Miscommunication about the US federal Tobacco 21 law: a content analysis of Twitter discussions.

OBJECTIVE: Tobacco 21 is a law that sets the minimum legal sales age of tobacco products to 21. On 20 December 2019, the USA passed a federal Tobacco 21 law. The objective of this study is to explore Twitter discussions about the federal Tobacco 21 law in the USA leading up to enacted. METHODS: Twitter messages about Tobacco 21 posted between September and December 2019 were collected via RITHM software. A 2% sample of all collected tweets were double coded by independent coders using a content analysis approach. RESULTS: Findings included three content categories of tweets (news, youth and young adults and methods of avoiding the law) with eight subcodes. Most news tweets incorrectly described the law as a purchase law (54.7%). However, Tobacco 21 is in fact a sales law-it only includes penalties for tobacco retailers who sell to under-age purchasers. About one-fourth (27%) of the tweets involved youth and young adults, with some claiming the law would reduce youth smoking and others doubting its ability to limit youth access to tobacco products. Few tweets (2.5%) mentioned methods of circumventing the policy, such as having an older peer purchase tobacco. CONCLUSIONS: As several countries explore raising their age of sale of tobacco laws to 21, they should couple policy enactment with clear and accurate communication about the law. Compliance agencies at all levels (eg, local, regional, national) can use social media to identify policy loopholes and support vulnerable populations throughout the policy implementation process.

Analyzing Twitter Chatter About Tobacco Use Within Intoxication-related Contexts of Alcohol Use: "Can Someone Tell Me Why Nicotine is So Fire When You're Drunk?"

INTRODUCTION: Alcohol and tobacco are commonly used together. Social influences within online social networking platforms contribute to youth and young adult substance use behaviors. This study used a sample of alcohol- and tobacco-related tweets to evaluate: (1) sentiment toward co-use of alcohol and tobacco, (2) increased susceptibility to tobacco use when consuming alcohol, and (3) the role of alcohol in contributing to a failed attempt to quit tobacco use. METHODS: Data were collected from the Twitter API from January 1, 2019 through December 31, 2019 using tobacco-related keywords (e.g., vape, ecig, smoking, juul*) and alcohol-related filters (e.g., drunk, blackout*). A total of 78,235 tweets were collected, from which a random subsample (n = 1,564) was drawn for coding. Cohen's Kappa values ranged from 0.66 to 0.99. RESULTS: Most tweets were pro co-use of alcohol and tobacco (75%). One of every ten tweets reported increased susceptibility to tobacco use when intoxicated. Non-regular tobacco users reported cravings for and tobacco use when consuming alcohol despite disliking tobacco use factors such as the taste, smell, and/or negative health effects. Regular tobacco users reported using markedly higher quantities of tobacco when intoxicated. Individuals discussed the role of alcohol undermining tobacco cessation attempts less often (2.0%), though some who had quit smoking for prolonged periods of time reported reinitiating tobacco use during acute intoxication episodes. CONCLUSIONS: Tobacco cessation interventions may benefit from including alcohol-focused components designed to educate participants about the association between increased susceptibility to tobacco use when consuming alcohol and the role of alcohol in undermining tobacco cessation attempts. IMPLICATIONS: Sentiment toward co-use of alcohol and tobacco on Twitter is largely positive. Individuals reported regret about using tobacco, or using more than intended, when intoxicated. Those who had quit smoking or vaping for prolonged periods of time reported reinitiating tobacco use when consuming alcohol. While social media-based tobacco cessation interventions like the Truth Initiative's "Ditch the Juul" campaign demonstrate potential to change tobacco use behaviors, these campaigns may benefit from including alcohol-focused components designed to educate participants about the association between increased susceptibility to tobacco use when consuming alcohol and the role of alcohol in undermining tobacco cessation attempts.

Puff Bars, Tobacco Policy Evasion, and Nicotine Dependence: Content Analysis of Tweets.

