Scalable Surveillance of E-Cigarette Products on Instagram and TikTok Using Computer Vision

INTRODUCTION

Instagram and TikTok, video-based social media platforms popular among adolescents, contain tobacco-related content despite the platforms' policies prohibiting substance-related posts. Prior research identified themes in e-cigarette-related social media posts using qualitative or text-based machine learning methods. We developed an image-based computer vision model to identify e-cigarette products in social media images and videos.

AIMS AND METHODS

We created a data set of 6999 Instagram images labeled for 8 object classes: mod or pod devices, e-juice containers, packaging boxes, nicotine warning labels, e-juice flavors, e-cigarette brand names, and smoke clouds. We trained a DyHead object detection model using a Swin-Large backbone, evaluated the model's performance on 20 Instagram and TikTok videos, and applied the model to 14 072 e-cigarette-related promotional TikTok videos (2019-2022; 10 276 485 frames).

RESULTS

The model achieved the following mean average precision scores on the image test set: e-juice container: 0.89; pod device: 0.67; mod device: 0.54; packaging box: 0.84; nicotine warning label: 0.86; e-cigarette brand name: 0.71; e-juice flavor name: 0.89; and smoke cloud: 0.46. The prevalence of pod devices in promotional TikTok videos increased by 15% from 2019 to 2022. The prevalence of e-juices increased by 33% from 2021 to 2022. The prevalence of e-juice flavor names and e-cigarette brand names increased by about 100% from 2019 to 2022.

CONCLUSIONS

Deep learning-based object detection technology enables automated analysis of visual posts on social media. Our computer vision model can detect the presence of e-cigarettes products in images and videos, providing valuable surveillance data for tobacco regulatory science (TRS).

IMPLICATIONS

Prior research identified themes in e-cigarette-related social media posts using qualitative or text-based machine learning methods. We developed an image-based computer vision model to identify e-cigarette products in social media images and videos. We trained a DyHead object detection model using a Swin-Large backbone, evaluated the model's performance on 20 Instagram and TikTok videos featuring at least two e-cigarette objects, and applied the model to 14 072 e-cigarette-related promotional TikTok videos (2019-2022; 10 276 485 frames). The deep learning model can be used for automated, scalable surveillance of image- and video-based e-cigarette-related promotional content on social media, providing valuable data for TRS. Social media platforms could use computer vision to identify tobacco-related imagery and remove it promptly, which could reduce adolescents' exposure to tobacco content online.

Association of over the counter "hangover remedy" use with alcohol use problems and consumption patterns among young adults

This cross-sectional study of young adults examined associations of hangover remedy use with alcohol use problems. Results suggest that ever-use of hangover remedy products was positively associated with alcohol use problem score, drinks per typical drinking day, and alcohol use disorder symptom count. Use of hangover remedies among young adults merits further scientific and regulatory attention.

Are the Relevant Risk Factors Being Adequately Captured in Empirical Studies of Smoking Initiation? A Machine Learning Analysis Based on the Population Assessment of Tobacco and Health Study

INTRODUCTION

Cigarette smoking continues to pose a threat to public health. Identifying individual risk factors for smoking initiation is essential to further mitigate this epidemic. To the best of our knowledge, no study today has used machine learning (ML) techniques to automatically uncover informative predictors of smoking onset among adults using the Population Assessment of Tobacco and Health (PATH) study.

AIMS AND METHODS

In this work, we employed random forest paired with Recursive Feature Elimination to identify relevant PATH variables that predict smoking initiation among adults who have never smoked at baseline between two consecutive PATH waves. We included all potentially informative baseline variables in wave 1 (wave 4) to predict past 30-day smoking status in wave 2 (wave 5). Using the first and most recent pairs of PATH waves was found sufficient to identify the key risk factors of smoking initiation and test their robustness over time. The eXtreme Gradient Boosting method was employed to test the quality of these selected variables.

RESULTS

As a result, classification models suggested about 60 informative PATH variables among many candidate variables in each baseline wave. With these selected predictors, the resulting models have a high discriminatory power with the area under the specificity-sensitivity curves of around 80%. We examined the chosen variables and discovered important features. Across the considered waves, two factors, (1) BMI, and (2) dental and oral health status, robustly appeared as important predictors of smoking initiation, besides other well-established predictors.

CONCLUSIONS

Our work demonstrates that ML methods are useful to predict smoking initiation with high accuracy, identifying novel smoking initiation predictors, and to enhance our understanding of tobacco use behaviors.

IMPLICATIONS

Understanding individual risk factors for smoking initiation is essential to prevent smoking initiation. With this methodology, a set of the most informative predictors of smoking onset in the PATH data were identified. Besides reconfirming well-known risk factors, the findings suggested additional predictors of smoking initiation that have been overlooked in previous work. More studies that focus on the newly discovered factors (BMI and dental and oral health status,) are needed to confirm their predictive power against the onset of smoking as well as determine the underlying mechanisms.

