CProject

Journal of Youth and Adolescence (2019) 48:1899–1911 https://doi.org/10.1007/s10964-019-01106-y

EMPIRICAL RESEARCH

Schools Influence Adolescent E-Cigarette use, but when? Examining the Interdependent Association between School Context and Teen Vaping over time

Adam M. Lippert 1 ● Daniel J. Corsi2 ● Grace E. Venechuk3

Received: 12 June 2019 / Accepted: 2 August 2019 / Published online: 24 August 2019 © Springer Science+Business Media, LLC, part of Springer Nature 2019

Abstract Schools are important contexts for adolescent health and health-risk behaviors, but how stable is this relationship? We develop a conceptual model based on Ecological Systems Theory describing the changing role of schools for adolescent health outcomes—in this case, teen e-cigarette use. To examine this change, we fit Bayesian multilevel regression models to two-year intervals of pooled cross-sectional data from the 2011–2017 U.S. National Youth Tobacco Survey, a school-based study of the nicotine use behaviors of roughly 65,000 middle and high school students (49.5% female; 41.1% nonwhite; x̄ age of 14.6 ranging from 9 to 18) from over 700 schools. We hypothesized that school-level associations with student e- cigarette use diminished over time as the broader popularity of e-cigarettes increased. Year-specific variance partitioning coefficients (VPC) derived from the multilevel models indicated a general decrease in the extent to which e-cigarette use clusters within specific schools, suggesting that students across schools became more uniform in their propensity to vape over the study period. This is above and beyond adjustments for personal characteristics and vicarious exposure to smoking via friends and family. Across all years, model coefficients indicate a positive association between attending schools where vaping is more versus less common and student-level odds of using e-cigarettes, suggesting that school contexts are still consequential to student vaping, but less so than when e-cigarettes were first introduced to the US market. These findings highlight how the health implications of multiply-embedded ecological systems like schools shift over time with concomitant changes in other ecological features including those related to policy, culture, and broader health practices within society. Though not uniformly reported in multilevel studies, variance partitioning coefficients could be used more thoughtfully to empirically illustrate how the influence of multiple developmentally-relevant contexts shift in their influence on teen health over time.

Keywords Adolescence ● Schools ● E-Cigarettes ● Multilevel

Introduction

Use of electronic cigarettes (“e-cigarettes”) among adoles- cents has rapidly proliferated. Between 2011 and 2015, the prevalence of current e-cigarette use climbed from 0.6% to 5.3% among US middle schoolers, and from 1.5% to 16% among high schoolers (Singh et al. 2016a). More recent estimates suggest continued increases from 2015-2018 with the widening popularity of novel “vaping” modalities including Juul devices (Gentzke et al. 2019). These patterns have emerged alongside historic lows in conventional tobacco use among teens who now favor vaping over smoking (Johnston et al. 2016). Conventional smoking remains among the most powerful correlates of adolescent e-cigarette use (Lippert 2018), though evidence suggests a

These authors contributed equally: Adam M. Lippert and Daniel J. Corsi

* Adam M. Lippert adam.lippert@ucdenver.edu

1 University of Colorado Denver, Sociology Department, 1380 Lawrence Street, Suite 420, Denver, CO 80204, USA

2 Ottawa Hospital Research Institute, Clinical Epidemiology Program, 501 Smyth Box 241, Ottawa, ON KIH 8L6, USA

3 University of Wisconsin-Madison, Sociology Department, 1180 Observatory Dr, Madison, WI 53706, USA

Supplementary information The online version of this article (https:// doi.org/10.1007/s10964-019-01106-y) contains supplementary material, which is available to authorized users.

1 2 3 4 5 6 7 8 9 0 () ;, :

12 34 56 78 90 () ;,:

bi-directional relationship with e-cigarette use being linked to a higher probability of future smoking, even among tobacco-abstinent teens (Bunnell et al. 2015; Coleman et al. 2014; Leventhal et al. 2015; Wills et al. 2015). In addition to the health risks associated with vaping (Chun et al. 2017; El-Hellani et al. 2018; Larcombe et al. 2017; Moheimani et al. 2017), the link between vaping and conversion to smoking has driven interest in how the traits of adolescents and the contexts they engage with most are associated with vaping.

Schools have long been recognized as critical contexts that shape youth health-risk behaviors (Alexander et al. 2001; Bonell et al. 2013a; Bonell et al. 2013b; Ellickson et al. 2003), including vaping (Corsi and Lippert 2016; Lippert 2018). However, recent population-wide trends call to question whether adolescent e-cigarette use has remained concentrated in particular schools or if the likelihood of vaping has become more uniform among students across schools. Responding to this question, the current study draws on Ecological Systems Theory (Bronfenbrenner 1977; Bronfenbrenner and Morris 1998) to investigate two research questions: (1) How has the school clustering of vaping—a measure of the extent to which schools differ based on their prevalence of student e-cigarette use— changed between 2011 and 2017 as the population-wide prevalence of youth vaping increased? (2) How is the pre- valence of school-level e-cigarette use associated with a student’s odds of vaping net of key influences from one’s family and peer group? To answer these questions, pooled cross-sectional data from the 2011 to 2017 National Youth Tobacco Surveys (NYTS) are used to decompose variance partitioning coefficients from multiyear multilevel regres- sion models of students nested within schools across the US. Variance partitioning components (VPC) provide an empirical estimate of school influence and a measure of the similarity of e-cigarette use or abstinence between students within schools. Thus, year-specific VPCs allow an assess- ment of change in the school clustering of teen vaping over a period when adolescent e-cigarette use emerged as a population health concern.

The Ecology of Youth Vaping

According to Ecological Systems Theory, youth develop- ment is guided by influences across multiply-embedded systems. These include features within the exosystem and macrosystem – broad societal influences such as public policy, culture, and population-wide trends—and the microsystem, or those institutions and relationships nearest to youth such as schools, families, and peers. Connections between and within the two systems are made via the mesosystem, a “system of systems” comprised of the reci- procal relationships among ecological features, such as

federal policies in the exosystem that alter family- or school-based processes in the microsystem. A fifth dimen- sion of the ecological model is the chronosystem, which reflects the developmental implications of both an indivi- dual’s stage in the life course and the sociohistorical context within which ecological systems are embedded and experienced. This aspect of ecological theory is critical for the purposes of this study, though the chronosystem has received less attention than features within the microsystem (Tudge et al. 2016).

Schools and Youth Health Behaviors

The association between school contexts and youth vaping is made clearer by the conceptual model developed by Frohlich and colleagues (Abel and Frohlich 2012; Frohlich et al. 2002; Frohlich, Corin and Potvin 2001), a complement to ecological theory. The model describes the interplay between contextual and agentic factors that produce—and are produced by—community features. According to this model, health behaviors are viewed as “generated prac- tices,” or products of both structural conditions and self- determination. This perspective is rooted in Weber’s dis- tinction between “life chances”—features of social structure that free or constrain human agency, and “life choices”— the health behaviors enacted by individuals responding to structural demands and opportunities (Hays 1994). Abel and Frohlich (2012) elaborate on self-determination by describing it in terms of its contextual embeddedness, or the “socially-structured development, acquisition, and applica- tion of structural and personal resources by individuals in a given context,” (p. 239). Under this view, the consequences of human agency can be understood as both the healthy and unhealthy practices among individuals responding to con- textual demands and opportunities, which reinforces the “collective health lifestyles” practiced among community members (Frohlich, Corin and Potvin 2001).

