Initiation, Progression, and Sustained Waterpipe Use: A Nationally Representative Longitudinal Study of U.S. Young Adults

As traditional cigarette smoking rates decrease in the United States (1), rates of new and emerging tobacco and nicotine product use have been steadily increasing (2). This includes waterpipe tobacco smoking (WTS, or “hookah”), which is increasingly popular among adolescent and young adult populations (3, 4). Although the FDA declared its intention to regulate WTS starting on August 8, 2016 (5), WTS remains exempt from many tobacco control policies (6). Moreover, the popular perception remains that WTS is a fun, relaxing, social activity with few negative associations and repercussions (7, 8). These beliefs and attitudes are concerning given our increasing understanding of the association of WTS with negative health effects, such as cancer, cardiovascular disease, and decreased pulmonary function (9), and potential for dependence (10).

There are three major gaps in the literature that hinder our ability to optimally intervene in this area. First, little is known regarding transitions between various WTS usage states over time. For example, although current cigarette smokers (usually defined as having use at least once in the past 30 days; ref. 11) tend to remain current users, these patterns have not been established regarding WTS. It would be valuable to obtain a more nuanced picture of transitions between various stages of WTS to tailor interventions in terms of timing. Related to this, it will be beneficial to conduct longitudinal studies with longer follow-up periods than have been conducted in the past. To our knowledge, the highest quality longitudinal study was conducted over 6 months (12).

Second, there is a need for better understanding of initiation of WTS, which is generally defined as the transition from never use to ever use. Rates of WTS initiation among U.S. late adolescents and young adults are between 13% and 23%, and factors associated with WTS initiation are not clearly understood (13–15). One established factor that has been clearly associated with WTS initiation is other substance use at baseline; these associations are well-grounded in theory, and the empiric evidence base is strong (12, 15). However, what is not established is whether there are associations between baseline theory-based perceptual predictors, such as knowledge, attitudes, normative beliefs, and intentions, and subsequent initiation, as would be predicted by relevant conceptual models (16). For example, some prior research suggests negative attitudes toward WTS protect against WTS initiation (13), whereas other studies suggest there may be more influence for positive attitudes and normative beliefs (12, 13). Likewise, studies examining knowledge of the harmful components of waterpipe tobacco and effects on WTS are inconclusive (13, 17). It will be important to elucidate associations between these types of predictive factors and initiation, because this will directly influence subsequent educational and policy initiatives.

Third, it is now important to conduct high-quality studies of WTS with large, nationally representative samples that include a broad cohort of emerging adults. Although research to date has been extremely valuable, the vast majority of it has been localized with limited external generalizability. In addition, emerging adulthood is increasingly recognized as a crucial time for the development and consolidation of lifelong habits and addictions (18). Although many studies have involved emerging adults, they have focused on college students (13, 14). Even though it is important to understand use among college students, it is also important to address these behaviors in emerging adults not in college, especially considering that other tobacco use is higher in noncollege young adults than their collegiate counterparts (19).

Therefore, we conducted a nationally representative longitudinal study of U.S. young adults over an 18-month period. We attempted to fill the specific gaps in the literature noted above by focusing on two key, specific aims. First, we aimed to characterize key transitions between WTS states (e.g., never use, ever use, and current use). Second, we aimed to assess independent associations between key predictors, such as knowledge, negative and positive attitudes, normative beliefs, and intentions and subsequent WTS initiation. On the basis of prior work in this area, we developed five a priori hypotheses: (i) Positive attitudes toward WTS will be associated with increased odds of WTS initiation; (ii) negative attitudes toward WTS will be associated with decreased odds of WTS initiation; (iii) favorable normative beliefs about WTS will be associated with increased odds of initiation; (iv) intention will be associated with increased odds of WTS initiation; and (v) WTS knowledge will be associated with WTS initiation.

Participants were members of KnowledgePanel, an online, non-volunteer access panel recruited and maintained by GfK (Growth from Knowledge, formerly Knowledge Networks). This panel was developed through both random digit dialing and address-based recruitment and was designed to be demographically representative of the U.S. population (20). At the time of the study, KnowledgePanel consisted of approximately 50,000 individuals ages 18 and older (20). Panel members are randomly selected to be invited to participate in online surveys and are supplied with e-mail addresses, computers, and Internet access if needed. However, considering the ubiquitous nature of electronic communications in today's world, facilitated access is rarely needed (20).

