Objective: To address observations that the smoking identities of youth are valid descriptors of their smoking behaviour, we examined the relationships between self-reported smoking identities, perceived levels of addiction, and established taxonomies of smoking behaviour of youth.
Method: Cross-sectional data were collected on demographics, perceived extent of addiction to tobacco, smoking history, and self-reported smoking identity from questionnaires administered to 8225 students in British Columbia, Canada. A total of 7246 participants were categorised according to four smoking taxonomies established in the literature. Differences in perceived physical and mental addiction between smoking identity groups were calculated. The strength of the associations between the taxonomies of smoking and the smoking identity groups was also assessed.
Results: There were significant differences in perceived levels of physical (Kruskal–Wallis χ2 3985.02, p<0.001) and mental (Kruskal–Wallis χ2 4046.09, p<0.001) addiction to tobacco by the participants’ self-reported smoking identity. Youth smoking identities were modestly associated with the established smoking taxonomies (Pearson C contingency coefficient 0.64–0.72).
Conclusion: Self-reported smoking identities appear to provide valid characterisations of the smoking behaviour of youths that complement and elaborate existing taxonomies of smoking behaviour. Questions about self-reported smoking identity should be used in conjunction with smoking behaviour taxonomies when investigating youth smoking behaviours.
Statistics from Altmetric.com
Several factors have been observed to foster cigarette use by youth, including those that are intrapersonal (eg, sensation/novelty seeking1 2 and genetic susceptibility),3 interpersonal (eg, peers’ smoking),4–6 and environmental (eg, tobacco industry advertising).7 8 Although interactions between environmental, interpersonal, and intrapersonal factors may influence adolescents’ progression towards regular tobacco use,9 10 it is not clear how adolescents’ conception of their own smoking status, ie, their “smoking identity”, informs this process.
Smoking identities represent widely used psychosocial constructions of how adolescents who smoke perceive themselves in relation to their smoking behaviour (eg, social smoker, occasional smoker).11–13 Youths that find the image of adult smokers desirable generally have greater intentions to smoke in the future.14 It has been suggested that the image of smoking projected to some youths (ie, looking “cool”, “grown up”, or “hard” (tough)), along with the perceived empowerment acquired by engaging in the behaviour, may proffer definition to their precarious social identities as they make the transition to adulthood and thus foster the adoption of particular smoking identities.15
Although different categorisations, or taxonomies, have been used to describe the stages through which youths progress in their tobacco use,16–18 increasing evidence of heterogeneity within the conventional classifications of smokers19 suggests that such categorisations may require reassessment. Measures used in the study of youths’ smoking behaviour frequently rely on categorisations used in the study of adults’ smoking behaviour20 21 (for a detailed review of smoking categorisations see Mayhew et al).17 Recent studies have suggested, however, that these categorisations may not converge with youths’ perceived smoking identity.22–24 These inconsistencies indicate a need to further explore whether youths’ perceptions of their own smoking behaviour are better (ie, valid) measures of their smoking behaviour.
Classifications of youths’ smoking behaviour that more closely match their perceived smoking identities will provide a means to properly assess youth smoking19 23 and potentially improve tailored health messages and interventions to which youth can relate. As Leatherdale and McDonald24 argued, if interventions fail to align with youths’ perceptions of their smoking behaviour, the strategies and communications employed run the risk of being viewed as irrelevant by the targeted youth audience, and may lead to inappropriate intervention (eg, cessation activities may be inappropriately targeted at youth who self-define as non-smokers although they use tobacco).
Despite uncertainty about youths’ ability to accurately report their degree of tobacco “addiction” or “dependence”, emerging evidence indicates that youths report tobacco dependence symptoms following minimal experience with tobacco products,25 26 and that increasing dependence in an individual may be responsible for his or her escalating use of tobacco products.16 This has led some researchers to assess adolescents’ perceived mental and physical addiction to tobacco in attempting to quantify emerging tobacco dependence.27–29 Whether youths make a distinction in their self-ratings of their mental and physical addiction is yet to be determined.