BACKGROUND: Puff Bars are e-cigarettes that continued marketing flavored products by exploiting the US Food and Drug Administration exemption for disposable devices. OBJECTIVE: This study aimed to examine discussions related to Puff Bar on Twitter to identify tobacco regulation and policy themes as well as unanticipated outcomes of regulatory loopholes. METHODS: Of 8519 original tweets related to Puff Bar collected from July 13, 2020, to August 13, 2020, a random 20% subsample (n=2661) was selected for qualitative coding of topics related to nicotine dependence and tobacco policy. RESULTS: Of the human-coded tweets, 2123 (80.2%) were coded as relevant to Puff Bar as the main topic. Of those tweets, 698 (32.9%) discussed tobacco policy, including flavors (n=320, 45.9%), regulations (n=124, 17.8%), purchases (n=117, 16.8%), and other products (n=110, 15.8%). Approximately 22% (n=480) of the tweets referenced dependence, including lack of access (n=273, 56.9%), appetite suppression (n=59, 12.3%), frequent use (n=47, 9.8%), and self-reported dependence (n=110, 22.9%). CONCLUSIONS: This study adds to the growing evidence base that the US Food and Drug Administration ban of e-cigarette flavors did not reduce interest, but rather shifted the discussion to brands utilizing a loophole that allowed flavored products to continue to be sold in disposable devices. Until comprehensive tobacco policy legislation is developed, new products or loopholes will continue to supply nicotine demand.

E-Cigarette-Related Nicotine Misinformation on Social Media.

Background. Twitter provides an opportunity to examine misperceptions about nicotine and addiction as they pertain to electronic nicotine delivery systems (ENDS). The purpose of this study was to systematically examine a sample of ENDS-related tweets that presented information about nicotine or addiction for the presence of potential misinformation.Methods. A total of 10.1 million ENDS-related tweets were obtained from April 2018 through March 2019 and were filtered for unique tweets containing keywords for nicotine and addiction. A subsample (n = 3,116) were human coded for type of account (individual, group, commercial, or news) and presence of potential misinformation.Results. Of tweets that presented ENDS-related nicotine or addiction information (n = 904), 41.7% (n = 377) contained potential misinformation coded as anti-vaping exaggeration, pro-vaping exaggeration, nicotine is not addictive or is never harmful, or unproven health benefits.Conclusions. Anti-vaping exaggeration tweets distorted or embellished claims about ENDS nicotine and addiction; pro-vaping exaggeration tweets misinterpreted results from scientific studies. Misinformation that nicotine is not addictive or is never harmful or has unproven health benefits appeared less but are potentially problematic. ENDS-related messaging should be designed to be easily understood by the public and monitored to detect the spread of misinterpretation or misinformation on social media.

Measures of Organizational Culture and Climate in Primary Care: a Systematic Review.

BACKGROUND: Primary care is increasingly contributing to improving the quality of patient care. This has imposed significant demands on clinicians with rising needs and limited resources. Organizational culture and climate have been found to be crucial in improving workforce well-being and hence quality of care. The objectives of this study are to identify organizational culture and climate measures used in primary care from 2008 to 2019 and evaluate their psychometric properties. METHODS: Data sources include PubMed, PsycINFO, HAPI, CINAHL, and Mental Measurements Yearbook. Bibliographies of relevant articles were reviewed and a cited reference search in Scopus was performed. Eligibility criteria include primary health care professionals, primary care settings, and use of measures representing the general concept of organizational culture and climate. Consensus-Based Standards for the selection of health Measurement Instruments (COSMIN) guidelines were followed to evaluate individual studies for methodological quality, rate results of measurement properties, qualitatively pool studies by measure, and grade evidence. RESULTS: Of 1745 initial studies, 42 studies met key study inclusion criteria, with 27 measures available for review (16 for organizational culture, 11 for organizational climate). There was considerable variability in measures, both conceptually and in psychometric quality. Many reported limited or no psychometric information. DISCUSSION: Notable measures selected for frequent use and strength and applicability of measurement properties include the Culture Questionnaire adapted for health care settings, Practice Culture Assessment, and Medical Group Practice Culture Assessment for organizational culture. Notable climate measures include the Nurse Practitioner Primary Care Organizational Climate Questionnaire, Practice Climate Survey, and Task and Relational Climate Scale. This synthesis and appraisal of organizational culture and climate measures can help investigators make informed decisions in choosing a measure or deciding to develop a new one. In terms of limitations, ratings should be considered conservative due to adaptations of the COSMIN protocol for clinician populations. PROSPERO REGISTRATION NUMBER: CRD 42019133117.