In March 2013, 3,254 panel members completed a baseline survey about WTS and other health behaviors. Approximately 18 months later, in October 2014, those individuals were invited to participate in the follow-up survey. Of the 3,254 individuals who had completed the baseline survey, 2,170 were still on the panel, 878 had left the panel but were still available for contact, and 206 had left the panel and were unable to be contacted. Therefore, the follow-up survey was sent to the 3,048 individuals with updated contact information. This study was approved by the University of Pittsburgh Institutional Review Board, and all study participants gave informed consent.

This study was guided by two relevant and complementary health behavior theories typically used to explore young adult substance use: the Theory of Reasoned Action (TRA) and the Health Belief Model (HBM). Consistent with the TRA, prior research demonstrates that those with more positive attitudes toward WTS have greater odds of current WTS, intention for future WTS, and WTS initiation, whereas those with more negative attitudes have lower odds of current WTS and WTS initiation (13, 21). Also consistent with the TRA, in a relatively small, regional sample, those with more favorable normative beliefs regarding WTS had greater odds of current WTS and WTS initiation (13, 21). These TRA constructs may in turn affect behavioral intention, which is believed to be a strong predictor of behavior (16). Indeed, self-reported intention to use tobacco products, including waterpipe tobacco, has been found to be an independent predictor of cigarette smoking initiation among young adults (22). The HBM has also been useful in understanding youth substance use, because increased perceptions related to severity and susceptibility have been associated with lower substance use (23). However, constructs such as knowledge that are part of the HBM have been inconsistently associated with outcomes related to WTS. For example, although both cross-sectional and longitudinal studies of college students have found no association between knowledge of the harmful components of waterpipe tobacco, a response of “don't know” has been shown to be protective against current WTS, intention to participate in WTS, and WTS initiation (13, 17).

At both baseline and follow-up, we surveyed participants regarding sociodemographic factors, WTS behavior, and theory-based predictors, such as knowledge, attitudes, and normative beliefs. Both the baseline and follow-up questionnaires were extensively pretested, and median times for completion were 15 and 10 minutes, respectively.

GfK maintains a database of key demographic information about panel members, including age, sex, race/ethnicity, household income, and level of education. To supplement this information, we included items to assess living situation and relationship status.

In the baseline survey, participants were asked to report on ever, past year, and current (i.e., within the past 30 days) WTS. Instructions immediately preceding the WTS items specifically instructed participants to answer about smoking tobacco from a hookah to differentiate from smoking other substances, such as marijuana. The term “hookah” was used instead of “waterpipe” for all WTS-related items, because it is the most common term used in the United States for this device (24). We used the WTS items to operationalize three key states of WTS. Never use was defined as answering “no” to the question, “Have you ever smoked tobacco from a hookah?” Ever use was defined as answering “yes” to “Have you ever smoked tobacco from a hookah?” but answering “0” to “Within the past 30 days, on how many days did you smoke tobacco from a hookah?” Finally, current use was defined as smoking tobacco from a hookah at least 1 day in the past month. Participants were presented with the same items in the follow-up questionnaire. This information was used to chart the transitions between the key states of WTS. Participants who transitioned from never use to ever use or current use were defined as initiators, whereas those who transitioned from ever use to current use were defined as progressors.

Attitudes toward WTS were assessed with a 6-item scale used in previous research studies (13, 21). These items asked participants whether they believed WTS was attractive, romantic, fun, relaxing, harmful, and addicting, each on a 5-point response scale (Definitely No, Probably No, Don't Know, Probably Yes, Definitely Yes). The four items assessing attitudes toward positive attributes (attractive, romantic, fun, relaxing) were grouped to reflect an overall positive attitude summary scale, whereas the two items assessing attitudes toward negative attributes (harmful, addicting) were grouped to reflect an overall negative attitude summary scale. A higher score on the positive attitude scale indicated a favorable attitude toward WTS, whereas a higher score on the negative attitude scale indicated a less favorable attitude toward WTS. Both summary scales demonstrated strong reliability (Cronbach's α = 0.88 and 0.85, respectively).