Like ecological theory, this model acknowledges the developmental importance of school environments and the reciprocal nature of relationships among features within and between ecological systems. Evidence from countless stu- dies lends empirical support to these claims. Given the recency of vaping as a population health concern much of this research is focused on other related outcomes, such as conventional tobacco use. These studies show that students —even low-risk students—attending schools with high rates of smoking are themselves more likely to smoke (Alexander et al. 2001; Leatherdale and Manske 2005). The association between school-level factors and adolescent smoking has proven robust to a range of confounders at the individual, family, and neighborhood levels (Dunn et al. 2015). Evidence from a limited number of studies on schools and e-cigarette use also show that youth attending

1900 Journal of Youth and Adolescence (2019) 48:1899–1911

schools where vaping is common are more likely to vape (Corsi and Lippert 2016).

The mechanisms linking schools to youth health-risk behaviors are not entirely clear, but may include norms and the provision of both social and material resources. Norms include the health beliefs, attitudes, and policies that inhere within school communities. While school-based policies have shown inconsistent and often null associations with adolescent health behaviors, especially smoking (Coppo et al. 2014; Galanti et al. 2014), the beliefs and attitudes held among one’s schoolmates have been shown to corre- late with one’s own beliefs, attitudes, and behaviors. For instance, there is a persistent and inverse association between average school-level student attainment and attendance and student-level substance use (Bonell et al. 2013b), suggesting that school environments supporting norms that emphasize student achievement discourage behaviors incompatible with academic success. Conversely, youth attending schools with lenient norms risk developing beliefs and attitudes compatible with substance use. For example, youth attending schools with high versus low rates of vaping are more likely to believe that e-cigarettes are harm-free and less addictive than combustible cigarettes, irrespective of their actual e-cigarette use (Lippert 2018).

In schools with lenient norms that fail to correct students’ misconceptions of the risks associated with e-cigarettes, pupils are more likely to have access to social and material resources needed to vape. This includes peer modeling of vaping behaviors and use of peers’ vaping devices. As a learned behavior, initiating e-cigarette use is facilitated by peer modeling and demonstrations. Indeed, a recent mixed- methods study of adolescents and young adults found that one-third of the sample first began experimenting with e- cigarettes because an e-cigarette using friend introduced them to the practice (Kong et al. 2015). Analysis of quali- tative data from the same study suggests that when one member of a peer group acquires an e-cigarette, offers to use the device are soon made to other group members. Other learned behaviors associated with e-cigarettes include performative aspects of use, such as “vaping tricks” demonstrated by friends or a desire to “look cool,” espe- cially for younger adolescents (Roditis et al. 2016). Given the social functions of e-cigarette use it is not surprising that a recent study showed leisure time activities among e- cigarette using teens is often centered on a mutual interest in vaping activities (Evans-Polce et al. 2018).

A Model for Changing School-Level Influence

The temporal stability of the link between school features and youth health behaviors has not been widely investi- gated. Ecological Systems Theory acknowledges that the influence of such systems on youth development is

conditioned by the sociohistorical context in which they are experienced (Bronfenbrenner and Morris 2006). In the case of adolescent e-cigarette use, recent changes in the exo and macrosystems warrant closer inspection of how the role of schools has changed for teen vaping. These cross-system relationships are described in Fig. 1.

Following the ecological model, adolescent e-cigarette use is considered a product of factors found across multiply- embedded systems, namely the microsystem (school, family, peer, and individual factors) and exo or macro- systems (policy, media, culture, and population-wide trends in e-cigarette use). The conceptual model in Fig. 1 under- scores the influence that features within the exo and mac- rosystems have on both schools as well as individual behaviors. At the level of the exosystem, state and federal policies on e-cigarettes have a bearing on both individual consumption (e.g., new minimum age restrictions for pur- chase) and school functionings (e.g., regulations on e- cigarette sales near schools). At the level of the macro- system, cultural shifts—the population-wide rise in e- cigarette use; youth-targeted marketing campaigns—not only have direct influences on youth behaviors, but they also modify the association between school features and youth behaviors. For instance, the role that peers play in disseminating beliefs and attitudes about the safety and benefits of vaping (e.g., looking cool or demonstrating adult-like behavior) could be augmented, or even sup- planted by, messages circulating throughout social media within the macrosystem. Indeed, recent evidence shows that advertising of e-cigarettes to adolescents has climbed dra- matically over recent years (Duke et al. 2014; McCarthy 2016), and exposure to such advertising, including celebrity endorsements via social media, is linked to higher odds of

Fig. 1 Conceptual model of the interdependent association between schools and e-cigarette use

Journal of Youth and Adolescence (2019) 48:1899–1911 1901

teen use (Camenga et al. 2018; Hammig, Daniel-Dobbs and Blunt-Vinti 2017; Pasch et al. 2018; Phua, Jin and Hahm 2018; Singh et al. 2016b). Under this scenario, the impor- tance of schools to teen vaping could have diminished over time as analogous influences from other ecological systems became more pronounced—a phenomenon we refer to as system drift.

Conversely, other changes in the exo and macrosystems could have enhanced the relevance of schools to teen vaping via the mechanisms described earlier. Recently expanded regulations on e-cigarettes by the FDA (U.S. Food and Drug Administration 2016) imposed mandatory minimum age requirements on e-cigarette sales, even though research shows that sales of e-cigarettes to minors is still common (Levinson 2018). Should e-cigarettes become more difficult for youth to purchase, adolescents will rely more on peer- mediated access organized within schools, raising the importance of schools to adolescent vaping. Additionally, important shifts in how teens vape could have strengthened the role that schools played in facilitating the initial emer- gence of adolescent e-cigarette use. Manufacturers of technologically-novel “next-gen” e-cigarette devices have targeted teenagers in their marketing (Chu et al. 2018) and to great effect, as devices like Juul have become popular among youth (Krishnan-Sarin et al. 2019). The evolution of vaping devices over time could have required localized peer-to-peer demonstrations of use, peer-mediated access to vaping materials, and flexible school-based cultural norms that permit vaping, bringing schools back into focus as the broader prevalence of teen vaping climbed.

Current Study

While either scenario—diminishing or increasing school- level influences—is possible, no empirical attention has been given to how the clustering of adolescent e-cigarette use within schools has shifted over recent years. Respond- ing to the lack of attention to shifting school influences on teen health, Ecological Systems Theory and conceptual models of school environments and adolescent health (Abel and Frohlich 2012; Frohlich et al. 2002; Frohlich, Corin and Potvin 2001) are used to address the following research questions: (1) How has the school clustering of vaping changed between 2011 and 2017 as the population-wide prevalence of youth vaping increased? (2) How is the pre- valence of school-level e-cigarette use associated with a student’s odds of vaping net of key influences from one’s family and peer group? The analyses presented here emphasize the importance of schools over time in distin- guishing risk for teen vaping while simultaneously adjust- ing for other features of the microsystem known to correlate with adolescent e-cigarette use including one’s own

conventional smoking status and that of their friends or family members. Poor academic achievement is also taken into account given the association this shares with nicotine use in prior studies. Additionally, as extant research has shown higher rates of vaping among older versus younger adolescents, males versus females, and among non-Hispanic Whites versus other racial/ethnic groups (Lippert 2018), students’ demographic characteristics are assessed. With adjustments made for student demographics, low scholastic achievement, personal smoking status and vicarious expo- sure to tobacco use among friends or family, the multilevel models presented here provide an estimate of the impor- tance of school context to teen vaping over time.