Machine Learning Classifiers for Twitter Surveillance of Vaping: Comparative Machine Learning Study.

BACKGROUND: Twitter presents a valuable and relevant social media platform to study the prevalence of information and sentiment on vaping that may be useful for public health surveillance. Machine learning classifiers that identify vaping-relevant tweets and characterize sentiments in them can underpin a Twitter-based vaping surveillance system. Compared with traditional machine learning classifiers that are reliant on annotations that are expensive to obtain, deep learning classifiers offer the advantage of requiring fewer annotated tweets by leveraging the large numbers of readily available unannotated tweets. OBJECTIVE: This study aims to derive and evaluate traditional and deep learning classifiers that can identify tweets relevant to vaping, tweets of a commercial nature, and tweets with provape sentiments. METHODS: We continuously collected tweets that matched vaping-related keywords over 2 months from August 2018 to October 2018. From this data set of tweets, a set of 4000 tweets was selected, and each tweet was manually annotated for relevance (vape relevant or not), commercial nature (commercial or not), and sentiment (provape or not). Using the annotated data, we derived traditional classifiers that included logistic regression, random forest, linear support vector machine, and multinomial naive Bayes. In addition, using the annotated data set and a larger unannotated data set of tweets, we derived deep learning classifiers that included a convolutional neural network (CNN), long short-term memory (LSTM) network, LSTM-CNN network, and bidirectional LSTM (BiLSTM) network. The unannotated tweet data were used to derive word vectors that deep learning classifiers can leverage to improve performance. RESULTS: LSTM-CNN performed the best with the highest area under the receiver operating characteristic curve (AUC) of 0.96 (95% CI 0.93-0.98) for relevance, all deep learning classifiers including LSTM-CNN performed better than the traditional classifiers with an AUC of 0.99 (95% CI 0.98-0.99) for distinguishing commercial from noncommercial tweets, and BiLSTM performed the best with an AUC of 0.83 (95% CI 0.78-0.89) for provape sentiment. Overall, LSTM-CNN performed the best across all 3 classification tasks. CONCLUSIONS: We derived and evaluated traditional machine learning and deep learning classifiers to identify vaping-related relevant, commercial, and provape tweets. Overall, deep learning classifiers such as LSTM-CNN had superior performance and had the added advantage of requiring no preprocessing. The performance of these classifiers supports the development of a vaping surveillance system.

Toward Real-Time Infoveillance of Twitter Health Messages.

There is growing interest in conducting public health research using data from social media. In particular, Twitter "infoveillance" has demonstrated utility across health contexts. However, rigorous and reproducible methodologies for using Twitter data in public health are not yet well articulated, particularly those related to content analysis, which is a highly popular approach. In 2014, we gathered an interdisciplinary team of health science researchers, computer scientists, and methodologists to begin implementing an open-source framework for real-time infoveillance of Twitter health messages (RITHM). Through this process, we documented common challenges and novel solutions to inform future work in real-time Twitter data collection and subsequent human coding. The RITHM framework allows researchers and practitioners to use well-planned and reproducible processes in retrieving, storing, filtering, subsampling, and formatting data for health topics of interest. Further considerations for human coding of Twitter data include coder selection and training, data representation, codebook development and refinement, and monitoring coding accuracy and productivity. We illustrate methodological considerations through practical examples from formative work related to hookah tobacco smoking, and we reference essential methods literature related to understanding and using Twitter data.

Comparison of toxicant load from waterpipe and cigarette tobacco smoking among young adults in the USA.

