An Investigation of Definitions of Experimental Vaping among Youth

Abstract

Objectives:

Identifying which youth are experimenting with vaping could aid the development and evaluation of targeted media campaigns and research and surveillance activities. In this study, we sought to identify behavioral definitions that best differentiate youth experimenters and established vapers.

Methods:

We conducted an online survey with a non-probability sample of 1500 youth aged 13–17 who reported ever vaping. Based on recency and lifetime vaping, we constructed 12 definitions of experimenters versus more established or recent users. We examined how well each definition discriminated between experimenters and established/recent users based on correlates (eg, vaping dependence, harm perceptions) using multivariate tests of mean differences, controlling for multiple variables, and ratios of between- to within-group variance.

Results:

Four definitions best distinguished between experimenters and more established/recent users (ie, had greater Hotelling T² statistics for the multivariate tests and higher ratios of between- to within-group variance). Three of these 4 identified experimenters as those with no past 30-day vaping.

Conclusion:

Ever vapers are not a monolithic group. Our results suggest that past 30-day use is an important criterion for distinguishing experimenters from more established users. Understanding nuances between user groups could lead to greater differentiation among ever vapers and aid campaign targeting.

Vaping (ie, using e-cigarettes and other electronic nicotine delivery systems) has been the most common type of nicotine/tobacco use among youth in the United States (US) since 2014.¹ The prevalence of current vaping among youth increased dramatically from 2017 to 2018.² Youth vaping poses a public health threat because it exposes users to nicotine, which harms adolescent brain development and leads to dependence,³ and exposes users to harmful chemicals in aerosol.⁴ Additionally, youth vaping potentially serves as a gateway to use of combustible tobacco products.^5,6 Nicotine exposure during adolescence enhances the rewarding and reinforcing effects of nicotine in adulthood, making it more likely to continue to be addicted to nicotine as an adult.^7,8

Experimentation with vaping may escalate to regular use,⁹ but few studies have examined how experimenters differ from regular users. In contrast, considerable research has addressed the uptake process for cigarette smoking and characterized smokers at different phases of that process.^10,11 Definitions for these phases of cigarette smoking (susceptibility, initial trying, experimentation, current smoking, and established smoking) are typically based on lifetime use, recency, and frequency.

There are no consistent or widely used behavioral definitions for the stages of youth vaping uptake, including the experimentation phase.^{12, 13, 14} Current literature defines experimental or infrequent vaping based on a myriad of definitions related to lifetime quantity (using general terms like “times” or “sessions” or specific quantities like number of refill pods),^{15, 16, 17, 18, 19, 20} number of days used in past 30 (eg, 1–5 days^12,21 or 1–19 days^22,23), or rate of use (eg, less than once a week^{24, 25, 26, 27} or “rarely”²⁸). Some definitions use combinations of these criteria.^29,30 Definitions also can vary by age; for example, the Population Assessment of Tobacco and Health (PATH) study includes definitions for “experimental” versus “regular” users among adults and “very light” versus “not light” for youth.^31,32 There is no scientific consensus or best practice guidance on measurement of youth vaping experimentation, and the evidence base to support any particular categorization is limited.

Gaining a better understanding of experimentation with vaping among youth is potentially important to inform the US Food and Drug Administration’s (FDA’s) research and surveillance efforts, as well as current and future campaigns to address the problem of youth vaping. The FDA recently expanded The Real Cost campaign to address the increase in youth vaping.³³ This campaign targets youth who are open to trying vaping and youth who have ever vaped. In developing the original The Real Cost campaign to prevent youth cigarette smoking, the FDA utilized an audience segmentation approach to identify teens who were most at risk for smoking – susceptible non-triers and smoking experimenters.³⁴ Identifying youth who are experimenting with vaping, as distinct from those who are more established vapers, could serve as an added step in a similar process of audience segmentation to create future targeted messages about vaping.