The purpose of this study was to examine the validity of smoking identities as measures of smoking behaviour by testing for differences in self-rated addiction across commonly reported identities. In addition, we examined the relationship between smoking identities and several established taxonomies of smoking behaviour.
Participants and sampling
The data for this analysis were obtained via self-administered questionnaires from the British Columbia Youth Survey on Smoking and Health II (BCYSOSH-II).30 The BCYSOSH-II was a cross-sectional survey that relied on a questionnaire administered to 8225 students, in grades 7–12, in schools in British Columbia (BC), Canada. The selection of students with the 49 participating schools ranged from the entire student body to all students attending grade-specific required classes. The average response rate was 84, with student absenteeism accounting for the majority of the non-response. The questionnaire was administered, either by paper and pencil or online, in the students’ classrooms. Before completing the questionnaire, the students were given information sheets to deliver to their parents or guardians regarding the study, but active parental consent was not required. The questionnaires remained anonymous; names and other identifiers were not provided. The study was approved by the University of British Columbia, Behavioural Research Ethics Board.
Sociodemographic characteristics from the questionnaire included the participant’s gender, age and school grade. Smoking history was determined by asking the participants questions about their lifetime and recent (ie, past month) cigarette use. The participants that affirmed having smoked in the past month were further asked, “In the past month, approximately how many days did you smoke cigarettes?” (with response choices: “1 to 2 days”, “3 to 5 days”, “6 to 10 days”, “11 to 20 days”, “21 to 29 days”, and “everyday”). The participants were also asked for the age at which they initiated smoking, and the current non-smokers were further asked about their likelihood of smoking in the future.
Perceived “mental” and “physical” addiction levels were measured on scales of 0 (“not at all addicted”) to 10 (“very addicted”). These items were derived from the work of O’Loughlin et al.29 These items have been shown to have moderate to strong associations with other measures of youths’ nicotine dependence.28
Various smoking identities, derived from our previous qualitative research in which youths were asked to describe types of smokers,31 32 were presented to the participants. The participants were asked, “How would you define yourself?”, with nine response options offered: “I am a non-smoker”, “I am an occasional smoker”, “I am an irregular smoker”, “I am a regular smoker”, “I am a heavy smoker”, “I am a social smoker”, “I am an ex-smoker”, “I am a pot smoker”, and “other, specify”. The participants were instructed to “check all that apply”. For the purposes of this study, we examined the participants’ tobacco smoking identities (ie, we omitted the “I am a pot smoker” response).
Taxonomies of smoking were constructed from the items that assessed the participants’ smoking history, based on the taxonomies available in the published literature (table 1). Taxonomy 1 consisted of “current smokers”, “current experimenters”, “ex-experimenters”, “ex-smokers”, “puffers”, and “never smokers”. These operational definitions have been employed elsewhere previously.33–36
Taxonomy 2 groupings were constructed from Mayhew et al’s recommended categorisations.17 “Non-smokers” were classified as those that responded, “I have never had a puff of a cigarette” to the question, “About how many cigarettes have you smoked in your entire life?” Sub-categories were constructed for this class with “susceptible” non-smokers being individuals that responded “very likely” or “somewhat likely” to the question, “How likely is it that you will ever smoke in the future?” and “non-susceptible” non-smokers responding “rather unlikely” or “very unlikely” to the same question. To construct the category, “tried once smoker”, we selected individuals that responded, “I have only had a puff or a few puffs” to the question, “About how many cigarettes have you smoked in your entire life?” and then subtracted the age at which they took their first puff of a cigarette from their current age to determine whether they had smoked in the past 12 months (criterion proposed by Mayhew et al’s17 categorisation for “tried once smoker”). “Experimental smokers” were classified as the individuals that reported smoking 1–99 cigarettes in their entire lifetime and had not smoked in the past month (ie, did not smoke monthly based on Mayhew et al’s17 definition). “Regular smokers” were classified as the individuals that had smoked within the past month and had smoked >100 cigarettes in their lifetime but did not smoke daily (ie, had smoked from 1–20 days in the past month). “Established smokers” were defined as individuals that had smoked within the past month, had smoked >100 cigarettes in their lifetime, and had smoked daily or almost daily (ie, had smoked from 21–30 days in the past month).