OBJECTIVES: To form population-level comparisons of total smoke volume, tar, carbon monoxide and nicotine consumed from waterpipe tobacco smoking (WTS) and cigarette smoking using data from a nationally representative sample of smokers and non-smokers aged 18-30 years. METHODS: In March and April 2013, we surveyed a nationally representative sample of 3254 US young adults to assess the frequency and volume of WTS and cigarette smoking. We used Monte Carlo analyses with 5000 repetitions to estimate the proportions of toxicants originating from WTS and cigarette smoking. Analyses incorporated survey weights and used recent meta-analytic data to estimate toxicant exposures associated with WTS and cigarette smoking. RESULTS: Compared with the additive estimates of WTS and cigarette smoking combined, 54.9% (95% CI 37.5% to 72.2%) of smoke volume was attributed to WTS. The proportions of tar attributable to WTS was 20.8% (95% CI 6.5% to 35.2%), carbon monoxide 10.3% (95% CI 3.3% to 17.3%) and nicotine 2.4% (95% CI 0.9% to 3.8%). CONCLUSIONS: WTS accounted for over half of the tobacco smoke volume consumed among young US adult waterpipe and cigarette smokers. Toxicant exposures to tar, carbon monoxide and nicotine were lower, but still substantial, for WTS alone compared with WTS and cigarette smoking. Public health and policy interventions to reduce harm from tobacco smoking in young US adults should explicitly address WTS toxicant exposures.

Use of Twitter to Assess Viewer Reactions to the Medical Drama, Code Black.

Fictional medical television programs are popular with viewers and have been shown to influence health-related outcomes. We sought to systematically analyze real-time viewer discourse on Twitter related to the new medical drama, Code Black. We retrieved all Twitter posts (tweets) and metadata around the time of the airing of Code Black for four consecutive weeks. We developed a codebook using both content assessment of Twitter messages (tweets) and theory-based variables used in entertainment education analyses. We coded all tweets that occurred during the Eastern Standard Time (EST) airing of the program. Tweets that fell into at least one coding category were further analyzed by two independent researchers. We collected a total of 19,369 tweets, with 54% of total tweets originating during the EST airing of the program. There were 1,888 tweets that fit into one or more of six broad coding categories. Qualitative analysis revealed several key themes including real-life motivation to pursue health sciences careers based on the program, engagement regarding medical accuracy, and respect for the nursing profession. Examination of tweets related to Code Black provides insight into viewer discourse and suggests that Twitter may provide a vehicle for leveraging program engagement into real-life discussion and inquiry.

The association between social media use and sleep disturbance among young adults.

INTRODUCTION: Many factors contribute to sleep disturbance among young adults. Social media (SM) use is increasing rapidly, and little is known regarding its association with sleep disturbance. METHODS: In 2014 we assessed a nationally representative sample of 1788 U.S. young adults ages 19-32. SM volume and frequency were assessed by self-reported minutes per day spent on SM (volume) and visits per week (frequency) using items adapted from the Pew Internet Research Questionnaire. We assessed sleep disturbance using the brief Patient-Reported Outcomes Measurement Information System (PROMIS®) sleep disturbance measure. Analyses performed in Pittsburgh utilized chi-square tests and ordered logistic regression using sample weights in order to estimate effects for the total U.S. RESULTS: In models that adjusted for all sociodemographic covariates, participants with higher SM use volume and frequency had significantly greater odds of having sleep disturbance. For example, compared with those in the lowest quartile of SM use per day, those in the highest quartile had an AOR of 1.95 (95% CI=1.37-2.79) for sleep disturbance. Similarly, compared with those in the lowest quartile of SM use frequency per week, those in the highest quartile had an AOR of 2.92 (95% CI=1.97-4.32) for sleep disturbance. All associations demonstrated a significant linear trend. DISCUSSION: The strong association between SM use and sleep disturbance has important clinical implications for the health and well-being of young adults. Future work should aim to assess directionality and to better understand the influence of contextual factors associated with SM use.

ASSOCIATION BETWEEN SOCIAL MEDIA USE AND DEPRESSION AMONG U.S. YOUNG ADULTS.