Restricting Tobacco Retail Outlets Around Middle and High Schools as a Way to Reduce Tobacco Marketing Disparities: A Simulation Study

INTRODUCTION

Communities with more people of color and economically disadvantaged residents are disproportionately exposed to tobacco marketing from tobacco companies. This study examined if banning tobacco retail outlets (TROs) within 1000 ft of schools would reduce these marketing disparities through a greater reduction in the amount of tobacco advertising around schools in these communities.

METHODS

Data from objectively audited advertisement data from 106 convenience stores and gas stations around 42 middle and high schools located in the four major metropolitan areas of Texas were linked with schools' enrollment data. ArcGIS (Aeronautical Reconnaissance Coverage Geographic Information System) was used to simulate a 1000-ft ban of tobacco sales around the schools. Independent sample T-tests and Mann-Whitney U tests were used to test mean differences where appropriate.

RESULTS

Schools with a higher enrollment of Hispanic/Latino (mean = 171.6, SD = 96.9) and economically disadvantaged students (mean = 168.9, SD = 102.3) were surrounded with significantly greater advertising at TROs than schools with lower enrollment of these groups (mean = 82.8, SD = 49.1 and mean = 89.2, SD = 50.6, respectively). A simulated 1000 ft ban of TROs around schools led to greater advertising reduction around schools with a higher enrollment of Hispanic/Latino students (13.3%-29.4% reductions) in comparison to schools with lower Hispanic/Latino student enrollment. However, the more economically disadvantaged schools had a smaller reduction in the number of advertisements (5.9%-21.9% reductions) in comparison to schools with less economically disadvantaged students.

CONCLUSION

The implementation of a ban of tobacco sales at TROs within 1000 ft of schools is one policy approach to reduce youth exposure to tobacco marketing, particularly among students of color.

STUDY IMPLICATION

Tobacco retail outlets (TROs) around schools with a higher enrollment of Hispanic/Latino and economically disadvantaged students had significantly more tobacco advertisements in comparison to schools with lower enrollment of these student groups. A simulated ban of TROs within 1000 ft of schools led to greater advertising reduction around schools with a higher enrollment of Hispanic/Latino students. For schools with more economically disadvantaged students, the ban led to a smaller reduction in advertisements in comparison to schools with less economically disadvantaged students. This proposed place-based strategy could be a successful means to reduce tobacco advertising and marketing disparity among communities of color.

How Smokers of Menthol Cigarettes and Flavored Cigars Might Respond to FDA's Proposed Bans

INTRODUCTION

A ban on menthol cigarettes and flavored cigars may reduce smoking and tobacco-related disparities.

AIMS AND METHODS

We aimed to examine the response of current smokers to a hypothetical ban on menthol cigarettes and flavored cigars. Current smokers were recruited online and reported the alternative products that they may switch to under a hypothetical ban, if they would try to obtain the banned products from illicit channels, and their support to the ban.

RESULTS

51% of current smokers would use nonflavored cigarettes and cigars as alternatives, 45% would switch to flavored heated tobacco products or e-cigarettes or quit smoking. 17% would try to obtain the banned flavored products from illicit markets. A majority of menthol only smokers opposed the ban. Daily smokers would be more likely to switch to nonflavored smoking, to try illicit market products, and were less supportive of the ban. Black smokers would be less likely to switch to nonflavored smoking and were more supportive of the ban. Smokers who used menthol cigarettes only would be more likely to switch to nonflavored smoking, less likely to try illicit market sellers, and were the least supportive of the ban.

CONCLUSIONS

In response to a ban of all added flavors for cigarettes or cigars, nearly half of the current smokers would quit smoking, largely by switching to nonsmoking products. However, smokers with more chronic use and those who used only menthol cigarettes would be more likely to switch to nonflavored smoking, diminishing the harm reduction potential. The ban may decrease the relatively higher prevalence of menthol cigarette smoking among Blacks compared with other groups.

IMPLICATIONS

A ban on the sale of menthol cigarettes and flavored cigars may decrease the prevalence of smoking because some current smokers may quit smoking and switch to nonsmoking products. However, smokers with more chronic use and those who used menthol cigarettes only were more likely to switch to nonflavored cigarettes or cigars, diminishing the harm reduction potential of the ban. Black smokers would be more likely to switch to products other than cigarettes and cigars thus decreasing their relatively higher prevalence of smoking compared with other groups.

Identifying Promising Themes for Adolescent Vaping Warnings: A National Experiment

INTRODUCTION

Adolescent vaping remains a problem in the United States, yet little is known about what health warning themes most discourage vaping among adolescents. We sought to identify the most compelling themes for vaping warnings for US adolescents.