In addressing these questions, three key hypotheses are tested. The first is that the extent of school clustering of adolescent e-cigarette use declined from 2011 to 2015 as vaping became more normative across society and alter- native sources of social and material resources needed to vape became available to teens across a variety of schools (Hypothesis 1). This hypothesis is tested by comparing the year-specific variance components from multilevel models where the null hypothesis is that there were not significant differences in the variance components across years, and the alternative is that school-level variance in student vaping declined between 2011 and 2015. The second is that between 2015 and 2017 this pattern reversed as novel methods of vaping again necessitated context-mediated access to vaping resources, peers, and permissive school norms (Hypothesis 2). Again, this hypothesis is evaluated by comparing year-specific variance components and via Wald tests evaluating whether the variance components from 2015 and 2017 are statistically identical (Ho) or if the variance components from 2017 were larger than those from 2015 (Ha). The third hypothesis is that net of the changing school clustering of e-cigarette use between 2011 and 2017, students attending a school with a higher versus lower prevalence of e-cigarette use will present a higher risk of vaping (Hypothesis 3). We test these hypotheses by using repeated cross-sectional data from the CDC-sponsored 2011–2017 National Youth Tobacco Surveys (NYTS) and by decomposing variance partitioning coefficients (VPC) from multiyear multilevel regression models of students nested within schools across the US.

Methods

Sample

Data are from four repeated cross-sections of the NYTS conducted in 2011, 2013, 2015, and 2017. The NYTS is a self-administered survey based on a national sample of US middle- and high-school students in public and private

1902 Journal of Youth and Adolescence (2019) 48:1899–1911

schools across all 50 states and DC. The 2011 survey—the first to include e-cigarette measures—covered 178 schools and 18,866 students (overall response rate, defined as the product of school- and student-level participation, was 73.2%); the 2013 sample covered 187 schools and 18,406 students (68.4% overall response rate); the 2015 sam- ple covered 183 schools and 17,711 students (63.4% overall response rate); the 2017 sample covered 185 schools and 17,872 students (68.1% overall response rate).

NYTS used a 3-stage cluster-based sampling design where primary sampling units (counties, several small counties, or portions of larger counties) were selected without replacement followed by schools within counties and students within schools. Purposeful oversampling was done for specific groups including Blacks and Hispanic students. The pooled sample size across all three rounds of NYTS was 72,855. Item missingness was rare and no single item exceeded 5% missing, a common threshold for judging the necessity of multiple imputation (Jakobsen et al. 2017). Thus, complete-case data were used by excluding those with any missing data across all items (n = 7788, 10.7%), arriving at a final analytic sample of 65,067 students.

Measures

E-cigarette use

Both lifetime and current (past-month) e-cigarette use were examined. In the 2011 and 2013 survey years, this was assessed dichotomously across two questions with those students who indicated use of “electronic cigarettes or e- cigarettes, such as Ruyan or NJOY” to the question: “Which of the following tobacco products have you ever tried?” Students reporting any lifetime use of e-cigarettes were coded “1” for the lifetime use measure (“0” other- wise), while students reporting any use of e-cigarettes in the month prior to the interview were coded “1” for the current use measure (“0” otherwise). A minor revision to these survey items was instituted in the 2015 and 2017 ques- tionnaires, where example brands were not provided and youth were simply asked if they had used e-cigarettes.

School prevalence of e-cigarette use

School-level e-cigarette usage was captured by aggregating student-level reports of lifetime e-cigarette use to the school level and was categorized in three groups—low, moderate, high—based on tertiles in regression models. This approach aligns with that of the CDC, which in 2017 began providing a similar aggregate measure of school-level lifetime e- cigarette use in data files released to the public.

Student, family, and peer smoking

Student-level conventional cigarette smoking was cap- tured in a three-category variable with categories for ‘never’, ‘experimenter’, and ‘current’. Never smokers had never smoked or tried cigarettes; experimenters had tried smoking at least 1-2 puffs but had lifetime usage of less than 100 cigarettes (5 packs) or no past-month smoking occurrences; current smokers had smoked at least 100 cigarettes in their lifetime and reported smoking at least once in the past 30 days. Vicarious exposure to smoking of family members was assessed via student reports of any household members who had smoked tobacco products in the previous week while the respondent was home (coded “1” for affirmative responses, “0” otherwise). Second- hand smoking exposure from peers was based on a question asking respondents whether in the past month they had breathed tobacco smoke from a smoker in places such as school buildings or grounds, indoors or outside, or other public places (coded “1” for affirmative responses, “0” otherwise).

Student demographics

Respondent age is a continuous variable and centred around its mean (14.6 years). Age appropriate-for-grade was calculated as a proxy for poor academic performance which may have resulted in students repeating a grade. A dichotomous indicator was created to indicate high age for grade (“1”) or appropriate age for grade (“0”) according to the following criteria: age >12 y in grade 6; >13 y in grade 7; >14 y in grade 8; >15 y in grade 9; >16 y in grade 10; >17 y in grade 11; and >18 y in grade 12. Race is self- reported and grouped into the following: non-Hispanic White (reference), non-Hispanic Black, Hispanic, or other race/ethnicity. Student gender is based on self-reports and dichotomized with “female” serving as the reference group. Finally, survey year is used to estimate year- specific variance components and VPCs from multilevel models and as a covariate in bivariate as well as multi- variate multilevel models, where 2011 is treated as the reference survey year.

Analysis

To assess the relationship between lifetime and past- month e-cigarette usage and survey year, a series of 2- level logistic regression models were fit to the response of lifetime and past-month e-cigarette usage (y, used vs had not used) for individual i in school j. The general mod- elling approach was as follows: e-cigarette usage, Pr(yij = 1), assumed to be binomially distributed yij ~ Binomial (1, πij) with probability πij, was related to indicators for all

Journal of Youth and Adolescence (2019) 48:1899–1911 1903

survey years (β0j, β1j, β2j) and other covariates X and a random effect for each survey year by a logit link function using the following specification:

Logit πij � �

¼ β0j2011 þ β1j2013 þ β2j2015 þ β3j2017

þβXij þ u0j þ u1j þ u2j þ u3j � �

The terms u0j, u1j, u2j, and u3j, in brackets represent school-level random effects which are allowed to be esti- mated separately for each survey year using a diagonal matrix for the variance-covariance matrix. These random effects are assumed to be independently and identically distributed with variances σ2u0, σ

2 u1, σ

2 u2, and σ

2 u3. The var-

iance parameters quantify heterogeneity in the log odds of e-cigarette usage between schools in 2011, 2013, 2015, and 2017, respectively. We expressed the year-specific school- level variance as a percentage of the total variance for that period using the variance partitioning coefficient (VPC). The VPC (for 2011) is calculated as,

VPC ¼ σ2u0= σ2u0 þ π 2

3

� � .

The first series of models were specified with indica- tors for survey year only. The second set included student-level characteristics (age, sex, age-for-grade, race) and school-level e-cigarette usage. A final set of models included individual smoking behaviour along with all of the covariates from the previous model. All models used Bayesian estimation procedures imple- mented via Monte Carlo Markov Chain (MCMC) meth- ods and Metropolis Hastings algorithm available in MLwiN 2.3. MCMC simulations were run for 50,000 to 75,000 iterations following a burn-in of 1000 iterations. MCMC diagnostics were examined for all parameters to ensure model convergence. The random variances and their associated credible intervals were obtained directly from summarizing the simulated posterior distributions of model parameters. This methodology has been demon- strated to reduce bias in estimated random effects in binomial multilevel models which can occur when using maximum-likelihood procedures (Browne et al. 2005). Further, the MCMC procedure provides the Deviance Information Criterion (DIC) which is an overall measure of model goodness-of-fit. A small difference in DIC between models indicates that they are empirically equivalent, whereas differences larger than 10 are taken to suggest support for the model with lower DIC value.