BACKGROUND: Social media (SM) use is increasing among U.S. young adults, and its association with mental well-being remains unclear. This study assessed the association between SM use and depression in a nationally representative sample of young adults. METHODS: We surveyed 1,787 adults ages 19 to 32 about SM use and depression. Participants were recruited via random digit dialing and address-based sampling. SM use was assessed by self-reported total time per day spent on SM, visits per week, and a global frequency score based on the Pew Internet Research Questionnaire. Depression was assessed using the Patient-Reported Outcomes Measurement Information System (PROMIS) Depression Scale Short Form. Chi-squared tests and ordered logistic regressions were performed with sample weights. RESULTS: The weighted sample was 50.3% female and 57.5% White. Compared to those in the lowest quartile of total time per day spent on SM, participants in the highest quartile had significantly increased odds of depression (AOR = 1.66, 95% CI = 1.14-2.42) after controlling for all covariates. Compared with those in the lowest quartile, individuals in the highest quartile of SM site visits per week and those with a higher global frequency score had significantly increased odds of depression (AOR = 2.74, 95% CI = 1.86-4.04; AOR = 3.05, 95% CI = 2.03-4.59, respectively). All associations between independent variables and depression had strong, linear, dose-response trends. Results were robust to all sensitivity analyses. CONCLUSIONS: SM use was significantly associated with increased depression. Given the proliferation of SM, identifying the mechanisms and direction of this association is critical for informing interventions that address SM use and depression.

Portrayal of alcohol intoxication on YouTube.

BACKGROUND: We aimed to characterize the content of leading YouTube videos related to alcohol intoxication and to examine factors associated with alcohol intoxication in videos that were assessed positively by viewers. METHODS: We systematically captured the 70 most relevant and popular videos on YouTube related to alcohol intoxication. We employed an iterative process to codebook development which resulted in 42 codes in 6 categories: video characteristics, character socio demographics, alcohol depiction, degree of alcohol use, characteristics associated with alcohol, and consequences of alcohol. RESULTS: There were a total of 333,246,875 views for all videos combined. While 89% of videos involved males, only 49% involved females. The videos had a median of 1,646 (interquartile range [IQR] 300 to 22,969) "like" designations and 33 (IQR 14 to 1,261) "dislike" designations each. Liquor was most frequently represented, followed by beer and then wine/champagne. Nearly one-half (44%) of videos contained a brand reference. Humor was juxtaposed with alcohol use in 79% of videos, and motor vehicle use was present in 24%. There were significantly more likes per dislike, indicating more positive sentiment, when there was representation of liquor (29.1 vs. 11.4, p = 0.008), brand references (32.1 vs. 19.2, p = 0.04), and/or physical attractiveness (67.5 vs. 17.8, p < 0.001). CONCLUSIONS: Internet videos depicting alcohol intoxication are heavily viewed. Nearly, half of these videos involve a brand-name reference. While these videos commonly juxtapose alcohol intoxication with characteristics such as humor and attractiveness, they infrequently depict negative clinical outcomes. The popularity of this site may provide an opportunity for public health intervention.

Characteristics and engagement among English-language online forums for addiction recovery available in the US.

In developing public resources for the Networks Enhancing Addiction Recovery - Forum Activity Roadmap (NEAR-FAR), we completed a systematic observational study of English-language online forums related to recovery from alcohol or other drug addiction in late 2021. Among 207 identified forums, the majority were classified as "general addiction" or alcohol-focused, though classifications related to other substances were common on websites hosting multiple forums. Commonly used social media platforms such as Reddit, Facebook, or Quora offered easily accessible venues for individuals seeking online support related to a variety of substances. Forums were related to established recovery programs such as 12-step and SMART Recovery as well as other nonprofit and for-profit recovery programs, and to community forums without formal recovery programming. Among 148 forums with any observed user activity, the median time between unique user engagements was 27 days (inter-quartile range: 2-74). Among 98 forums with past-month posting activity, we found a median of <10 posts per week (inter-quartile range: 1-78). This study compares three metrics of observed forum activity (posts per week, responses per post, time between unique user engagements) and operationalizes forum characteristics that may potentiate opportunities for enhanced engagement and social support in addiction recovery.

Characterizing online social support for alcohol use disorder: A mixed-methods approach.

