Paper Session 23: Tobacco Control and Nicotine-Related Behavior

We use the nationally representative Tobacco Use Supplement to the Current Population Survey (TUS-CPS) with data from 1992 -2022 and project state-specific smoking prevalence to 2035 using Autoregressive Integrated Moving Average (ARIMA) models. The TUS-CPS also included up to 6 questions on each survey on which settings smoking should not be allowed. Across all the surveys, these items formed a scale, which we label a social norm scale, that has strong validity, and consistent psychometric properties, including high factor loadings and acceptable model fit indices (SRMSR ~0.05, CFI ~0.95-0.99).
In order to include all 50 states and the District of Columbia, a number of which have small populations, we grouped them into tertiles on their average smoking prevalence between 1992-2001. The tertile with the lowest prevalence (tertile 1) in the 1990s declined from 20.5% to 7.4% in 2022 while the tertile with the highest prevalence saw a decline from 26.3% to 12.7%. Projections indicate that all three tertiles will continue to decline substantially, although only tertile 1 will fall below the Healthy People target of 5%, and this only by 2035.
Over the duration of the TUS-CPS, state-specific social norms increased substantially over time in each state. In the 1990s, states with the highest prevalence had social norm scores a quarter standard deviation lower than those with the lowest prevalence and the average difference across tertiles narrowed over time. By 2022, the social norm gap between the tertiles had narrowed significantly.
Averaged over the 30-year period, state level social norm scores were highly correlated with smoking prevalence by tertile (r’s range =-0.88 to -0.91). Weighted regression models estimated a standard deviation increase in social norms was associated with an 13% reduction in smoking prevalence among states with the lowest prevalence, while states with the highest prevalence saw a 7% reduction. These findings suggest that stronger social norms may have played a major role in the remarkable reduction in smoking prevalence across the US population.
Limitations of this research include the fact that the TUS-CPS are repeated cross-sectional studies. We will outline additional research that is needed to demonstrate that changes in social norms are causally associated with the change in prevalence. The evidence so far suggests that this social norm scale may be an important surrogate marker for future changes in population prevalence.

Presenter - Matthew D. Stone, PhD,
University of California, San Diego

Co-Author - John P. Pierce, PhD, PhD, FSBM,
University of California, San Diego

Co-Author - Brian Q. Dang, MS,
University of California San Diego

Co-Author - Thet Khin, MPH,
University of California San Diego

Co-Author - Sara B. McMenamin, PhD,
University of California San Diego

Co-Author - Yuyan Shi, PhD, PhD,
University of California San Diego

Co-Author - Karen Messer, PhD,
University of California San Diego

Co-Author - Dennis R. Trinidad, PhD,
University of California San Diego

Background: Text messaging interventions for smoking cessation are widely disseminated across the U.S. However, their uptake when offered as part of a state quit line is understudied. Understanding demographic and regional factors influencing the use of these programs can help optimize their reach and effectiveness.
Methods: Individuals aged 18 and older who used the Virginia Quitline (VAQL) from 2018 to 2022 were analyzed. VAQL users were categorized into two groups: those who enrolled into Text2Quit (users) and those who did not enroll in Text2Quit but may have used another VAQL resource (non- users). VAQL also provided individual-level information (age, gender, race, ethnicity, education) County-level smoking rates, socioeconomic, and other features (e.g., cellular data plan coverage) were obtained from the Behavioral Risk Factor Surveillance System and the American Community Survey. Logistic regression analysis assessed the association of individual and county-level factors with Text2Quit enrollment.
Results: Of 14,959 VAQL users, 36.58% enrolled in Text2Quit. Text2Quit users were younger (54.18 vs. 57.08 years, P<0.001), more likely to have a college education (45.50% vs. 40.08%, P<0.001), and more likely to live in counties with higher smoking rates (14.89% vs. 13.94%, P<0.001), and greater cellular data coverage (71.58% vs. 66.41%, P<0.001). Adjusted logit regression models revealed that increasing age (OR=0.98, P<0.001), lower educational attainment (OR=0.84, P<0.001), and residence in Appalachian counties (OR=0.87, P=0.002) were associated with lower odds of Text2Quit enrollment. Conversely, higher county-level cellular data plan coverage (OR=1.04, P<0.001) was significantly associated with increased Text2Quit enrollment.
Conclusions: Text2Quit users differ from non-users in key demographic and regional aspects, with cellular data coverage being a significant factor in enrollment in this specific cessation program. These findings support the notion that technology, such as cellular data coverage are a health equity issue, particularly in rural areas such as Appalachia.

Presenter - Asal Pilehvari, PhD, PhD,
University of Virginia

Co-Author - Becca Anne Krukowski, PhD,
University of virginia

Co-Author - Kara Wiseman, PhD,
University of Virginia

Co-Author - Melissa A. Little, PhD,
University of Virginia

Background
Approximately 15.4 million Americans attempt smoking cessation, yet 83.6% of smokers relapse, highlighting the need to understand the underlying causes of relapse. Through repetition, smoking behaviors become associated with their environmental contexts (e.g., social settings, stressful situations), which can later trigger relapse during quit attempts. Designing effective interventions requires greater understanding of individuals’ smoking triggers, yet current self-report and wearable sensing tools only capture smoking episodes after the event – thus limiting the ability to capture smoking triggers in the moment.