Multivariate results were similar for both lifetime and past-month e-cigarette use. Thus, results only from mod- els predicting current e-cigarette use are reported, though in Table 3 year-specific variance components from all models for lifetime and current use are provided.

Results

Table 1 indicates that the unweighted prevalence of lifetime and current e-cigarette usage increased in the NYTS from 3.0% and 1.0% in 2011, respectively, to 20.8% and 7.8% by 2017. The sample descriptives also reveal an unstandar- dized average age of 14.6 years across all years, a sample well-balanced by gender (50.3% male), and one that gen- erally follows population-wide racial and ethnic composi- tional attributes for younger cohorts, though deliberate oversampling led to a higher share of Hispanic students in the NYTS samples than is observed in the general popula- tion. A small minority of students (5.6% across all years) met criteria for a high age-for-grade. Similarly, a small and diminishing share of survey participants reported experi- menting with cigarettes or actively using them at the time of the survey. Following this pattern, fewer students reported co-residing with a smoker across each survey year, though the fraction of students reporting second-hand smoke exposure from peers (e.g., while socializing with friends) increased between 2013 and 2015.

Bivariate analyses are presented in Table 2. These are based on multilevel models of lifetime and current e- cigarette use and each covariate with no other model adjustments. Results indicate higher odds of e-cigarette use among students attending moderate or high-use schools compared to low-use schools, older versus younger ado- lescents, males versus females, non-Hispanic Whites com- pared to non-Whites (with the exception of current e- cigarette use among Hispanic youth), those whose age was high for their grade, youth exposed to second-hand smoking from family members or friends, and those who experi- mented with conventional cigarettes or smoked them reg- ularly versus abstainers. Bivariate results also indicate higher odds of e-cigarette use among youth interviewed in 2013 and beyond versus those surveyed in the baseline year of 2011.

From a variance components analysis, school-level residuals and corresponding standard errors for lifetime and current e-cigarette usage by survey year were estimated. Plots of the residuals ordered from smallest to largest (Supplementry Fig. 1) indicate an increase in current usage at the school level over time. Across all years, a quarter of schools (105/733) demonstrated significantly higher than average levels of current e-cigarette use, increasing from 5/ 178 (2.8%) of schools in 2011 to 32/183 (17.4%) in 2017 although different samples of schools were drawn at each survey wave. Results for lifetime use followed a similar pattern.

Multilevel ‘null’ models fit separately to the responses for lifetime and current e-cigarette usage and with random intercepts for schools by each survey year indicated that the between-school variance increased between 2011 and 2013

1904 Journal of Youth and Adolescence (2019) 48:1899–1911

before declining in 2015 and rising again by 2017 in the case of current usage. The between-school variance in lifetime e-cigarette usage was 0.99 in 2011, equivalent to 23.1% of the variability in the outcome attributed to the school level. This increased to 1.04 (23.9%) in 2013 before falling to 0.94 (22.1%) by 2015 and 0.73 (18.2%) by 2017. A distinctive pattern was found for current usage, with variance from the null model increasing from 0.64 (16.2%) in 2011 to 0.91 in 2013 (21.7%) before returning to 0.7 in 2015 (17.6%) followed by a second increase to 0.77 (19%) in 2017 (Table 3).

Additional model specifications shown in Table 3 intro- duced school- and student-level covariates—including high age-for-grade and student demographics—to the null models for current and lifetime usage. The introduction of covariates reduced between-school variance indicating that some of the school-to-school variability apparent in the null models was due to between-school differences in student composition and school-level e-cigarette usage. Across all models, the VPC estimate was greatest for the 2011 survey indicating a higher level of correlation between two students within the same school in terms of their lifetime and current e-cigarette usage.

Model 4, which adds adjustments for student-level con- ventional smoking status and second-hand smoking expo- sure from family or peers, paints a clearer picture of how the school-level clustering of adolescent e-cigarette use shifted over time net of these controls. While models of lifetime use revealed lower VPC estimates across all years, fully- adjusted models for current use showed larger but steadily declining VPCs from 2011 to 2015, suggesting that the prevalence of current e-cigarette use was becoming more uniform across schools over time. However, this pattern was disrupted between 2015 and 2017 when the VPC estimates increased from 5.02 to 12.1%. This suggests that schools had become more differentiated by the prevalence of stu- dents’ current e-cigarette use. This pattern is shown more clearly in Fig. 2.

Significant year-to-year changes in the variance compo- nents derived from Model 4 for current e-cigarette use were investigated more closely via Wald tests. Year-specific school-level variance estimates, τ, were compared for each pairing of subsequent years. These results indicate that, while the VPCs for models examining current use generally declined between 2011 and 2015, the year-to-year changes were not significantly different. However, between 2015

Table 1 Sample description (N = 65,067)

Unweighted sample percentages

Student-level variables 2011 2013 2015 2017 All years

Lifetime e-cigarette use 3.2 7.9 26.8 21.3 14.7

Current e-cigarette use 1.0 2.9 11.1 8.1 5.8

Age in years (unstandardized)a 14.6 14.7 14.6 14.7 14.6

Gender

Female 49.8 49.8 49.9 49.5 49.7

Male 50.2 50.2 50.1 50.5 50.3

Race/ethnicity

White 61.1 58.2 58.3 57.8 58.9

Black 14.2 15.5 13.7 12.5 14.0

Hispanic 20.0 21.0 22.0 23.9 21.9

Other race/ethnicit 4.6 5.2 5.1 5.8 5.2

High age-for-grade 6.9 5.9 5.3 4.2 5.6

Smoking status

Never 70.6 74.8 78.8 83.7 76.9

Experimenter 24.4 21.5 18.8 14.5 19.8

Current smoker 5.0. 3.8 2.4 1.8 3.3

Public secondhand smoke exposure

Yes 41.3 40.3 53.1 51.1 46.4

No 58.7 59.7 46.9 48.9 53.6

Coresides with smoker

Yes 26.5 23.8 22.7 19.9 23.2

No 73.5 76.2 77.3 80.1 76.8

aShown as the arithmetic mean

Journal of Youth and Adolescence (2019) 48:1899–1911 1905

and 2017 the change was significant for current use (χ = 10.84, p < 0.01). Further, the variance components between 2011 and 2017 showed no significant difference, indicating a rebound in the extent of school-clustering of current e- cigarette use to levels comparable with 2011.

Table 4 summarizes parameters from multilevel logistic models assessing associations between school-level vap- ing prevalence and student-level traits with current e- cigarette use (results for lifetime use are available upon request). Results indicate that school-level prevalence of lifetime e-cigarette use was robustly associated with cur- rent e-cigarette usage at the individual level conditional on age, sex, race, age-for-grade, and second-hand smoke exposure (Table 4, Model 1). The odds ratio for current e- cigarette use was 2.37 (95% credible interval [CI]: 1.97, 2.85) for students in schools with moderate e-cigarette usage, and 4.47 (95% CI: 3.69, 5.43) among students in high usage schools. These effect sizes were attenuated but robust in a second model which included student-level conventional smoking status.