BACKGROUND: Online social media communities are increasingly popular venues for discussing alcohol use disorder (AUD) and recovery. Little is known about distinct contexts of social support that are exchanged in this milieu, which are critical to understanding the social dynamics of online recovery support. METHODS: We randomly selected one post per day over the span of a year from the StopDrinking recovery forum. Direct responses to posts were double coded within an established theoretical framework of social support. Within a mixed-methods research framework, we quantified the linguistic characteristics of 1386 responses (i.e., text length, complexity, and sentiment) and qualitatively explored themes within and among different types of social support. RESULTS: Emotional support was most prevalent (74% of responses) and appeared as the sole form of support in 38% of responses. Emotionally supportive responses were significantly shorter, less complex, and more positively valenced than other support types. Appraisal support was also common in 55% of responses, while informational support was identified in only 17%. There was substantial overlap among support types, with 40% of responses including two or more types. Salient themes included the common use of community-specific acronyms in emotional support. Appraisal support conveyed feedback about attitudes and behaviors that are perceived as (un-) favorable for AUD recovery. Informational support responses were composed primarily of recommendations for self-help literature, clinical treatment approaches, and peer recovery programs. CONCLUSIONS: Social support in this sample was primarily emotional in nature, with other types of support included to provide feedback and guidance (i.e., appraisal support) and supplemental recovery resources (i.e., informational support). The provided social support framework can be helpful in characterizing community dynamics among heterogeneous online AUD recovery support forums. This framework could also be helpful in considering changes in support approaches that correspond to progress in recovery.

Twitter Sentiment About the US Federal Tobacco 21 Law: Mixed Methods Analysis.

BACKGROUND: On December 20, 2019, the US "Tobacco 21" law raised the minimum legal sales age of tobacco products to 21 years. Initial research suggests that misinformation about Tobacco 21 circulated via news sources on Twitter and that sentiment about the law was associated with particular types of tobacco products and included discussions about other age-related behaviors. However, underlying themes about this sentiment as well as temporal trends leading up to enactment of the law have not been explored. OBJECTIVE: This study sought to examine (1) sentiment (pro-, anti-, and neutral policy) about Tobacco 21 on Twitter and (2) volume patterns (number of tweets) of Twitter discussions leading up to the enactment of the federal law. METHODS: We collected tweets related to Tobacco 21 posted between September 4, 2019, and December 31, 2019. A 2% subsample of tweets (4628/231,447) was annotated by 2 experienced, trained coders for policy-related information and sentiment. To do this, a codebook was developed using an inductive procedure that outlined the operational definitions and examples for the human coders to annotate sentiment (pro-, anti-, and neutral policy). Following the annotation of the data, the researchers used a thematic analysis to determine emergent themes per sentiment category. The data were then annotated again to capture frequencies of emergent themes. Concurrently, we examined trends in the volume of Tobacco 21-related tweets (weekly rhythms and total number of tweets over the time data were collected) and analyzed the qualitative discussions occurring at those peak times. RESULTS: The most prevalent category of tweets related to Tobacco 21 was neutral policy (514/1113, 46.2%), followed by antipolicy (432/1113, 38.8%); 167 of 1113 (15%) were propolicy or supportive of the law. Key themes identified among neutral tweets were news reports and discussion of political figures, parties, or government involvement in general. Most discussions were generated from news sources and surfaced in the final days before enactment. Tweets opposing Tobacco 21 mentioned that the law was unfair to young audiences who were addicted to nicotine and were skeptical of the law's efficacy and importance. Methods used to evade the law were found to be represented in both neutral and antipolicy tweets. Propolicy tweets focused on the protection of youth and described the law as a sensible regulatory approach rather than a complete ban of all products or flavored products. Four spikes in daily volume were noted, 2 of which corresponded with political speeches and 2 with the preparation and passage of the legislation. CONCLUSIONS: Understanding themes of public sentiment-as well as when Twitter activity is most active-will help public health professionals to optimize health promotion activities to increase community readiness and respond to enforcement needs including education for retailers and the general public.

Policy and Behavior: Comparisons between Twitter Discussions about the US Tobacco 21 Law and Other Age-Related Behaviors.

