Applying topic modelling and qualitative content analysis to identify and characterise ENDS product promotion and sales on Instagram.

BACKGROUND: Increased public health and regulatory scrutiny concerning the youth vaping epidemic has led to greater attention to promotion and sales of vaping products on social media platforms. OBJECTIVES: We used unsupervised machine learning to identify and characterise sale offers of electronic nicotine delivery systems (ENDS) and associated products on Instagram. We examined types of sellers, geographic ENDS location and use of age verification. METHODS: Our methodology was composed of three phases: data collection, topic modelling and content analysis. We used data mining approaches to query hashtags related to ENDS product use among young adults to collect Instagram posts. For topic modelling, we applied an unsupervised machine learning approach to thematically categorise and identify topic clusters associated with selling activity. Content analysis was then used to characterise offers for sale of ENDS products. RESULTS: From 70 725 posts, we identified 3331 engaged in sale of ENDS products. Posts originated from 20 different countries and were roughly split between individual (46.3%) and retail sellers (43.4%), with linked online sellers (8.8%) representing a smaller volume. ENDS products most frequently offered for sale were flavoured e-liquids (53.0%) and vaping devices (20.5%). Online sellers offering flavoured e-liquids were less likely to use age verification at point of purchase (29% vs 64%) compared with other products. CONCLUSIONS: Instagram is a global venue for unregulated ENDS sales, including flavoured products, and access to websites lacking age verification. Such posts may violate Instagram's policies and US federal and state law, necessitating more robust review and enforcement to prevent ENDS uptake and access.

Campus Smoking Policies and Smoking-Related Twitter Posts Originating From California Public Universities: Retrospective Study.

BACKGROUND: The number of colleges and universities with smoke- or tobacco-free campus policies has been increasing. The effects of campus smoking policies on overall sentiment, particularly among young adult populations, are more difficult to assess owing to the changing tobacco and e-cigarette product landscape and differential attitudes toward policy implementation and enforcement. OBJECTIVE: The goal of the study was to retrospectively assess the campus climate toward tobacco use by comparing tweets from California universities with and those without smoke- or tobacco-free campus policies. METHODS: Geolocated Twitter posts from 2015 were collected using the Twitter public application programming interface in combination with cloud computing services on Amazon Web Services. Posts were filtered for tobacco products and behavior-related keywords. A total of 42,877,339 posts were collected from 2015, with 2837 originating from a University of California or California State University system campus, and 758 of these manually verified as being about smoking. Chi-square tests were conducted to determine if there were significant differences in tweet user sentiments between campuses that were smoke- or tobacco-free (all University of California campuses and California State University, Fullerton) compared to those that were not. A separate content analysis of tweets included in chi-square tests was conducted to identify major themes by campus smoking policy status. RESULTS: The percentage of positive sentiment tweets toward tobacco use was higher on campuses without a smoke- or tobacco-free campus policy than on campuses with a smoke- or tobacco-free campus policy (76.7% vs 66.4%, P=.03). Higher positive sentiment on campuses without a smoke- or tobacco-free campus policy may have been driven by general comments about one's own smoking behavior and comments about smoking as a general behavior. Positive sentiment tweets originating from campuses without a smoke- or tobacco-free policy had greater variation in tweet type, which may have also contributed to differences in sentiment among universities. CONCLUSIONS: Our study introduces preliminary data suggesting that campus smoke- and tobacco-free policies are associated with a reduction in positive sentiment toward smoking. However, continued expressions and intentions to smoke and reports of one's own smoking among Twitter users suggest a need for more research to better understand the dynamics between implementation of smoke- and tobacco-free policies and resulting tobacco behavioral sentiment.

Characterizing alternative and emerging tobacco product transition of use behavior on Twitter.

OBJECTIVE: The objective of this study was to develop an inductive coding approach specific to characterizing user-generated social media conversations about transition of use of different tobacco and alternative and emerging tobacco products (ATPs). RESULTS: A total of 40,206 tweets were collected from the Twitter public API stream that were geocoded from 2018 to 2019. Using data mining approaches, these tweets were then filtered for keywords associated with tobacco and ATP use behavior. This resulted in a subset of 5718 tweets, with 657 manually annotated and identified as associated with user-generated conversations about tobacco and ATP use behavior. The 657 tweets were coded into 9 parent codes: inquiry, interaction, observation, opinion, promote, reply, share knowledge, use characteristics, and transition of use behavior. The highest number of observations occurred under transition of use (43.38%, n = 285), followed by current use (39.27%, n = 258), opinions about use (0.07%, n = 46), and product promotion (0.06%, n = 37). Other codes had less than ten tweets that discussed these themes. Results provide early insights into how social media users discuss topics related to transition of use and their experiences with different and emerging tobacco product use behavior.

Characterizing Self-Reported Tobacco, Vaping, and Marijuana-Related Tweets Geolocated for California College Campuses.