Beyond school factors, multiple variables capturing family, peer, and student-level risk indicators were also

significantly related to one’s odds of e-cigarette use. Results from Model 2 shown that youth living with a smoker or those who were friends with one had 1.76 (95% CI: 1.62, 1.93) and 1.54 (95% CI: 1.41, 1.68) the odds, respectively, of having used an e-cigarette in the past month. One’s own smoking status was also strongly associated with the odds of current e-cigarette use. Among sociodemographic mea- sures, results indicate higher odds of current e-cigarette use for males versus females and non-Hispanic Whites com- pared to non-Hispanic Blacks.

Discussion

The climbing prevalence of teen vaping is well-documented (Gentzke et al. 2019), though it is unclear how schools are implicated in this trend. Ecological Systems Theory (Bronfenbrenner 1977; Bronfenbrenner and Morris 1998) holds that youth development is influenced by factors very near to adolescents including family, friends, and personal characteristics, as well as more distal factors including school settings and the broader societal and sociohistorical

Table 2 Bivariate associations with e-cigarette use (N = 65,067)

Current e-cigarette use Lifetime e-cigarette use

OR 95% CI OR 95% CI

School e-cigarette use

Low 1.00 1.00

Moderate 2.61*** (2.27, 3.00) 2.75*** (2.52, 3.00)

High 6.57*** (5.77, 7.49) 7.18*** (6.62, 7.80)

Age (centered) 1.31*** (1.28, 1.33) 1.34*** (1.32, 1.35)

Male sex 1.52*** (1.42, 1.63) 1.26*** (1.21, 1.32)

Race

White 1.00 1.00

Black 0.41*** (0.37, 0.47) 0.49*** (0.46, .53)

Hispanic 0.91* (0.84, 0.98) 0.96 (0.92, 1.01)

Other 0.70*** (0.60, 0.82) 0.64*** (0.57, 0.71)

High age for grade 1.47*** (1.30, 1.67) 1.18*** (1.08, 1.29)

Peer secondhand smoke exposure 2.95*** (2.74, 3.17) 2.45*** (2.34, 2.57)

Family secondhand smoke exposure 2.95*** (2.75, 3.16) 2.39*** (2.28, 2.50)

Own smoking

Never smoker 1.00 1.00

Experimenter 6.65*** (6.15, 7.18) 6.62*** (6.30, 6.96)

Current smoker 21.47*** (19.24, 23.97) 18.28*** (16.62, 20.09)

Year

2011 1.00 1.00

2013 2.89*** (2.42, 3.45) 2.66*** (2.39, 2.95)

2015 12.29*** (10.48, 14.41) 11.76*** (10.69, 12.93)

2017 8.47*** (7.20, 9.95) 8.58*** (7.79, 9.44)

CI credible interval

*p < 0.05; **p < 0.01; ***p < 0.001

1906 Journal of Youth and Adolescence (2019) 48:1899–1911

contexts in which they are embedded. Scholars have extended this view to adolescent health outcomes by developing conceptual models describing the interplay between school environments and teen health behaviors (Frohlich et al. 2002; Frohlich, Corin and Potvin 2001), though insufficient attention has been paid to how school- level influences on adolescent health behaviors change over

time as broader societal factors—policy, culture, media, and population-wide trends—also change. This omission leaves untested a key aspect of ecological theory—that the influ- ence of ecological systems on adolescent health and development is conditional on the sociohistorical context in which they are embedded and experienced (Tudge et al. 2016). Thus, the current study is focused on understanding how the association between school contexts and adolescent e-cigarette use changed over the same period within which the broader prevalence of teen vaping skyrocketed.

To address this, multiyear nationally-representative data from the National Youth Tobacco Survey (NYTS) were utilized with a conceptual model describing school influ- ences on adolescent health-risk behaviors as conditional on the sociohistorical context within which they are experi- enced. Variance components including variance partitioning coefficients (VPCs) from the multilevel models were decomposed into year-specific estimates in order to identify whether teen vaping remained concentrated in certain schools over time, or if the school-level clustering of e- cigarette use diminished as the practice generally grew more popular among youth. We hypothesized that (1) the clus- tering of teen e-cigarette use diminished between 2011 and 2015 as the broader popularity of vaping increased; (2) that this trend would have reversed between 2015 and 2017 as

Table 3 Year-specific variance estimates from multilevel models of lifetime and current e-cigarette use (N = 65,067)

Outcome Model specification

Model 1: null Model 2: school prevalence

Model 3: M2 + controls Model 4: M3 + smoking

Var (τ)a VPC (%) Var (τ)a VPC (%) Var (τ)a VPC (%) Var (τ)a VPC (%)

Lifetime e-cigarette use

2011 0.99 23.08 0.18 5.29 0.18 5.06 0.19 5.45

(0.66, 1.43) (0.10, 0.30) (0.10, 0.28) (0.09, 0.31)

2013 1.03 23.93 0.04 1.33 0.03 0.76 0.10 3.03

(0.74, 1.41) (0.01, 0.09) (0.00, 0.07) (0.04, 0.18)

2015 0.94 22.13 0.07 2.18 0.05 1.57 0.07 2.11

(0.72, 1.21) (0.05, 0.11) (0.03, 0.09) (0.04, 0.11)

2017 0.73 18.25 0.10 2.89 0.10 2.90 0.12 3.48

(0.56, 0.95) (0.06, 0.14) (0.06, 0.14) (0.08, 0.17)

Current e-cigarette use

2011 0.64 16.20 0.23 6.41 0.30 8.43 0.36 9.91

(0.32, 1.11) (0.03, 0.51) (0.04, 0.60) (0.09, 0.68)

2013 0.91 21.70 0.17 4.84 0.17 4.86 0.29 8.06

(0.61, 1.31) (0.07, 0.30) (0.06, 0.31) (0.15, 0.48)

2015 0.70 17.56 0.15 4.32 0.13 3.71 0.17 5.02

(0.51, 0.94) (0.10, 0.22) (0.07, 0.20) (0.11, 0.26)

2017 0.77 18.98 0.32 8.81 0.39 10.70 0.45 12.11

(0.57, 1.03) (0.22, 0.44) (0.28, 0.54) (0.32, 0.62)

a95% credible intervals for year-specific variance estimate τ shown in parentheses

Fig. 2 Year-specific variance partitioning coefficients (VPC) from fully-adjusted multilevel models (N = 65,067) Estimates adjusted for student characteristics (age, gender, race/ethnicity, high age-for-grade), secondhand exposure to family or peer smoking, own smoking status, and school prevalence of e-cigarette use

Journal of Youth and Adolescence (2019) 48:1899–1911 1907

new vaping modalities were introduced and required peer- to-peer demonstrations of use, access to materials, and flexible school norms; and (3) that even as schools became more, and eventually less, uniform with respect to the prevalence of student vaping, students attending schools with high rates of e-cigarette use would themselves be more likely to vape than their peers at schools where vaping was rare.

To the first hypothesis, study findings yielded mixed support. On the one hand, evidence from fully-adjusted models revealed that VPCs between 2011 and 2015 steadily declined, indicating greater uniformity between schools in student-level e-cigarette use. However, tests of these dif- ferences were null. This was not the case for the second hypothesis, which held that the school-level variance in student e-cigarette use should have increased between 2015 and 2017 as the initiation of new vaping practices would have required school-mediated access to peers, resources, and specific norms. The empirical results bear this out, with Wald tests indicating a significant difference in the VPCs

from 2015 and 2017 net of school- and student-level risk indicators. Finally, the third hypothesis was that students attending school with high rates of e-cigarette use would be more likely to use themselves, a supposition that the results supported.