To combat the e-cigarette epidemic among young audiences, a federal law was passed in the US that raised the minimum legal sales age of tobacco to 21 years (commonly known as Tobacco 21). Little is known about sentiment toward this law. Thus, the purpose of our study was to systematically explore trends about Tobacco 21 discussions and comparisons to other age-restriction behaviors on Twitter. Twitter data (n = 4628) were collected from September to December of 2019 that were related to Tobacco 21. A random subsample of identified tweets was used to develop a codebook. Two trained coders independently coded all data, with strong inter-rater reliability (κ = 0.71 to 0.93) found for all content categories. Associations between sentiment and content categories were calculated using χ2 analyses. Among relevant tweets (n = 955), the most common theme­the disjunction between ages for military enlistment and tobacco use­was found in 17.8% of all tweets. Anti-policy sentiment was strongly associated with the age of military enlistment, alcohol, voting, and adulthood (p < 0.001 for all). Opposition to Tobacco 21 propagates on social media because the US federal law does not exempt military members. However, the e-cigarette epidemic may have fueled some support for this law.

Classification of Twitter Vaping Discourse Using BERTweet: Comparative Deep Learning Study.

BACKGROUND: Twitter provides a valuable platform for the surveillance and monitoring of public health topics; however, manually categorizing large quantities of Twitter data is labor intensive and presents barriers to identify major trends and sentiments. Additionally, while machine and deep learning approaches have been proposed with high accuracy, they require large, annotated data sets. Public pretrained deep learning classification models, such as BERTweet, produce higher-quality models while using smaller annotated training sets. OBJECTIVE: This study aims to derive and evaluate a pretrained deep learning model based on BERTweet that can identify tweets relevant to vaping, tweets (related to vaping) of commercial nature, and tweets with provape sentiment. Additionally, the performance of the BERTweet classifier will be compared against a long short-term memory (LSTM) model to show the improvements a pretrained model has over traditional deep learning approaches. METHODS: Twitter data were collected from August to October 2019 using vaping-related search terms. From this set, a random subsample of 2401 English tweets was manually annotated for relevance (vaping related or not), commercial nature (commercial or not), and sentiment (positive, negative, or neutral). Using the annotated data, 3 separate classifiers were built using BERTweet with the default parameters defined by the Simple Transformer application programming interface (API). Each model was trained for 20 iterations and evaluated with a random split of the annotated tweets, reserving 10% (n=165) of tweets for evaluations. RESULTS: The relevance, commercial, and sentiment classifiers achieved an area under the receiver operating characteristic curve (AUROC) of 94.5%, 99.3%, and 81.7%, respectively. Additionally, the weighted F1 scores of each were 97.6%, 99.0%, and 86.1%, respectively. We found that BERTweet outperformed the LSTM model in the classification of all categories. CONCLUSIONS: Large, open-source deep learning classifiers, such as BERTweet, can provide researchers the ability to reliably determine if tweets are relevant to vaping; include commercial content; and include positive, negative, or neutral content about vaping with a higher accuracy than traditional natural language processing deep learning models. Such enhancement to the utilization of Twitter data can allow for faster exploration and dissemination of time-sensitive data than traditional methodologies (eg, surveys, polling research).

Acceptability and Efficacy of the SMARxT Media Literacy Education Program to Counter Pharmaceutical Marketing Influences among Medical Trainees.

Background: Evidence-based prescribing (EBP) results in decreased morbidity and reduces medical costs. However, pharmaceutical marketing influences medication requests and prescribing habits, which can detract from EBP. Media literacy, which teaches critical thinking, is a promising approach for buffering marketing influences and encouraging EBP. The authors developed the "SMARxT" media literacy education program around marketing influences on EBP decision-making. The program consisted of six videos and knowledge assessments that were delivered as an online educational intervention through the Qualtrics platform. Methods: In 2017, we assessed program feasibility, acceptability, and efficacy of enhancing knowledge among resident physicians at the University of Pittsburgh. Resident physicians (n=73) responded to pre-test items assessing prior knowledge, viewed six SMARxT videos, and responded to post-test items. A 6-month follow-up test was completed to quantitatively assess sustained changes in knowledge and to qualitatively assess summative feedback about the program (n=54). Test scores were assessed from pre- to post-test and from pre-test to follow-up using paired-sample t-tests. Qualitative results were synthesized through content analysis. Results: Proportion of correct knowledge responses increased from pre-test to immediate post-test (31% to 64%, P<0.001) at baseline. Correct responses also increased from pre-test to 6-month follow-up (31% to 43%, P<0.001). Feasibility was demonstrated by 95% of enrolled participants completing all baseline procedures and 70% completing 6-month follow-up. Quantitative measures of acceptability yielded positive scores and qualitative responses indicated participants' increased confidence in understanding and countering marketing influences due to the intervention. However, participants stated they would prefer shorter videos, feedback about test scores, and additional resources to reinforce learning objectives. Conclusion: The SMARxT media literacy program was efficacious and acceptable to resident physicians. Participant suggestions could be incorporated into a subsequent version of SMARxT and inform similar clinical education programs. Future research should assess program impact on real-world prescribing practices.