Given limited research on how to operationalize lifetime quantity and recency of use for vaping experimentation, we reviewed the literature on cigarette experimentation definitions for guidance. In defining experimental cigarette use based on lifetime quantity of smoking, researchers have varied the number of cigarettes smoked or number of smoking sessions. Most commonly a cut-off of having smoked fewer than 100 cigarettes in one’s lifetime was used to categorize experimental smokers.^{10,35, 36, 37, 38} The most common operational definition of experimental cigarette use incorporating recency of use is having ever smoked, but not having smoked in the past 30 days or past month.^{39, 40, 41, 42, 43, 44, 45} These past studies of cigarette use experimentation can inform the development of new measures of experimental vaping that varies different cut-offs for lifetime use of vaping products and recency of use.

The purpose of this study was to examine definitions for experimental versus more established vaping and identify those that best differentiate those groups in terms of key correlates, such as perceptions of harm and dependence. Results of this study can help inform future communication efforts around youth experimentation with vaping.

METHODS

Participants and Procedures

In June-July 2019, 1500 youth aged 13–17 who reported ever vaping completed an online survey. Youth were recruited through their parents, who were members of an online convenience sample of consumers managed by Kantar (formerly Lightspeed). Kantar recruits panel members on an ongoing basis via social media, online ads (eg, banner placements), and affiliate corporate networks. For this survey, parents who were part of the Kantar panel at the time of this study received a recruitment email and provided permission for their child to complete the survey. Children then assented, completed a screener, and, if eligible, completed the survey. Eligibility criteria were being aged 13–17 and responding “yes” to the question “Have you ever vaped, even one or two times?” The parent’s Kantar account received compensation of 100 points (approximately $1) if the youth completed the survey and 50 points (approximately $0.50) if the youth assented to participate but was not eligible.

Measures

Some measures used in this study were novel, whereas others have been used and validated previously.

Lifetime use and recency of vaping.

The behavioral definitions for experimental vaping (described in the Data Analysis section and in Table 1) relied on measures of lifetime vaping and recency. We measured lifetime use of vaping products and recency of use and constructed 12 potential definitions of experimental vaping with different cut-offs for number of times vaped in lifetime and/or last time vaped.

Table 1.

Definitions of Experimental versus More Established/Recent Vaping among Youth

Definition	Experimental Use	More Established/Recent Use (Referent)	Excluded from Definitions
1	Used 1 to 99 times in lifetime	Used 100+ times in lifetime	None
2	Used 1 to 99 times in lifetime AND used in past year	Used 100+ times in lifetime AND used in past year	Last use was more than one year ago
3	Used 1 to 99 times in lifetime AND used in past 6 months	Used 100+ times in lifetime AND used in past 6 months	Last use was more than 6 months ago
4	Used 1 to 99 times in lifetime AND did not use in past 30 days	Used in past 30 days	Used more than 99 times in lifetime AND no past 30-day use
5	Used 1 to 5 times in lifetime	Used more than 5 times in lifetime	None
6	Used 1 to 5 times in lifetime AND used in past year	Used more than 5 times in lifetime AND used in past year	Last use was more than one year ago
7	Used 1 to 5 times in lifetime AND used in past 6 months	Used more than 5 times in lifetime AND used in past 6 months	Last use was more than 6 months ago
8	Used 1 to 5 times in lifetime AND did not use in past 30 days	Used in past 30 days	Used more than 5 times in lifetime AND did not use in past 30 days
9	Used 1 to 50 times in lifetime	Used more than 50 times in lifetime	None
10	Used 1 to 50 times in lifetime AND used in past year	Used more than 50 times in lifetime AND used in past year	Last use was more than one year ago
11	Used 1 to 50 times in lifetime AND used in past 6 months	Used more than 50 times in lifetime AND used in past 6 months	Last use was more than 6 months ago
12	Used 1 to 50 times in lifetime AND did not use in past 30 days	Used in past 30 days	Used more than 50 times in lifetime AND no past 30-day use

The survey assessed lifetime use of vaping products with the item: “How many times have you vaped in your entire life? Your best guess is fine.” with response options: 1 time, 2 to 5 times, 6 to 15 times, 16 to 25 times, 26 to 50 times, 50 to 99 times, and 100 or more. Participants indicated last use (ie, recency) in response to the item: “When did you last use a vaping product, even one or two times?” with response options: Earlier today, Not today but sometime during the past 7 days, Not during the past 7 days but sometime during the past 30 days, Not during the past 30 days but sometime during the past 6 months, and Not during the past 6 months but sometime in the past year, 1 to 4 years ago, and 5 or more years ago.