Normative beliefs were measured in two complementary ways used in substance use research (13, 21, 25). The first asked participants to estimate the percentage of people their age that have ever smoked tobacco from a hookah, and the second asked participants to indicate how socially acceptable it is for people their age to smoke tobacco from a hookah. This item included a 4-point response scale ranging from “Very socially acceptable” to “Not socially acceptable.”

WTS intention was assessed through the item, “Do you intend to smoke tobacco from a hookah at any time in the rest of your life?” The 4-item response scale for this item asked respondents to select “Definitely Yes,” “Probably Yes,” “Probably No,” or “Definitely No.” Consistent with prior research (26), for analysis, any answer other than “Definitely No” was defined as intention to participate in WTS.

Knowledge about the harmful components of waterpipe tobacco was measured using a set of items asking participants to compare smoking a single cigarette with a single hookah tobacco smoking session and estimate which contained more of the following: tar, nicotine, carcinogens, carbon monoxide, and heavy metals (13, 17). Although a typical hookah session is longer than smoking a single cigarette, this wording was used to be consistent with, and comparable with, other literature on this topic (13, 27, 28). A 4-item response scale allowed respondents to choose “Definitely Hookah,“ “Probably Hookah,“ “Probably Cigarettes,“ or “Definitely Cigarettes.” On the basis of established research (27, 29), an answer of “Definitely Hookah” or “Probably Hookah” was considered to be correct, whereas an answer of “Definitely Cigarettes” or “Probably Cigarettes” was considered to be incorrect. Internal consistency was measured using Cronbach's α and demonstrated strong reliability among items (α = 0.87).

We included all individuals with complete data on the dependent variable at baseline and follow-up. Because less than 1% were excluded for incomplete data, this is unlikely to have affected results. We assessed differences in sociodemographic characteristics between responders and nonresponders using Rao-χ² tests and Cramer's V. In addition, we calculated simple frequencies and percentages to characterize use patterns of the dependent variable from baseline to follow-up. We assessed associations between independent variables at baseline and our dependent variable at follow-up using logistic regression, adjusting for a comprehensive set of sociodemographic covariates in multivariable models. We conducted sensitivity analyses with a more parsimonious set of covariates to confirm the robustness of our results. All analyses were conducted using survey weights provided by GfK to estimate effects for the general U.S. population, as well as correct for any under- or oversampling. We defined statistical significance with a two-tailed α of 0.05. Data were analyzed using Stata 14 (StataCorp; ref. 30).

A total of 1,796 individuals completed the survey (RR = 59%). The final sample for analysis consisted of 1,785 individuals with complete data on WTS items for both baseline and follow-up. Responders and nonresponders did not significantly differ on WTS (P = 0.08), age (P = 0.15), sex (P = 0.07), or race/ethnicity (P = 0.19). Therefore, those with complete data represented the complete population in terms of basic socio-demographics. In addition, study-specific survey weights were applied to all analyses to adjust for nonresponses well as under- or oversampling.

The weighted age distribution of our sample was approximately equal, with the largest group of 27 to 30 years (31%). The sample was 50% female, and just over half of the sample identified as White, non-Hispanic (57%); followed by Hispanic (22%), Black, non-Hispanic (13%); and Other (9%). In addition, half of the sample reported being single (50%), whereas a plurality reported living with a parent or guardian (40%), earning an income of $75,000 or greater (47%), and having an education level of high school or less (43%; data not shown).

The trajectories of WTS behavior between baseline and follow-up are depicted in Fig. 1. Of the baseline never smokers (n = 1,232, 69%), 94% remained never smokers at follow-up. Similarly, 96% of ever smokers at baseline (n = 446, 25%) remained ever smokers at follow-up (i.e., they did not progress to current smoking). This is in contrast to baseline current smokers (n = 107, 6%), of whom 17% remained current smokers. Of note were the 4% of baseline ever smokers that progressed to current smoking at follow-up and were considered progressors. In addition, of the 1,232 participants reporting never WTS at baseline, 7% reported initiation at follow-up and were considered initiators.

Sociodemographic characteristics of baseline WTS nonsmokers (Table 1) are similar and consistent with whole sample characteristics. In bivariable analyses, we found significant associations between WTS initiation and race/ethnicity and yearly household income (Table 1). Those who initiated WTS were more often White, non-Hispanic individuals with an annual household income of $30,000 to $74,999.