One interpretation of the first finding is that school-to- school rates of student e-cigarette use increasingly converged over the period between 2011 and 2015 as the popularity of vaping diffused among students across schools. Shortly after e-cigarettes were introduced to US consumers, adolescents may have depended on context-mediated access to vaping materials (e.g., “cig-a-like” devices and customized e-pens), peer models to demonstrate use techniques, and permissive school norms. Over time, as teen-directed advertising pro- liferated (Duke et al. 2014; El-Toukhy and Choi 2016; Kim, Arnold and Makarenko 2014; McCarthy 2016), access to peer and celebrity demonstrators grew via social media (Merianos, Gittens and Mahabee-Gittens 2016; Phua, Jin and Hahm 2018), and purchasing opportunities expanded (Levinson 2018; Williams, Derrick and Ribisl 2015), the relevance of

Table 4 Multilevel logistic regression models results for current e-cigarette use (N = 65,067)

Model 1 Model 2

OR 95% CI OR 95% CI

School e-cigarette use

Moderate 2.37*** (1.97, 2.85) 2.13*** (1.75, 2.60)

High 4.47*** (3.69, 5.43) 3.34*** (2.73, 4.09)

Age (centered) 1.15*** (1.12, 1.18) 1.02 (0.99, 1.05)

Sex

Male 1.70*** (1.58, 1.83) 1.55*** (1.43, 1.68)

Race

Black 0.61*** (0.53, 0.69) 0.68*** (0.59, 0.79)

Hispanic 1.17** (1.06, 1.28) 1.08 (0.98, 1.19)

Other 0.84* (0.70, 0.99) 0.87 (0.73, 1.05)

High age for grade 1.26** (1.09, 1.45) 1.12 (0.96, 1.30)

Peer secondhand smoke exposure 1.92*** (1.77, 2.09) 1.54*** (1.41, 1.68)

Family secondhand smoke exposure 2.81*** (2.60, 3.04) 1.76*** (1.62, 1.93

Own smoking

Experimenter 7.04*** (6.43, 7.72)

Current smoker 29.03*** (24.93, 33.80)

Random effects variances

2011 0.30* (0.04, 0.60) 0.36* (0.09, 0.68)

2013 0.17*** (0.06, 0.31) 0.29** (0.15, 0.48)

2015 0.13*** (0.07, 0.20) 0.17*** (0.11, 0.26)

2017 0.39*** (0.28, 0.54) 0.45*** (0.32, 0.62)

Model fit

DIC 22133.43 19202.72

Difference in DIC 2930.71

CI credible interval, DIC differing information criterion

*p < 0.05; **p < 0.01; ***p < 0.001

1908 Journal of Youth and Adolescence (2019) 48:1899–1911

local school contexts to teen vaping may have diminished. Such an explanation is consistent with the falling VPC esti- mates identified between 2011 and 2015, though the year-to- year differences in school VPCs did not reach statistical significance.

With respect to the second finding, the introduction of new vaping methods, devices, and flavors required a reas- sessment of adolescent e-cigarette use and its correlates (Jenssen and Wilson 2019). A central focus of this study was to reassess how the relevance of schools changed as the practice of teen vaping also changed (Krishnan-Sarin et al. 2019). The recent increase in the school clustering of e- cigarette use suggests a renewed role of schools for both the broader uptick in teen e-cigarette use (Gentzke et al. 2019) and the emerging popularity of novel vaping methods. Future assessments may show again that school clustering in relation to what are now considered novel vaping methods will diminish, but the findings hint that schools played a pronounced part in recent trends. Importantly, the magnitude of clustering was higher for current versus life- time use. This stands to reason, as experimental use is likely to be perceived as less risky and requires less steady access to material resources and permissive norms than current use. Indeed, that there was a greater degree of school clustering for current versus lifetime use lends support to the con- clusion that schools are important contexts influencing adolescent e-cigarette use.

The third finding is consistent with the second. Net of powerful risk factors for vaping including one’s own tobacco use and second-hand exposure to use among friends or family, the school-level prevalence of vaping was sig- nificantly associated with student-level use. Such findings suggest that initiatives devised to discourage adolescent e- cigarette use must not only consider personal risk indicators such as conventional tobacco use, but also those that inhere at the school community level such as restrictions on e-cigarette sales and marketing near schools (Giovenco et al. 2016).

Several limitations to the study are noteworthy. First, use of repeated cross-sectional data obscures the sequencing of conventional and electronic cigarette use. Second, although important covariates were included in the multilevel models, the possibility of residual confounding by unmeasured vari- ables such as family socioeconomic status (SES) or broader community characteristics such as neighborhood-level nico- tine availability and advertising environments cannot be dis- counted. Third, given the NYTS design, it is likely that the same schools are not represented in each of the annual survey waves. The conclusion that the school-level clustering of adolescent e-cigarette use fell and rose again between 2015- 2017 would be strengthened with panel data that assessed both students and school environments repeatedly over time.

Notwithstanding these limitations, the current study adds to what is known about the shifting links between schools and adolescent e-cigarette use. While the overall variance in the outcomes attributed to the school level was small than that attributed to the student level, it is worth noting that this is common to most studies of context and health, and the par- titioning of the variance between schools and students yields greater VPCs than are generally observed for other common health outcomes (e.g., obesity). Thus, school factors may be more strongly associated with teen vaping than is the case for other commonly-studied health-risk behaviors.

Conclusion

Though the relationship between school contexts and ado- lescent e-cigarette use has shifted over time, schools are still consequential for teen vaping – a finding that is consistent with understudied tenets of Ecological Systems Theory. In light of these findings, prevention measures could incor- porate individual-level aims such as educating teens on the potential for nicotine addictions and health risks caused by e-cigarette use. Indeed, given recent results showing increasing rates of e-cigarette use among otherwise tobacco- abstinent youth, the need seems clear for health education programs that correct teenage misunderstandings of the addictive potential of e-cigarette use (Hilton et al. 2016; Lippert 2018). Preventative measures should also extend to school-level factors, including greater restrictions on retail sales and consumption of e-cigarettes near school property (Giovenco et al. 2016) and the realignment of school norms towards health-promoting behaviors and understandings of the risks associated with vaping. This may involve added restrictions on advertising, which has not only increasingly targeted youth (Duke et al. 2014; Kim, Arnold and Makarenko 2014)—especially teens in areas with low rates of conventional tobacco use among (Dai, Deem and Hao 2017)—but has largely remained untouched by recent expansions to federal regulatory oversight of e-cigarettes.

Authors’ Contributions AL planned the study, drafted the manuscript, and contributed to the analyses and results interpretation; DC con- tributed to the planning of the study, conducted descriptive and mul- tilevel analyses, and interpreted results; GV assisted with the drafting of the manuscript, designed the presentation of results, and conducted systematic literature reviews for the study. All authors approved of the study.

Funding This research received no specific grant from any funding agency in the public, commercial, or not-for-profit sectors.

Data Sharing and Declaration The datasets analyzed during the cur- rent study are available through the Centers for Disease Control, http://www.cdc.gov/tobacco/data_statistics/surveys/nyts

Journal of Youth and Adolescence (2019) 48:1899–1911 1909

Compliance with Ethical Standards

Conflict of Interest The authors declare that they have no conflict of interest.