METHODS

Participants were a national probability sample of 623 US adolescents aged 13-17 years, recruited in the summer of 2020. Adolescents were randomized to one of the five warning message themes about the potential health effects of vaping: 1. chemical harms, 2. lung harms, 3. Coronavirus Disease 2019 (COVID-19) harms, 4. nicotine addiction, or 5. control (messages about vape litter). The primary outcome was perceived message effectiveness (PME; 3-item scale). Secondary outcomes were negative affect (fear), attention, anticipated social interactions, and message novelty.

RESULTS

Adolescents rated the chemical, lung, and COVID-19 harms warning messages higher on PME than nicotine addiction and control (all p < .05), while nicotine addiction was rated higher than control (p < .05). The chemical, lung, and COVID-19 harms warning themes also elicited greater negative affect than nicotine addiction and control (all p < .05). For all other secondary outcomes, the COVID-19 harms warning message theme was rated higher than nicotine addiction and control (all p < .05).

CONCLUSION

Adolescents perceived warning message themes about lung, chemical and COVID-19 health effects of vaping as more effective than nicotine addiction. To discourage vaping, the FDA and others should communicate to youth about the health effects of vaping beyond nicotine addiction.

IMPLICATIONS

Adolescents rated warning message themes about the lung, chemical, and COVID-19 health effects of vaping as more effective than nicotine addiction, while nicotine addiction was rated as more effective than control themes about vaping litter. To discourage vaping among adolescents, health messaging should expand message themes to communicate about a broader set of health effects of vaping beyond nicotine addiction.

A Magic Bullet? The Potential Impact of E-Cigarettes on the Toll of Cigarette Smoking

INTRODUCTION

We examine the proportion of US smoking-produced mortality that e-cigarettes might eliminate under assumptions regarding vaping's ability to increase smoking cessation, vaping's health risks, and the possibility that vaping will increase smoking among young people.

METHODS

We employ a dynamic population simulation model that tracks individuals from ages 0 to 110, differentiated by gender and smoking status. Using data from the US Census, the National Vital Statistics Reports, Cancer Prevention Study II, and the National Health Interview Survey, we estimate the number of smoking-related life-years lost (LYL) from 2018 to 2100 in a no-vaping scenario. We then compare results for model runs that assess the impact of vaping under a variety of assumptions.

RESULTS

The combination of assumptions produces 360 possible scenarios. 357 (99%) yield positive estimates of life-years saved (LYS) due to vaping by 2100, from 143 000 to 65 million. Most scenarios result in millions of individuals quitting smoking due to vaping. On average, vaping-induced quitters gain an extra 1.2-2.0 years of life compared to smokers who quit without vaping. The impact of vaping is greatest when it most helps smokers who otherwise have the greatest difficulty quitting smoking. While the numbers of LYS are generally large across all scenarios, they often represent a small fraction of the toll of smoking.

CONCLUSIONS

Vaping is highly likely to reduce smoking-produced mortality. Still, vaping is not "the" answer to the public health crisis created by smoking. Rather, it may well be a tool to add to the armamentarium of effective tobacco control measures.

IMPLICATIONS

E-cigarettes hold the potential to reduce cigarette smoking's enormous toll. By itself, however, tobacco harm reduction, as embodied in vaping, is no magic bullet. Going forward, tobacco control will require vigilant application of the evidence-based measures that have brought us so much success in combatting smoking. It will require, as well, the search for and adoption of novel means of attacking the remaining problem. Harm reduction can, and many would say should, be a part of the complex formula that will eventually bring about the demise of smoking.

Quantification of Flavorants and Nicotine in Waterpipe Tobacco and Mainstream Smoke and Comparison to E-cigarette Aerosol

INTRODUCTION

Waterpipe use remains popular among youth with the availability of flavored shisha tobacco being one of the main drivers of waterpipe use. Although waterpipe mainstream toxicant emissions are well understood, less is known about the carryover of flavorants such as vanillin, benzaldehyde, and eugenol. In this study, flavored waterpipe tobacco was analyzed for flavorants and nicotine, and subsequent carryover to mainstream smoke.

METHODS

Flavorants vanillin, benzaldehyde, and eugenol, and nicotine were quantified in vanilla-, cherry-, and cinnamon-flavored shisha tobacco by gas chromatography/flame ionization detector and subsequently in waterpipe mainstream smoke generated by a smoking machine. The setup allowed for sampling before and after the water-filtration step.

RESULTS

Flavorant and nicotine content in smoke was reduced 3- to 10-fold and 1.4- to 3.1-fold, respectively, due to water filtration. Per-puff content of filtered waterpipe mainstream smoke ranged from 13 to 46 µg/puff for nicotine and from 6 to 55 µg/puff for flavorants.