Discussions and Misinformation About Electronic Nicotine Delivery Systems and COVID-19: Qualitative Analysis of Twitter Content.

BACKGROUND: Misinformation and conspiracy theories related to COVID-19 and electronic nicotine delivery systems (ENDS) are increasing. Some of this may stem from early reports suggesting a lower risk of severe COVID-19 in nicotine users. Additionally, a common conspiracy is that the e-cigarette or vaping product use-associated lung injury (EVALI) outbreak of 2019 was actually an early presentation of COVID-19. This may have important public health ramifications for both COVID-19 control and ENDS use. OBJECTIVE: Twitter is an ideal tool for analyzing real-time public discussions related to both ENDS and COVID-19. This study seeks to collect and classify Twitter messages ("tweets") related to ENDS and COVID-19 to inform public health messaging. METHODS: Approximately 2.1 million tweets matching ENDS-related keywords were collected from March 1, 2020, through June 30, 2020, and were then filtered for COVID-19-related keywords, resulting in 67,321 original tweets. A 5% (n=3366) subsample was obtained for human coding using a systematically developed codebook. Tweets were coded for relevance to the topic and four overarching categories. RESULTS: A total of 1930 (57.3%) tweets were coded as relevant to the research topic. Half (n=1008, 52.2%) of these discussed a perceived association between ENDS use and COVID-19 susceptibility or severity, with 42.4% (n=818) suggesting that ENDS use is associated with worse COVID-19 symptoms. One-quarter (n=479, 24.8%) of tweets discussed the perceived similarity/dissimilarity of COVID-19 and EVALI, and 13.8% (n=266) discussed ENDS use behavior. Misinformation and conspiracy theories were present throughout all coding categories. CONCLUSIONS: Discussions about ENDS use and COVID-19 on Twitter frequently highlight concerns about the susceptibility and severity of COVID-19 for ENDS users; however, many contain misinformation and conspiracy theories. Public health messaging should capitalize on these concerns and amplify accurate Twitter messaging.

COVID-19 Social Distancing and Online Mutual Help Engagement for Alcohol Use Recovery.

INTRODUCTION: COVID-19 and associated social distancing has presented challenges for individuals engaging in face-to-face mutual help groups (MHGs) such as Alcoholics Anonymous for alcohol use recovery. Online MHGs may be particularly appealing to individuals with limited access or inclination to attend in-person MHGs. We examined engagement within the popular "StopDrinking" online MHG, hypothesizing that engagement would increase due to demand for virtual peer support as COVID-19 social distancing progressed. METHODS: We collected publicly available engagement data for StopDrinking from February 19, 2018 through April 30, 2020 while considering March and April of 2020 as months initially impacted by voluntary or mandated COVID-19 social distancing. Using seasonal autoregressive integrated moving average models, we predicted daily engagement for this social distancing time period based on all available engagement data collected before April 2020. Kalman filtering with 95% prediction limits was employed to define significant thresholds for observed data to reside within. RESULTS: All days of observed engagement in March and April 2020 were lower than corresponding predicted values. Observed engagement fell below the lower 95% prediction limit for 36% of days, with 15 days in March and 7 days in April having significantly lower than predicted engagement. CONCLUSIONS: Relatively low activity on StopDrinking may signal broader population trends of problematic alcohol use and recovery disengagement during the initial COVID-19 social distancing timeframe. Continued investigation of online MHGs is needed to understand their potential for monitoring population health trends and to understand how such groups might support alcohol use recovery in contexts of crisis and isolation.