Ethical Approval This study was deemed “exempt from review” by the Colorado Multiple Institution Review Board.

Informed Consent Centers for Disease Control personnel obtained written consent and permission by parents or legal guardians of selected youth prior to student participation in the National Youth Tobacco Survey.

Publisher’s note: Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.

References

Abel, T., & Frohlich, K. L. (2012). Capitals and capabilities: linking structure and agency to reduce health inequalities. Social Science & Medicine, 74(2), 236–244.

Alexander, C., Piazza, M., Mekos, D., & Valente, T. (2001). Peers, schools, and adolescent cigarette smoking. Journal of Adolescent Health, 29(1), 22–30.

Bonell, C. P., Fletcher, A., Jamal, F., Wells, H., Harden, A., Murphy, S., & Thomas, J. (2013a). Theories of how the school environment impacts on student health: systematic review and synthesis. Health & Place, 24, 242–249.

Bonell, C., Parry, W., Wells, H., Jamal, F., Fletcher, A., Harden, A., Thomas, J., Campbell, R., Petticrew, M., Murphy, S., Whitehead, M., & Moore, L. (2013b). The effects of the school environment on student health: a systematic review of multi-level studies. Health and Place, 21, 180–191.

Bronfenbrenner, U. (1977). Toward an experimental ecology of human-development. American Psychologist, 32(7), 513–531.

Bronfenbrenner, U., & Morris, P. (1998). The ecology of develop- mental processes. In Handbook of child psychology: theoretical models of human development (vol. 1, 5th ed. pp. 993–1028). Hoboken, NJ, US: John Wiley & Sons Inc.

Bronfenbrenner, U., & Morris, P. (2006). The Bioecological Model of Human Development. In Handbook of child psychology: theo- retical models of human development (vol. 1, 6th ed. pp. 793–828). Hoboken, NJ, US: John Wiley & Sons Inc.

Browne, W. J., Subramanian, S. V., Jones, K., & Goldstein, H. (2005). Variance partitioning in multilevel logistic models that exhibit over-dispersion. Journal of the Royal Statistical Society: Series A, 168(3), 599–613.

Bunnell, R. E., Agaku, I. T., Arrazola, R. A., Apelberg, B. J., Car- aballo, R. S., Corey, C. G., Coleman, B. N., Dube, S. R., & King, B. A. (2015). Intentions to smoke cigarettes among never- smoking US middle and high school electronic cigarette users: National Youth Tobacco Survey, 2011–2013. Nicotine and Tobacco Research, 17(2), 228–235.

Camenga, D., Gutierrez, K. M., Kong, G., Cavallo, D., Simon, P., & Krishnan-Sarin, S. (2018). E-cigarette advertising exposure in e- cigarette naive adolescents and subsequent e-cigarette use: a longitudinal cohort study. Addictive Behaviors, 81, 78–83.

Chu, K. H., Colditz, J. B., Primack, B. A., Shensa, A., Allem, J. P., Miller, E., Unger, J. B., & Cruz, T. B. (2018). JUUL: spreading online and offline. Journal of Adolescent Health, 63(5), 582–586.

Chun, L. F., Moazed, F., Calfee, C. S., Matthay, M. A., & Gotts, J. E. (2017). Pulmonary toxicity of e-cigarettes. American Journal of Physiology-Lung Cellular and Molecular Physiology, 313(2), L193–L206.

Coleman, B. N., Apelberg, B. J., Ambrose, B. K., Green, K. M., Choiniere, C. J., Bunnell, R., & King, B. A. (2014). Association between electronic cigarette use and openness to cigarette smoking among U.S. young adults. Nicotine & Tobacco Research, 17(2), 212–218.

Coppo, A., M. R. Galanti, L. Giordano, D. Buscemi, S. Bremberg, and F. Faggiano. (2014). School policies for preventing smoking among young people. Cochrane Database of Systematic Reviews (10):Cd009990.

Corsi, Daniel. J., & Lippert, Adam. M. (2016). An examination of the shift in school-level clustering of US adolescent electronic cigarette use and its multilevel correlates, 2011–2013. Health & Place, 38, 30–38.

Dai, H. Y., Deem, M. J., & Hao, J. Q. (2017). Geographic variations in electronic cigarette advertisements on Twitter in the United States. International Journal of Public Health, 62(4), 479–487.

Duke, J. C., Lee, Y. O., Kim, A. E., Watson, K. A., Arnold, K. Y., Nonnemaker, J. M., & Porter, L. (2014). Exposure to electronic cigarette television advertisements among youth and young adults. Pediatrics, 134(1), e29–e36.

Dunn, E. C., Richmond, T. K., Milliren, C. E., & Subramanian, S. V. (2015). Using cross-classified multilevel models to disentangle school and neighborhood effects: an example focusing on smoking behaviors among adolescents in the United States. Health & Place, 31, 224–232.

El-Hellani, A., Salman, R., El-Hage, R., Talih, S., Malek, N., Baal- baki, R., Karaoghlanian, N., Nakkash, R., Shihadeh, A., & Saliba, N. A. (2018). Nicotine and carbonyl emissions from popular electronic cigarette products: correlation to liquid composition and design characteristics. Nicotine & Tobacco Research, 20(2), 215–223.

El-Toukhy, S. M., & Choi, K. (2016). Magazine hyped: trends in tobacco advertising and readership characteristics, 2010–2014. Preventive Medicine, 91, 132–137.

Ellickson, P. L., Bird, C. E., Orlando, M., Klein, D. J., & McCaffrey, D. F. (2003). Social context and adolescent health behavior: does school-level smoking prevalence affect students' subsequent smoking behavior? Journal of Health and Social Behavior, 44(4), 525–535.

Evans-Polce, Rebecca. J., Patrick, Megan. E., Lanza, Stephanie. T., Miech, Richard. A., O'Malley, Patrick. M., & Johnston, Lloyd. D. (2018). Reasons for vaping Among U.S. 12th Graders. Journal of Adolescent Health, 62(4), 457–462.

Frohlich, K. L., Potvin, L., Chabot, P., & Corin, E. (2002). A theo- retical and empirical analysis of context: neighbourhoods, smoking and youth. Social Science and Medicine, 54(9), 1401–1417.

Frohlich, Katherine.L., Corin, Ellen, & Potvin, Louise (2001). A theoretical proposal for the relationship between context and disease. Sociology of Health & Illness, 23(6), 776–797.

Galanti, M. R., Coppo, E., Jonsson, E., Bremberg, S., & Faggiano, F. (2014). Anti-tobacco policy in schools: upcoming preventive strategy or prevention myth? A review of 31 studies. Tobacco Control, 23(4), 295–301.

Gentzke, Andrea. S., Creamer, Me. Lisa., Cullen, Karen. A., Ambrose, Bridget. K., Willis, Gordon., Jamal, Ahmed., & King, Brian. A. (2019). Vital signs: tobacco product use among middle and high school students—United States, 2011–2018. Morbidity and Mortality Weekly Report, 68, 157–164.

Giovenco, D. P., Casseus, M., Duncan, D. T., Coups, E. J., Lewis, M. J., & Delnevo, C. D. (2016). Association Between Electronic Cigarette Marketing Near Schools and E-cigarette Use Among Youth. Journal of Adolescent Health, 59(6), 627–634.