Other tobacco use.

Participants also indicated whether they had ever smoked cigarettes, cigars, or cigarillos, used smokeless tobacco, and smoked hookah and the most recent time they had used each of those products; past 30-day use of a given tobacco product was considered current use.

Correlates

We examined a range of different correlate measures that would differentiate experimental vapers from more established vapers. These knowledge, attitude and perception correlates were adapted from the PATH study or other published studies.

Vaping among friends.

One item assessed vaping among friends: “How many of your friends vape?” (from 1 [4 or more of my friends] to 5 [None of my friends], adapted from Cooper et al).⁴⁶

Dependence.

Two items assessed dependence on vaping: “Right now, how much do you want to vape?” (1 [Extremely] to 5 [Not at all]; adapted from Pearson et al⁴⁷) and “How much of the time have you felt the urge to vape in the past 24 hours?” (1 [All the time] to 6 [Not at all]; adapted from West and Hajek⁴⁸).

Knowledge, attitudes, and perceptions.

Participants indicated how wrong they thought it was for “someone your age to vape,” “your friends feel it would be for you to vape,” and “your parents feel it would be for you to vape” on a 4-point scale (1 [Not wrong at all] to 4 [Very wrong]; adapted from Brooks et al⁴⁹). They also rated vaping on a scale from 1 [Very unenjoyable] to 5 [Very enjoyable], adapted from Farrelly et al.⁵⁰ The survey assessed comparative harm perception with the item: “Compared to smoking cigarettes, would you say that vaping products are…” (1 [Much less harmful than cigarettes] to 5 [Much more harmful than cigarettes] adapted from Tan and Bigman.)⁵¹ Finally, participants indicated their agreement on a 5-point scale (1 [Strongly Disagree] to 5 [Strongly Agree]) with 13 potential consequences of vaping (“If I vape, I will…”): Inhale microscopic metal particles such as nickel, tin, and lead; Develop serious health problems; Miss out on activities I enjoy; Damage my body; Get stunted lungs; Become addicted; Decrease my sports performance; Be controlled by nicotine; Be controlled by vaping; Be unable to stop when I want to; Expose my brain to nicotine; Be a bad influence on others; and Be more likely to smoke cigarettes.

Demographics.

The survey also assessed sex, age, and race and ethnicity.

Data Analysis

Definitions and sample sizes.

We constructed 12 potential definitions for experimental vaping versus a referent group, based on lifetime and last use; these definitions (Table 1) were modeled off parallel categorizations for cigarette smoking and vaping^{12,17,18,30,52} as identified in our literature review.

Depending on the definition, the referent group represented more established and/or more recent users. For each definition, we calculated the proportion of ever-users who would be considered experimenters or more established/recent users. Of the 12 definitions, 9 excluded some study participants as falling outside these 2 groups because the participants could be considered former users; for example, Definition 2 excludes those whose last use was more than one year ago and Definitions 4, 8, and 12 exclude those who reached a certain threshold of number of times of use and whose last use was more than 30 days ago.

Of the 1500 survey respondents, 23 did not answer the item about number of incidences of lifetime vaping. Of those 23, 3 did not respond to the item about recency, so their past 30-day vaping behavior was unknown; 12 indicated that they had not vaped in the past 30 days; and 11 indicated that they had vaped in the past 30 days. Based on the construction of the definitions, the resulting sample size was N = 1477 for Definitions 1–3, 5–7, and 9–11 because these definitions require information about lifetime use to categorize respondents. The sample size was N = 1488 for Definitions 4, 8, and 12 because these definitions can categorize respondents who have vaped in the past 30 days as “more established/recent vapers,” regardless of whether they provide information about lifetime vaping experience.

Analytic approach.