We developed the first wearable system to detect smoking events in real-time using thermal sensors and simultaneously collect contextual data through ecological momentary assessment (EMA). We evaluate the reliability of our system, comprising a thermal-sensing necklace and paired smartwatch, through a free-living user study.

Methods
Eight smokers wore the necklace and smartwatch for 7 days. When the necklace predicted smoking (defined as 2-5 contiguous thermal pixels registering between 50-100 deg Celsius), an EMA was sent to the smartwatch prompting the participant to confirm or refute the prediction (“Are you smoking?” [yes/no]). A camera alongside the necklace thermal sensor recorded video, allowing us to visually confirm smoking events.

Results
Across all participants, 354.9 hours of sensor data were recorded, and 217 smoking episodes were captured. Of the 229 smoking predictions made by the necklace, 209 (91%) were correct and 20 (9%) were incorrect. 8 undetected smoking episodes were discovered in the RGB videos (3% of total episodes). Participants responded to 212 EMA alerts (92.6%). Of the 212 responses, 206 (97.2%) were correct (i.e., responding ‘yes’ if smoking and vice-versa).

Conclusions
Our results suggest our system is viable for automated, objective assessment of contextual triggers that underpin smoking relapse. We show accurate smoking prediction and high response rates to smoking-triggered EMAs. Future work will expand the EMAs to query contexts of smoking episodes, leverage the contextual data to design smartwatch-deliverable interventions, and explore automation of the EMA-intervention pipeline, which would yield a system adaptive to the individual differences that elude extant approaches to smoking cessation.

Presenter - Christopher Romano,
Northwestern University

Co-Author - Soroush Shahi,
Northwestern University

Co-Author - Glenn Fernandes,
Northwestern University

Co-Author - Annie Lin,
University of Minnesota

Co-Author - Jiayi Zheng,
Northwestern University

Co-Author - Brian L. Hitsman, PhD, FSBM, PhD, FSBM,
Northwestern University Feinberg School of Medicine

Co-Author - Tanmeet Butani,
Northwestern University

Co-Author - Nabil Alshurafa, PhD, PhD,
Northwestern University

Background: Nicotine and cannabis vaping are prevalent among young adults. Understanding real-time predictors of vaping is crucial to inform the development of Just-In-Time Adaptive Interventions (JITAI) for vaping cessation. We examined the associations between a variety of contextual factors with momentary vaping of nicotine and cannabis among young adults.
Methods: We collected ecological momentary assessments (EMAs) via smartphone app among California young adults (ages 18-29) who vaped nicotine and cannabis in 2023. Participants completed four random prompts each day for over 30 days. Three binary outcomes were defined as whether participants reported being about to vape nicotine, cannabis, or both substances (same-occasion co-vaping) on a given EMA. Contextual factors included craving for nicotine vaping/cannabis vaping, seeing tobacco/cannabis products/advertisements, mood, alcohol use, being with other people, and time of the day. We used mixed-effects logistic regression models to examine real-time predictors of each outcome, controlling for demographics and e-cigarette and cannabis dependence measured at baseline.
Results: Overall, 113 participants (mean age = 23.8 years, 62.5% female) completed 9,001 EMAs during a 30-day period. Of all the EMAs, 35.8% reported nicotine vaping, 9.3% reported cannabis vaping, and 6.3% reported same-occasion co-vaping of nicotine and cannabis. Similar predictors of all three vaping outcomes were using alcohol and staying alone or being with roommates/friends/a partner (vs. being with different people or strangers). Seeing tobacco advertisements was associated with higher odds of reporting nicotine vaping and co-vaping. Feeling happier was associated with higher odds of reporting co-vaping while feeling more stressed was associated with lower odds of reporting vaping nicotine or cannabis. Higher craving levels of a given substance were associated with an increased likelihood of vaping that substance or co-vaping. However, craving for cannabis vaping was associated with lower odds of reporting nicotine vaping. Cannabis vaping and co-vaping were more likely to occur in the afternoon and nighttime than in the morning, but this did not hold for vaping nicotine.
Conclusions: We found different contextual predictors that drive nicotine vaping, cannabis vaping, and same-occasion co-vaping among young adults. Our findings provide important targets for JITAI to help young adults quit vaping these substances.

Co-Author - Vuong Van Do,
University of California San Francisco

Co-Author - Pamela M. Ling,
University of California San Francisco

Co-Author - Salomeh Keyhani,
University of California San Francisco

Co-Author - Gregory Marcus,
University of California San Francisco

Presenter - Nhung Nguyen, PhD,
University of California San Francisco