Hammig, B., Daniel-Dobbs, P., & Blunt-Vinti, H. (2017). Electronic cigarette initiation among minority youth in the United States. American Journal of Drug and Alcohol Abuse, 43(3), 306–310.

1910 Journal of Youth and Adolescence (2019) 48:1899–1911

Hays, Sharon (1994). Structure and agency and the sticky problem of culture. Sociological Theory, 12(1), 57–72.

Hilton, Shona, Weishaar, Heide, Sweeting, Helen, Trevisan, Filippo, & Katikireddi, SrinivasaVittal (2016). E-cigarettes, a safer alter- native for teenagers? A UK focus group study of teenagers’ views. BMJ Open, 6(11), e013271.

Jakobsen, Janus, Christian, Christian, Gluud, J. ørn, Wetterslev, & Winkel, Per (2017). When and how should multiple imputation be used for handling missing data in randomised clinical trials—a practical guide with flowcharts. BMC Medical Research Metho- dology, 17(1), 162.

Jenssen, B. P., & Wilson, K. M. (2019). What is new in electronic- cigarettes research? Current Opinion in Pediatrics, 31(2), 262–266.

Johnston, L. D., O'Malley, P. M., Miech, R. A., Bachman, J. G., & Schulenberg, J. E. (2016). Monitoring the future national results on adolescent drug use: overview of key findings, 2015. Ann Arbor, MI: University of Michigan Institute for Social Research.

Kim, A. E., Arnold, K. Y., & Makarenko, O. (2014). E-cigarette advertising expenditures in the U.S. 2011–2012. American Journal of Preventive Medicine, 46(4), 409–412.

Kong, G., Morean, M. E., Cavallo, D. A., Camenga, D. R., & Krishnan-Sarin, S. (2015). Reasons for electronic cigarette experimentation and discontinuation among adolescents and young Adults. Nicotine & Tobacco Research, 17(7), 847–854.

Krishnan-Sarin, S., Jackson, A., Morean, M., Kong, G., Bold, K. W., Camenga, D. R., Cavallo, D. A., Simon, P., & Wu, R. (2019). E- cigarette devices used by high-school youth. Drug and Alcohol Dependence, 194, 395–400.

Larcombe, A. N., Janka, M. A., Mullins, B. J., Berry, L. J., Bredin, A., & Franklin, P. J. (2017). The effects of electronic cigarette aerosol exposure on inflammation and lung function in mice. American Journal of Physiology-Lung Cellular and Molecular Physiology, 313(1), L67–L79.

Leatherdale, S. T., & Manske, S. (2005). The relationship between student smoking in the school environment and smoking onset in elementary school students. Cancer Epidemiology, Biomarkers & Prevention, 14(7), 1762–1765.

Leventhal, A. M., Strong, D. R., Kirkpatrick, M. G., Unger, J. B., Sussman, S., Riggs, N. R., Stone, M. D., Khoddam, R., Samet, J. M., & Audrain-McGovern, J. (2015). Association of electronic cigarette use with initiation of combustible tobacco product smoking in early adolescence. JAMA, 314(7), 700–707.

Levinson, A. H. (2018). Nicotine sales to minors: store-level com- parison of e-cigarette versus cigarette violation rates. Nicotine & Tobacco Research, 20(2), 267–270.

Lippert, Adam. M. (2018). Association between school-level pre- valence of electronic cigarette use and student-level use beha- viors, pre-use intentions, and risk perceptions: evidence from the 2014 U.S. National Youth Tobacco Survey. Nicotine & Tobacco Research, 20(2), 231–238.

McCarthy, M. (2016). Teens’ e-cigarette use rises as spending on advertising soars, says CDC. BMJ, 352, i93.

Merianos, A. L., Gittens, O. E., & Mahabee-Gittens, E. M. (2016). Depiction of health effects of electronic cigarettes on YouTube. Journal of Substance Use, 21(6), 614–619.

Moheimani, R. S., Bhetraratana, M., Yin, F., Peters, K. M., Gombein, J., Araujo, J. A., & Middlekauff, H. R. (2017). Increased cardiac sympathetic activity and oxidative stress in habitual electronic cigarette users implications for cardiovascular risk Jama. Cardi- ology, 2(3), 278–284.

Pasch, K. E., Nicksic, N. E., Opara, S. C., Jackson, C., Harrell, M. B., & Perry, C. L. (2018). Recall of point-of-sale marketing predicts cigar and e-cigarette use among Texas youth. Nicotine & Tobacco Research, 20(8), 962–969.

Phua, Joe, Jin, SeungaVenus, & Hahm, Jung Min (2018). Celebrity- endorsed e-cigarette brand Instagram advertisements: effects on young adults’ attitudes towards e-cigarettes and smoking inten- tions. Journal of Health Psychology, 23(4), 550–560.

Roditis, M., Delucchi, K., Cash, D., & Halpern-Felsher, B. (2016). Adolescents' perceptions of health risks, social risks, and benefits differ across tobacco products. Journal of Adolescent Health, 58 (5), 558–566.

Singh, T., Arrazola, R. A., Corey, C. G., Husten, C. H., Neff, L. J., Homa, D. M., & King, B. A. (2016a). Tobacco use among middle and high school students—United States, 2011–2015. Morbidity and Mortality Weekly Report, 65(14), 361–367.

Singh, Tushar., Israel, T. Agaku., Arrazola, René. A., Marynak, Kristy. L., Neff, Linda. J., Rolle, Italia. T., & King, Brian. A. (2016). Exposure to advertisements and electronic cigarette use among US middle and high school students. Pediatrics, 137(5), e20154155

Tudge, J. R. H., Payir, A., Mercon-Vargas, E., Cao, H. J., Liang, Y., Li, J. Y., & O'Brien, L. (2016). Still misused after all these years? A reevaluation of the uses of bronfenbrenner's bioecological theory of human development. Journal of Family Theory & Review, 8(4), 427–445.

U.S. Food and Drug Administration. (2016). Deeming tobacco pro- ducts to be subject to the federal food, drug, and cosmetic act, as amended by the family smoking prevention and tobacco control act; restrictions on the sale and distribution of tobacco products and required warning statements for tobacco products. Final rule. Federal Register, 81(90), 28973–29106.

Williams, R. S., Derrick, J., & Ribisl, K. M. (2015). Electronic cigarette sales to minors via the internetelectronic cigarette sales to minors via the internetelectronic cigarette sales to minors via the internet. JAMA Pediatrics, 169(3), e1563–e1563.

Wills, T. A., Sargent, J. D., Knight, R., Pagano, I., & Gibbons, F. X. (2015). E-cigarette use and willingness to smoke: a sample of adolescent non-smokers. Tobacco Control, 25, e52–e59.

Adam M. Lippert is Assistant Professor of Sociology at the University of Colorado Denver. His research interests include school and neighborhood influences on health, work-family conflict and stress, and family management of austerity.

Daniel J. Corsi is Adjunct Professor at the University of Ottawa and Epidemiologist with the Ottawa Hospital Research Institute. His research interests include substance use among vulnerable populations including adolescents and pregnant women.

Grace E. Venechuk is a graduate student at the University of Wisconsin-Madison and trainee at the Center for Demography of Health and Aging. Her research interests concern the intersections among work, family, and long-term mental and physical health.

Journal of Youth and Adolescence (2019) 48:1899–1911 1911

Journal of Youth & Adolescence is a copyright of Springer, 2019. All Rights Reserved.