We considered definitions to be superior if they provided better discrimination between the experimenter and more recent/established groups. To identify which definitions were best for discriminating between the groups, we conducted a set of analyses using key correlates. These key correlates were the items on vaping among friends (1 item), vaping dependence (2 items), and perceptions (eg, perceived harm, perceived enjoyment; total of 18 items). Key correlates were analyzed separately. We felt that data reduction techniques with these measures would be inappropriate because response options varied by question and were often on different scales.

To assess the performance of the different definitions, we contrasted those classified as experimental vapers versus those defined as more established or recent vapers. Our assumption is that a better definition would result in groups that were homogeneous within group but different across groups on measures related to vaping (key correlates). We used 2 types of analyses: tests of a set of mean differences of correlates, controlling for multiple correlates, between groups and ratios of between-group variance over within-group variance with respect to correlates. Both tests provide information about the extent to which the experimenter group and the more established/recent group are homogenous within group and heterogenous between groups. The test combines information from the set of correlates and presents a test statistic at the level of the definition, whereas results from the ratio of between- to within-group variance are at the level of the correlate. We conducted analyses using Stata v.15.1 (College Station, TX).

Test of mean differences.

We conducted a test of mean differences in correlates between experimental and more established/recent use groups as determined by our experimental vaping definitions. Specifically, using Stata’s Hotelling command, we calculated the Hotelling T² statistic.⁵³ Greater T² statistics indicate greater differences between the experimenter and referent groups.

Ratio of between to within-group variance.

We examined the extent to which experimentation definitions result in distinct, homogeneous segments of experimenters and more established/recent users based on methods used in audience segmentation and cluster analysis to identify distinct groups of people based on shared characteristics. For each of the correlates, we calculated the Caliński-Harabasz pseudo-F index (C-H index), which represents the ratio of between-group variance over within-group variance for a particular grouping.⁵⁴ Higher scores on this measure indicate greater between-group heterogeneity with lower within-group heterogeneity, which suggests more distinct segmentation. We identified the definitions with the first, second, and third highest C-H scores for each correlate.

Profiles of experimenter groups.

Once we determined which definitions best discriminated experimenters and referent groups, we examined the sociodemographic and tobacco use characteristics of those groups of experimenters.

RESULTS

Among the 1488 respondents who were eligible for inclusion in at least one definition, the majority were aged 15–17 (81.1%) and female (60.3%). The sample was 56.9% white non-Hispanic, 20.4% Hispanic, 12.0% black/African-American non-Hispanic, and 10.6% other races. A minority used at least one of the 4 other tobacco products in the past 30 days: 15.1% cigarettes, 9.5% cigars or cigarillos, 5.5% smokeless tobacco, and 4.6% hookah.

As Figure 1 shows, the proportion of respondents classified as experimenters varied from 25.5% (Definition 7) to 81.4% (Definition 1) and more established/recent users from 17.7% (Definition 3) to 58.7% (Definition 5). Among the 9 definitions that excluded participants because they were considered former users, the proportion of the sample that was excluded ranged from 3.4% (Definition 4) to 22.0% (Definitions 3, 7, and 11).

Figure 1. Proportion of Youth Categorized as an Experimenter, More Established/Recent User (Referent Group), or Excluded from Analysis, by Experimental Use Definition, among Youth Who Had Ever Vaped (N = 1477–1488).

Note. Eleven participants provided information about past 30-day but not lifetime vaping. These participants were eligible for inclusion in the referent group for definitions 4, 8, and 12. Thus, the sample size is 1488 for definitions 4, 8, and 12 and 1477 for the other definitions.

Test of Mean Differences

As Figure 2 shows, the definitions with the greatest Hotelling T² statistics (indicating greater differences between the experimenter and referent groups) were Definitions 4 (T² = 705.0), 5 (667.5), 8 (815.2), and 12 (714.5). T² statistic values for the other definitions ranged from 302.5 (Definition 3) to 529.1 (Definition 6).

Figure 2. Tests of Differences in Means of Correlates between Experimenters and More Established/Recent Users Using Hotelling T² Statistic.