Introduction: College-aged youth are active on social media yet smoking-related social media engagement in these populations has not been thoroughly investigated. We sought to conduct an exploratory infoveillance study focused on geolocated data to characterize smoking-related tweets originating from California 4-year colleges on Twitter. Methods: Tweets from 2015 to 2019 with geospatial coordinates in CA college campuses containing smoking-related keywords were collected from the Twitter API stream and manually annotated for discussions about smoking product type, sentiment, and behavior. Results: Out of all tweets detected with smoking-related behavior, 46.7% related to tobacco use, 50.0% to marijuana, and 7.3% to vaping. Of these tweets, 46.1% reported first-person use or second-hand observation of smoking behavior. Out of 962 tweets with user sentiment, the majority (67.6%) were positive, ranging from 55.0% for California State University, Long Beach to 95.8% for California State University, Los Angeles. Discussion: We detected reporting of first- and second-hand smoking behavior on CA college campuses representing possible violation of campus smoking bans. The majority of tweets expressed positive sentiment about smoking behaviors, though there was appreciable variability between college campuses. This suggests that anti-smoking outreach should be tailored to the unique student populations of these college communities. Conclusion: Among tweets about smoking from California colleges, high levels of positive sentiment suggest that the campus climate may be less receptive to anti-smoking messages or adherence to campus smoking bans. Further research should investigate the degree to which this varies by campuses over time and following implementation of bans including validating using other sources of data.

Big Data, Natural Language Processing, and Deep Learning to Detect and Characterize Illicit COVID-19 Product Sales: Infoveillance Study on Twitter and Instagram.

BACKGROUND: The coronavirus disease (COVID-19) pandemic is perhaps the greatest global health challenge of the last century. Accompanying this pandemic is a parallel "infodemic," including the online marketing and sale of unapproved, illegal, and counterfeit COVID-19 health products including testing kits, treatments, and other questionable "cures." Enabling the proliferation of this content is the growing ubiquity of internet-based technologies, including popular social media platforms that now have billions of global users. OBJECTIVE: This study aims to collect, analyze, identify, and enable reporting of suspected fake, counterfeit, and unapproved COVID-19-related health care products from Twitter and Instagram. METHODS: This study is conducted in two phases beginning with the collection of COVID-19-related Twitter and Instagram posts using a combination of web scraping on Instagram and filtering the public streaming Twitter application programming interface for keywords associated with suspect marketing and sale of COVID-19 products. The second phase involved data analysis using natural language processing (NLP) and deep learning to identify potential sellers that were then manually annotated for characteristics of interest. We also visualized illegal selling posts on a customized data dashboard to enable public health intelligence. RESULTS: We collected a total of 6,029,323 tweets and 204,597 Instagram posts filtered for terms associated with suspect marketing and sale of COVID-19 health products from March to April for Twitter and February to May for Instagram. After applying our NLP and deep learning approaches, we identified 1271 tweets and 596 Instagram posts associated with questionable sales of COVID-19-related products. Generally, product introduction came in two waves, with the first consisting of questionable immunity-boosting treatments and a second involving suspect testing kits. We also detected a low volume of pharmaceuticals that have not been approved for COVID-19 treatment. Other major themes detected included products offered in different languages, various claims of product credibility, completely unsubstantiated products, unapproved testing modalities, and different payment and seller contact methods. CONCLUSIONS: Results from this study provide initial insight into one front of the "infodemic" fight against COVID-19 by characterizing what types of health products, selling claims, and types of sellers were active on two popular social media platforms at earlier stages of the pandemic. This cybercrime challenge is likely to continue as the pandemic progresses and more people seek access to COVID-19 testing and treatment. This data intelligence can help public health agencies, regulatory authorities, legitimate manufacturers, and technology platforms better remove and prevent this content from harming the public.

Machine Learning to Detect Self-Reporting of Symptoms, Testing Access, and Recovery Associated With COVID-19 on Twitter: Retrospective Big Data Infoveillance Study.

BACKGROUND: The coronavirus disease (COVID-19) pandemic is a global health emergency with over 6 million cases worldwide as of the beginning of June 2020. The pandemic is historic in scope and precedent given its emergence in an increasingly digital era. Importantly, there have been concerns about the accuracy of COVID-19 case counts due to issues such as lack of access to testing and difficulty in measuring recoveries. OBJECTIVE: The aims of this study were to detect and characterize user-generated conversations that could be associated with COVID-19-related symptoms, experiences with access to testing, and mentions of disease recovery using an unsupervised machine learning approach. METHODS: Tweets were collected from the Twitter public streaming application programming interface from March 3-20, 2020, filtered for general COVID-19-related keywords and then further filtered for terms that could be related to COVID-19 symptoms as self-reported by users. Tweets were analyzed using an unsupervised machine learning approach called the biterm topic model (BTM), where groups of tweets containing the same word-related themes were separated into topic clusters that included conversations about symptoms, testing, and recovery. Tweets in these clusters were then extracted and manually annotated for content analysis and assessed for their statistical and geographic characteristics. RESULTS: A total of 4,492,954 tweets were collected that contained terms that could be related to COVID-19 symptoms. After using BTM to identify relevant topic clusters and removing duplicate tweets, we identified a total of 3465 (<1%) tweets that included user-generated conversations about experiences that users associated with possible COVID-19 symptoms and other disease experiences. These tweets were grouped into five main categories including first- and secondhand reports of symptoms, symptom reporting concurrent with lack of testing, discussion of recovery, confirmation of negative COVID-19 diagnosis after receiving testing, and users recalling symptoms and questioning whether they might have been previously infected with COVID-19. The co-occurrence of tweets for these themes was statistically significant for users reporting symptoms with a lack of testing and with a discussion of recovery. A total of 63% (n=1112) of the geotagged tweets were located in the United States. CONCLUSIONS: This study used unsupervised machine learning for the purposes of characterizing self-reporting of symptoms, experiences with testing, and mentions of recovery related to COVID-19. Many users reported symptoms they thought were related to COVID-19, but they were not able to get tested to confirm their concerns. In the absence of testing availability and confirmation, accurate case estimations for this period of the outbreak may never be known. Future studies should continue to explore the utility of infoveillance approaches to estimate COVID-19 disease severity.