambers of commerce to establish community relations among Spanish-speaking consumers and potential consumers,¹⁸ little is known about how the industry is targeting Spanish-speaking populations through online channels. Spanish-language tobacco content has implications for Spanish-speaking individuals who may be exposed to this content. As such, Spanish-language health communication programs and interventions may be needed to counter both tobacco industry marketing efforts and positive valence references to tobacco use behaviors that appear in the online environment.

Similar to past research,^6,19,20 this study found that tobacco-related tweets contained positive tobacco use portrayals with an e-cigarette, little cigar, or hookah product. In addition to pro-tobacco themes, this study showed that Spanish-language tweets contained anti-tobacco themes, such as health warnings, and other less prevalent themes like cessation. The presence of health warnings on tweets echoes similar research that has found that content discouraging youth use of tobacco products and/or steps to curb youth access were found on the official channels of e-cigarette companies on YouTube.²¹ Other Twitter research has shown that themes related to cessation were uncommon compared to pro-tobacco themes like discussing new product flavors.⁶ However, Twitter has been used as a tool to assist in smoking cessation efforts and smoking cessation campaigns.^22,23 Research has shown that online banner advertising can be an effective and cost-efficient strategy to reach and engage Spanish-speaking Latino smokers.¹⁴ Twitter may be considered for future efforts to promote smoking cessation among Spanish speakers. Given the findings from this study, smoking cessation campaigns could use a combination of colloquialisms and technical terms to identify those talking about tobacco online and focus campaign messages on the health effects or adverse effects of e-cigarettes, or other tobacco products.

Tweets related to cannabis use and, in some instances, polysubstance use were observed in this study. Prior research relying on Twitter data has shown similar findings.²⁰ Social cognitive theory suggests that behavior is learned and reinforced by observing or modeling others.²⁴ Social media content that depicts polysubstance use in a positive valence may serve as a socializing force for young social media users, persuading them to believe that substance use, including tobacco use, is normative, attractive, and/or rewarding.^25,26 Spanish-language educational programs and interventions may help Spanish speakers build skills to counteract pro-substance use themes on social media, and develop the necessary tools to become critical consumers of digital content.²⁷

Limitations

This study was limited to the analysis of discussions on three tobacco product types (electronic cigarettes, cigarettes, and cigars) and may not generalize to other tobacco products. Aside from “vapear” (ie, vaping), this study did not search for posts that contained behavioral terms (eg, smoking, puffing). Additionally, language colloquialisms can vary by class. The study team may not be reflective of the socio-economic demographics of those most likely to use tobacco. As such, the colloquialisms used in this study may not be reflective of the population of interest. This study focused on the text of each Twitter post but did not code website links or accompanying images. Additional themes may have emerged had the research team done so. The tweets collected in this study were unrestricted by geography. Given the limited geographic information available in the data, tweets went unexamined by location. This is similar to prior research that found that 1% of posts included geolocation data.²⁸ Twitter (like many social media platforms) has an abundance of bots, false or duplicate profiles, and other non-“face-value” sources of content (eg, “astroturfing”).²⁸ It was beyond the scope of this study to conduct a formal or informal review of the data to classify these data by propensity for user, industry, or non-“face-value” posts. Findings from this study may not generalize to other time periods or other social media platforms. Findings from this study may not extend to all Spanish-speaking Twitter users or to those who do not speak Spanish.

Conclusions

This study demonstrated that Spanish-speaking Twitter users could be exposed to pro-tobacco tweets, and these tweets may convey similar themes seen in previous analyses of English-language tweets.^7,15,20 In the United States, language barriers can limit Spanish speakers’ ability to access tobacco education resources and communicate with their physicians about tobacco cessation.⁸ Public health initiatives that increase the prevalence of accurate information about the health consequences of tobacco and provide links to cessation resources in Spanish could potentially decrease tobacco-related health disparities among Spanish speakers. Social media-based tobacco prevention and cessation programs have been effective in the general population and among Spanish speakers, although recruitment of Spanish speakers remains a challenge.^29–32 Continued efforts are needed to connect Spanish speakers with accurate tobacco prevention and cessation information and resources to combat the potential effects of pro-tobacco posts on Twitter.

This study’s findings have implications for preventive, youth-focused interventions. In the United States, federally funded, youth-focused prevention campaigns for the general population (eg, The Real Cost) and subpopulations (eg, Fresh Empire), are primarily discussed via social media and are done so exclusively in English. Expanding these to include Spanish would be a similar, likely more effective, approach to prevention as that of Spanish-language Twitter advertisements for cessation programs. Future research should determine how frequently Spanish-speaking individuals are exposed to pro-tobacco content on social media and if such exposure increases susceptibility to use tobacco among never users or continued use among current users. Future research should scrutinize user profiles to understand the source of posts, determining if posts originated directly from industry accounts (eg, influencers, brands, vendors) or the general public.

Supplementary Material

Supplementary material is available at Nicotine and Tobacco Research online.

Funding

This study was supported by the California Tobacco Control Branch, Center for Healthy Communities of the California Department of Public Health through contract number 21-10032 Tobacco Industry Monitoring Evaluation, Grant #U54 CA 180905 from the National Cancer Institute and the Food and Drug Administration Center for Tobacco Products, and The Regents of the University of California, Research Grants Program Office, Tobacco-Related Diseases Research Program, Grant Number No. T30IR0891. The findings and conclusions in this article are those of the authors and do not necessarily represent the views or opinions of the California Department of Public Health, the NIH, or the California Health and Human Services Agency.

Declaration of Interests

JPA has received fees for consulting services in court cases pertaining to the content on social media platforms. The authors have no other conflicts of interest to disclose.

Author Contributions

Jon-Patrick Allem (Conceptualization [Equal], Data curation [Equal], Funding acquisition [Equal], Investigation [Equal], Methodology [Equal], Project administration [Equal], Supervision [Equal], Writing—original draft [Equal], Writing—review & editing [Equal]), Viviana Rodriguez (Methodology [Equal]), Monica Pattarroyo (Methodology [Equal]), Carla Michelle Ramirez (Methodology [Equal]), Trista Beard (Formal analysis [Equal], Writing—original draft [Equal], Writing—review & editing [Equal]), Daniel Soto (Writing—review & editing [Equal]), Scott Donaldson (Formal analysis [Equal], Writing—original draft [Equal], Writing—review & editing [Equal]), and Jennifer Unger (Funding acquisition [Equal], Validation [Equal], Writing—review & editing [Equal])

Data Availability

Data will be made available upon request.