Ratio of Between to Within-group Variance

Table 2 and Table 3 include the C-H index scores for all ordinal correlates by definition. The definitions with the greatest number of high C-H index scores were Definition 4 (the highest score for 2 of the 21 correlates, second highest for one correlate, and third highest for 2 correlates), Definition 5 (the highest score for 8 correlates, second highest for 8 correlates, and third highest for one correlate), Definition 6 (second highest score for 3 correlates and third highest for 5 correlates), Definition 8 (the highest score for 10 correlates, second highest for 6 correlates, and third highest for 4 correlates), and Definition 12 (second highest score for 3 correlates and third highest for 7 correlates).

Table 2.

Ratio of between to within-group Variance Based on Caliński-Harabasz (C-H) Index Scores for Each Correlate for Definitions 1–6

Table 3.

Ratio of between to within-group Variance Based on Caliński-Harabasz (C-H) Index Scores for Each Correlates for Definitions 7–12

Profiles of Experimenter Groups

We examined age, sex, race/ethnicity, and current tobacco use among the experimenters in the definitions that appeared to provide the best discrimination between groups, specifically Definitions 4 (N = 605), 5 (N = 610), 8 (N = 424), and 12 (N = 585). Full results appear in Supplemental Table 1. Overall, the demographic make-up of experimenters according to Definitions 4, 8, and 12 was similar. Those classified as experimenters according to Definition 5 tended to be younger (20.5% were aged 13–14 vs 13.6% to 16.0% for the other definitions), male (35.1% male vs 30.4% to 32.4%), and a greater percentage used cigarettes (5.6% vs 3.2% to 4.4%), cigars or cigarillos (4.8% vs 2.9% to 3.7%), smokeless tobacco (2.7% vs 1.0% to 1.5%), and hookah (2.2% vs 0.7% to 1.0%). Distributions of race/ethnicity among experimenters were similar across definitions. Across each of the analyses, past 30-day use was included as a component for all experimenter definitions.

DISCUSSION

Across tests, the definitions that showed the greatest distinction between vaping experimenters and more established/recent vapers were Definitions 4, 5, 8, and 12. These definitions had the highest Hotelling T² statistics in the tests of mean differences and highest ratios of between- to within-group variance.

Definitions 4, 8, and 12 are constructed similarly; experimenters are vapers who have not used in the past 30 days and have not exceeded a certain number of incidences of lifetime use (no more than 99 times, 5 times, and 50 times, respectively), and the referent category is past 30-day vapers. In contrast, Definition 5 is based solely on a threshold of lifetime use (less than 5 times in lifetime), irrespective of current use. Thus, experimentation according to Definition 5 includes a wide range of experiences; these experimenters could be adolescents who tried vaping for the first time today with the intention of using again tomorrow, or they may have tried vaping one time a year ago and never had the intention to vape again. Definitions 4, 8, and 12 result in a narrower interpretation of experimentation because not only are the number of episodes of use limited, but the timing is limited (ie, not in the past 30 days).

The ability of Definition 5 to discriminate well between experimenters and more established users, despite not considering past 30-day use like Definitions 4, 8, and 12 do, may be explained, in part, by a correlation between lifetime and past 30-day use. That is, youth who have vaped on more occasions in their lifetime are also more likely to have vaped in the past 30 days. This results in extensive overlap between Definition 5 (which does not account for past 30-day use) and Definition 8 (which does account for past 30-day use); specifically, the majority (74%) of respondents classified as established users in Definition 5 were also considered established users in definition 8. Nevertheless, the fact that Definition 8 outperformed Definition 5 (ie, more instances of high C-H scores and a greater Hotelling T² statistic) suggests that considering past 30-day use improves discrimination between experimental and established users.