The taxonomy 3 categories were constructed from the smoking history questions in keeping with the operational definitions employed by other researchers.37 It consisted of three groups: “non-smokers”, “former smokers”, and “current smokers”.
Taxonomy 4 was based on smoking in the past month, which is a commonly used categorisation of the smoking behaviour of youths and adults.20 21 37 38 The “non-smokers” were individuals that reported not smoking in the past month, whereas the “current smokers” were individuals that had smoked in the past month.
The data for this analysis were obtained from the participants that selected at least one smoking identity (fig 1). Of the 8225 students in the original sample, 1.1 had missing responses and 7.8 either defined themselves strictly as users of products other than cigarettes (ie, cigars, chewed tobacco, cocaine, marijuana) or gave responses that could not be categorised (eg, “I am a quitting smoker” and “quitting but with relapses”). Another 3.1 gave incongruent responses (eg, “I am an ex-smoker” combined with “I am a regular smoker”). As a result, the present analysis is based on data obtained from 7246 adolescents (88 of the original sample). Individuals whose data were not included in this analysis were on average of a younger age (14.4 years vs 15.3 years, p<0.001) and were more likely to be female (13.2 vs 10.7, p 0.001).
Our preliminary analyses of the participants’ smoking identities resulted in three groups (fig 1, shown in superscript letters): (a) individuals with a single smoking identity (n 6950; 95.9), (b) individuals with two identities selected (n 273; 3.8), and (c) individuals with three identities selected (n 23; 0.3). Because of the small sub-samples of individuals with multiple smoking identities, our analysis is focused on the individuals with a single smoking identity; a description of those with multiple identities is provided (table 2).
Descriptive statistics of the total sample including frequency and mean (SD) were calculated. We reported medians and means with 95 CI for the participants’ perceived mental and physical addiction by their self-reported smoking identities. Because the perceived levels of physical and mental addiction were non-normally distributed, we tested for differences between the smoking-identity groups using the Kruskal–Wallis one-way analysis of variance (ANOVA) nonparametric test. We further determined differences in perceived physical and mental addiction between smoking identities, post hoc, by using Mann–Whitney U tests and the Bonferroni correction for multiple comparisons by setting a significance level of alpha 0.0024 for 21 multiple comparisons (ie, α 0.05/21). Cross-tabulations were used to compare the self-reported smoking identities of the participants by the four smoking taxonomies and a contingency coefficient (Pearson C) was used to measure and compare the strength of association between the smoking taxonomies and the self-reported smoking identities of the youths. The Pearson C is recommended for contingency tables with 5-by-5 categories or larger. All analyses were performed using SPSS V.14 (SPSS, Chicago, Illinois, USA).
The students were evenly distributed by gender (51.6 were female) with a mean (SD) age of 15.2 (1.6) years. Table 2 describes the distribution of the participants by their self-reported single and multiple smoking identities.