CONCLUSIONS

Although water filtration reduced flavor and nicotine content in waterpipe mainstream smoke, the detected flavorant concentrations were similar or higher to those previously reported in e-cigarette aerosol. Therefore, users could be drawn to waterpipes due to similar flavor appeal as popular e-cigarette products. Absolute nicotine content of waterpipe smoke was lower than in e-cigarette aerosol, but the differential use patterns of waterpipe (>100 puffs/session) and e-cigarette (mostly <10 puffs/session, multiple session throughout the day) probably result in higher flavorant and nicotine exposure during a waterpipe session. Strategies to reduce youth introduction and exposure to nicotine via waterpipe use may consider similar flavor restrictions as those for e-cigarettes.

IMPLICATIONS

Although waterpipe mainstream smoke is well characterized for toxicants content, little is known about carryover of molecules relevant for appeal and addiction: flavorants and nicotine. This study shows that flavorant content of waterpipe mainstream smoke is comparable or higher than e-cigarette aerosol flavorant content. Regulatory action to address tobacco use behaviors targeting the availability of flavors should also include other tobacco products such as flavored shisha tobacco.

Predicting age groups of Twitter users based on language and metadata features

Health organizations are increasingly using social media, such as Twitter, to disseminate health messages to target audiences. Determining the extent to which the target audience (e.g., age groups) was reached is critical to evaluating the impact of social media education campaigns. The main objective of this study was to examine the separate and joint predictive validity of linguistic and metadata features in predicting the age of Twitter users. We created a labeled dataset of Twitter users across different age groups (youth, young adults, adults) by collecting publicly available birthday announcement tweets using the Twitter Search application programming interface. We manually reviewed results and, for each age-labeled handle, collected the 200 most recent publicly available tweets and user handles’ metadata. The labeled data were split into training and test datasets. We created separate models to examine the predictive validity of language features only, metadata features only, language and metadata features, and words/phrases from another age-validated dataset. We estimated accuracy, precision, recall, and F1 metrics for each model. An L1-regularized logistic regression model was conducted for each age group, and predicted probabilities between the training and test sets were compared for each age group. Cohen’s d effect sizes were calculated to examine the relative importance of significant features. Models containing both Tweet language features and metadata features performed the best (74% precision, 74% recall, 74% F1) while the model containing only Twitter metadata features were least accurate (58% precision, 60% recall, and 57% F1 score). Top predictive features included use of terms such as “school” for youth and “college” for young adults. Overall, it was more challenging to predict older adults accurately. These results suggest that examining linguistic and Twitter metadata features to predict youth and young adult Twitter users may be helpful for informing public health surveillance and evaluation research.

Tobacco Use, Insulin Resistance, and Risk of Type 2 Diabetes: Results from the Multi-Ethnic Study of Atherosclerosis

INTRODUCTION

Tobacco use is associated with insulin resistance and incident diabetes. Given the racial/ethnic differences in smoking patterns and incident type 2 diabetes our objective was to evaluate the association between tobacco use and insulin resistance (IR) as well as incident type 2 diabetes mellitus in a contemporary multiethnic cohort.

METHODS AND RESULTS

We studied 5,931 Multi- Ethnic Study of Atherosclerosis (MESA) participants who at baseline were free of type 2 diabetes (fasting glucose ≥7.0 mmol/l (126 mg/dl) and/or use of insulin or oral hypoglycemic medications) categorized by self-reported tobacco status and reclassified by urinary cotinine (available in 58% of participants) as never, current or former tobacco users. The association between tobacco use, IR (fasting plasma glucose, insulin, and the homeostatic model assessment of insulin resistance (HOMA-IR)) and incident diabetes over 10 years was evaluated using multivariable linear regression and Cox proportional hazards models, respectively. Mean age of the participants was 62 (±10) years, 46% were male, 41% Caucasian, 12% Chinese, 26% African American and 21% Hispanic/Latino. IR biomarkers did not significantly differ between current, former, and never cigarette users (P >0.10) but showed limited unadjusted differences for users of cigar, pipe and smokeless tobacco (All P <0.05). Fully adjusted models showed no association between dose or intensity of tobacco exposure and any index of IR. When stratified into participants that quit smoking vs. those who continued smoking during the 10-year study there was no difference in serum glucose levels or frequency of diabetes. In fully adjusted models, there was no significant difference in diabetes risk between former or current cigarette smokers compared to never smokers [HR (95% CI) 1.02 (0.77,1.37) and 0.81 (0.52,1.26) respectively].

CONCLUSION

In a contemporary multi-ethnic cohort, there was no independent association between tobacco use and IR or incident type 2 diabetes. The role smoking plays in causing diabetes may be more complicated than originally thought and warrants more in-depth large contemporary multi-ethnic studies.