None of the definitions is inherently superior, but each has different implications for placing adolescents along the continuum of uptake for vaping. The experimenter group in Definitions 4, 8, and 12 could be at risk for re-initiating or continuing experimentation. The absence of past 30-day use among the experimenters in these definitions does not inherently classify them as former users; rather, this lack of recent use could represent a “pause” in their use or simply reflect the irregularity of the uptake process. The experimenter group in Definition 5 could fit at many different points along the uptake continuum, at risk for re-initiation, continuation of experimentation, or escalation. In short, Definitions 4, 8, and 12 capture the non-current experimenters, whereas Definition 5 could detect experimenters from a broader group that are at different points of the uptake process. These results suggest that past 30-day use is an important criterion for distinguishing experimenters from more established users. Practitioners could consider using past 30-day use as one of the key measures when discerning experimenters from established vapers.

Defining experimenter and established vapers has potential implications for youth focused vaping prevention campaign research and targeting. Additional segmentation, especially within the group of past 30-day users, could help the FDA develop profiles of at-risk youth using these definitions, as they did previously for cigarette smoking.³⁴ Target audience profiles improve health communication campaigns by providing specific and accurate target audience definitions. The more accurately campaign administrators can define their target audience, the better able they will be to tailor and target campaign messages to the audience segment most in need of intervention. Understanding behavioral definitions that best indicate when youth become established users has implications for vaping policy and health marketing campaigns as these definitions can be used to illuminate crucial intervention points at which opportunities exist to interrupt vaping progression.

Limitations

Due to an oversight, the measure of lifetime use was not mutually exclusive for the number 50. If a respondent vaped 50 times, they had 2 response options (26 to 50 times, 50 to 99 times). We used 50 as a cut-off in definitions 9–12 so it is possible that there was some misclassification for these definitions. However, we think the misclassification impact is relatively minimal. Even if the respondent noticed the mistake in the response options, they likely chose the category (26–50 or 50–99) that best reflected their level of use given the broad range in these categories. Thus, we do not expect that oversight in the response options for this measure to have had a substantial effect on our results and conclusions. Within the groups of experimenters and established/recent vapers created by the present definitions, there could be further differences based on other key correlates (ie, knowledge, attitudes, and behaviors). This research used a nonprobability sample, which limits generalizability. An additional limitation is that the survey item: “How many times have you vaped in your entire life?” is open to interpretation. Users could have considered one “time” to be one puff, one session, or one day. Future qualitative research could investigate how youth interpret “time” in this context and potentially identify terminology that is clearer among youth who vape. Finally, there are other possible definitions of experimental use than those we used in this study. Future qualitative research could explore how youth experiment with vaping and perceive experimentation may uncover new ways to define experimentation differently in future studies. The correlates used in this study had not all been derived from validated measures. Future studies may investigate the validity of these measures and replicate the current exploratory study.

IMPLICATIONS FOR TOBACO REGULATION

The FDA’s The Real Cost e-cigarette prevention campaign focuses on youth susceptible to vaping and those who have ever vaped. Greater differentiation within the “ever-vaped” group could aid campaign targeting. We created 12 definitions of experimental vaping (versus more established and/or recent vaping) based on lifetime use and recency and examined differences between groups in risk perceptions and other correlates. Definitions using “past 30 days” as the cut-off for recency tended to provide the greatest differentiation between youth who have experimented and more established/recent users.

Utilizing a standard definition of youth experimentation with vaping versus established vaping could improve surveillance by making findings easier to compare across studies. Future research could examine how well the different definitions identified here predict changes in vaping behavior over time and how definitions might vary for adults versus youth.

These definitions also have the potential to enhance the design and targeting of interventions, such as public education campaigns. For example, these definitions could be used to characterize experimenters as an audience segment for vaping-related public health messages based on the group’s shared knowledge, attitudes, beliefs, and behaviors; moreover, they also could be used to develop messages targeting experimenters.

Human Subjects Approval Statement

Study procedures and materials were approved by the Institutional Review Boards at RTI International (IRB #STUDY00020494) and the US Food and Drug Administration Center for Tobacco Products (IRB #2019-CTP-046) and by the Office of Management and Budget (OMB #0910–0810).

Supplementary Material

1

Supplemental Table 1: Sociodemographic and Tobacco Use Behavior Characteristics of Experimenter Groups According to the Best Discriminating Definitions