In multivariable analyses, participants with increased positive attitudes toward WTS had significantly greater odds of WTS initiation (Table 2). No significant associations were found between negative attitudes toward WTS and WTS initiation or WTS normative beliefs and WTS initiation. However, WTS intention, which was reported by 22% of baseline nonsmokers, was significantly associated with increased odds of WTS initiation [adjusted OR (AOR) = 7.0; confidence interval (CI), 3.5–13.7]. Of the five knowledge items, only the knowledge that WTS exposes the user to greater amounts of nicotine compared with cigarettes was significantly associated with WTS initiation (AOR = 2.8; CI, 1.1–7.0). Overall knowledge was not significantly associated with WTS initiation. Sensitivity analyses using a more parsimonious set of covariates yielded consistent findings.

In this nationally representative longitudinal study of U.S. young adults, we found that, over an 18-month period, 4% of baseline ever smokers progressed to current smoking and that approximately 7% of baseline never smokers initiated WTS. A closer examination of WTS initiation found that positive attitudes toward WTS, knowledge of WTS-related nicotine exposure compared with cigarettes, and baseline WTS intention were significantly associated with initiation. In contrast, negative attitudes toward WTS, normative beliefs toward WTS, and knowledge of other WTS components were not significantly associated with WTS initiation.

We also found that the vast majority of baseline never smokers (94%) and ever smokers (96%) maintained their status at the 18-month follow-up. However, it is notable that the prevalence of never smoking decreased by 4 percentage points over the 18-month period, whereas ever smoking increased by approximately 7 percentage points. This suggests that more individuals are being introduced to WTS in some capacity. In contrast to the sustained never and ever smoking, the majority of baseline current smokers (83%) reverted to ever smoking by follow-up. It is possible that some of these individuals quit WTS over the follow-up period. Indeed, in studies of U.S. waterpipe tobacco users, many have expressed intention to quit at some point in the future (31, 32). However, almost 17% of current smokers sustained their use through the 18-month period, consistent with some international studies (3). In addition, 4% of baseline ever smokers progressed to current smoking over the follow-up period. These findings suggest a potential for dependence on WTS. Some users of WTS report feeling “hooked” on WTS (33), and a recent study found that half of past year WTS users endorsed at least one dependence item on a 6-item WTS dependence scale (10).

Of the baseline nonsmokers, 7% initiated WTS at follow-up. Although this initiation rate is lower than in studies focused on college students, it is similar to the only other study of initiation among a more general young adult population (12). A possible reason for the lower initiation rate in this study is that this sample was not limited solely to college students. College entry and progression through the initial years of college have been found to be associated with WTS (13, 15). Thus, by including noncollege populations, we may have reduced the number of higher risk individuals in our sample.

Positive attitudes toward WTS significantly predicted WTS initiation, whereas negative attitudes did not. The association between positive attitudes and WTS initiation is consistent with studies of college students (13). This is not surprising, as many U.S. waterpipe users view WTS to be a fun activity during which they can socialize, party, or relax (7). However, the lack of association between negative attitudes and WTS initiation is contrary to studies that found negative attitudes to be protective against WTS use or initiation (13, 21). This is akin to cigarette smokers, for whom negative outcome expectations of cigarette smoking may not influence the decision to start smoking (34). It is possible that, similar to some cigarette smokers, those who participate in WTS may rationalize their behavior by focusing on the positive aspects of WTS and modulating the negative aspects (35).

We also found that knowledge about greater nicotine exposure in WTS compared with cigarettes was associated with higher odds of smoking initiation. It is possible that individuals who had already begun to show interest in WTS were more knowledgeable about this fact than those who had not. Also, it is possible that the novelty and general positive appeal of WTS outweighs the knowledge of the presence of nicotine. This phenomenon has been noted among users of other non-cigarette nicotine and tobacco products, such as e-cigarettes, snus, and nicotine dissolvables (36). In a series of focus groups, about half of the young adult participants reported being willing to try the products despite knowledge of nicotine content and potential negative health effects, because they were perceived as attractive, modern, and fun (36).

The strongest association with WTS initiation was baseline intention. Those individuals who reported that they intended to smoke tobacco from a hookah at one point in the rest of their lives had approximately 7 times greater odds of initiating WTS compared with those who reported no intention to smoke tobacco from a hookah. This is especially concerning, because, consistent with other research (22), approximately one quarter of baseline nonsmokers reported an intention to initiate WTS at one point in the rest of their lives. Extrapolating this to the greater population, almost 9 million baseline nonsmokers ages 18 to 30 may have some intention to participate in WTS (2014 estimated U.S. Census population of 57,702,876 × 69% baseline never smokers × 22% reporting intention; ref. 37). Considering the strong association between baseline intention to participate in WTS and subsequent initiation of WTS, it is clear that this is a ripe opportunity for primary prevention efforts among this particular age group.

Although certain theory-based predictors, which may be addressable through intervention, were associated with WTS initiation, the majority of sociodemographic factors were not. This suggests that primary prevention efforts to curb WTS initiation in the general 18- to 30-year-old population may be more effective if they address these theory-based predictors rather than focusing on any one group of individuals. This is consistent with recommendations for the college population, which also found no association between sociodemographic factors and WTS initiation (13). However, this is in contrast to secondary prevention efforts addressing current or ever use, for which efforts may be more beneficial when focused on populations that are younger, white, and male (38). These findings provide important distinctions that can help guide targeting of interventions for future research or for the practice of health professionals.

The two strongest predictors of WTS initiation, positive attitudes and intention, may be addressed through similar prevention and intervention means. Despite the associated negative health effects and potential for addiction, WTS is still viewed as a fun, attractive social activity. This is especially true in online advertisements, where WTS is portrayed as social and pleasurable; however, health warnings or mention of nicotine content are rare (39). Likewise, studies have shown that the waterpipe tobacco industry uses questionable marketing and labeling practices, such as misleading descriptors on packaging and omission of health effects (40, 41).

This study was limited by its self-report survey design. Biochemical validation was not used to confirm WTS status. However, participants were assured of the confidentiality of their answers and had little incentive to be untruthful. This study was also limited by the fact that, despite having a large sample size, there was not enough power to fully examine factors associated with WTS progression (i.e., baseline use followed by increases). The progression from WTS experimentation to more regular use is an area that warrants deeper investigation, and future studies should seek to examine this trajectory specifically by recruiting a greater volume of baseline experimenters. Another limitation is the stem used for the knowledge item, which asked participants to compare their knowledge of toxicant exposure from a WTS session compared with smoking a single cigarette. Because a typical WTS session takes longer than smoking a single cigarette, this item could have confused participants, potentially leading them to answer incorrectly. In addition, this study examined the association between WTS initiation and certain theory-based predictors and did not include information about other substance use. Future studies may seek to examine both sets of factors simultaneously. Finally, although this study had a longitudinal design, data collection at more time points over a longer period would have yielded data that may have been amenable to more complete longitudinal analysis techniques, such as growth mixture modeling.

In conclusion, we found that almost 20% of current WTS users maintained their use through an 18-month follow-up period. We also found initiation and progression rates of 7% and 4%, respectively. Initiation was associated with certain theory-based predictors, such as positive attitudes toward WTS and intention to participate in WTS, but not most sociodemographic factors, suggesting specific targets for prevention efforts. These efforts may be most effective if they counteract positive attitudes toward WTS and intention to participate in WTS in the future.

No potential conflicts of interest were disclosed.

The funding agency had no role in the design and conduct of the study, collection, management, analysis, and interpretation of the data, and preparation, review, or approval of the manuscript.

Conception and design: J.E. Sidani, A. Shensa, B.A. Primack

Development of methodology: J.E. Sidani, A. Shensa, B.A. Primack

Acquisition of data (provided animals, acquired and managed patients, provided facilities, etc.): J.E. Sidani, B.A. Primack

Analysis and interpretation of data (e.g., statistical analysis, biostatistics, computational analysis): J.E. Sidani, A. Shensa, J.G. Yabes, B.A. Primack

Writing, review, and/or revision of the manuscript: J.E. Sidani, A. Shensa, M.R. Naidu, J.G. Yabes, B.A. Primack

Administrative, technical, or material support (i.e., reporting or organizing data, constructing databases): A. Shensa, B.A. Primack

Study supervision: B.A. Primack

This work was supported by the NCI at the NIH (R01-CA140150; awarded to B.A. Primack).

The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked advertisement in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.