Perceived level of physical and mental addiction by smoking identity
There were significant differences in the participants’ perceived levels of physical (Kruskal–Wallis χ2 3985.02, p<0.001) and mental (Kruskal–Wallis χ2 4046.09, p<0.001) addiction to tobacco by their selected smoking identities (fig 2). For level of perceived physical addiction, individuals identifying themselves as non-smokers (median 0.00, mean 0.08) had significantly lower levels of perceived physical addiction compared with all the other smoking identities. Individuals identifying themselves as ex-smokers (median 0.00, mean 1.32) had significantly lower levels of physical addiction compared with those identifying themselves as social, irregular, regular, and heavy smokers, but not occasional smokers. Individuals identifying themselves as occasional smokers (median 1.00, mean 2.04) had significantly lower levels of physical addiction than those identifying themselves as social, irregular, regular, and heavy smokers. Individuals identifying themselves as social smokers (median 2.00, mean 3.02) had significantly lower levels of physical addiction than those identifying themselves as regular and heavy smokers, but not individuals identifying themselves as irregular smokers (p 0.08). Individuals identifying themselves as irregular smokers (median 4.00, mean 3.70) had significantly lower levels of physical addiction than those identifying themselves as regular and heavy smokers. There was no significant difference in perceived physical addiction between individuals identifying themselves as regular (median 7.00, mean 6.06) and heavy smokers (median 6.00, mean 5.45). Similar results were obtained for differences in the perceived level of mental addiction between the groups, with the exception that the difference between perceived mental addiction between individuals identifying themselves as ex-smokers and occasional smokers was also significant (p<0.001).
Smoking identities and smoking taxonomies
Table 3 describes the distribution of participants that selected a single smoking status by the four smoking taxonomies. Within taxonomy 1, of the individuals identifying themselves as non-smokers, 72.7 fell within the “never smoker” category and 17.7 fell within the “puffer” category. Among participants that identified themselves as ex-smokers, 54.9 fell within the “ex-experimenter” category, whereas 20.7 were in the “ex-smoker” category. Of the participants that identified themselves as irregular, occasional, or social smokers, 33.9, 47.0, and 38.7, respectively, fell within the “current experimenter” category, and 42.7, 18.3, and 36.3, respectively, were in the “current smoker” category. Among participants identifying themselves as regular and heavy smokers, 19.9 and 29.4, respectively, fell within the “current experimenter” category and 76.0 and 60.3, respectively, fell within the “current smoker” category.
Within taxonomy 2, of the participants that identified themselves as non-smokers, 71.1 fell within the “never smoker” category (60.6 were non-susceptible non-smokers and 10.5 were susceptible non-smokers) and 15.3 fell into the “tried once” category; however, about 5.0 could not be classified (see Discussion for details). For the participants that identified themselves as ex-smokers, 54.9 fell into the “experimental smoker” category and 10.3 fell within the “regular smoker” category, but 29.3 could not be classified. Of the participants that identified themselves as irregular, occasional, and social smokers, 10.4, 12.4, and 13.7, respectively, fell within the “regular smoker” category, and 34.7, 6.8, and 21.8 fell within the “established smoker” category (32.2 of the participants identifying themselves as irregular smokers, 45.0 of those identifying themselves as occasional smokers, and 37.9 of those identifying themselves as social smokers could not be classified). Of the participants that identified themselves as regular smokers, 13.2 fell within the “regular smoker” category and 66.4 fell within the “established smoker” category, although 16.4 could not be classified. Finally, for the participants that identified themselves as heavy smokers, 11.1 fell within the “regular smoker” category and 50.8 fell within the “established smoker” category, but 27.0 could not be grouped into any of the taxonomy 2 categories.
Within taxonomy 3, the participants that identified themselves as non-smokers mostly fell within the “non-smoker” (90.4) and “former smoker” (8.3) categories; and those that classified themselves as ex-smokers fell mostly within the “former smoker” (75.5) and the “current smoker” (20.1) categories. Of the participants that identified themselves as irregular, occasional, and social smokers, 76.6, 65.4, and 75.0, respectively, fell within the “current smoker” category. Finally, of the participants that identified as regular smokers, 95.9 fell within the “current smoker” category and of the participants that identified themselves as heavy smokers, 89.7 fell within the “current smoker” category.
Within taxonomy 4, almost all of the participants that identified themselves as non-smokers fell within the “non-smoker” category (98.7); the participants that classified themselves as ex-smokers fell mostly within the “non-smoker” category (79.9); the participants that identified themselves as irregular, occasional, and social smokers, fell mostly within the “current smoker” category (76.6, 65.4, and 75.0, respectively); and the participants that identified themselves as regular and heavy smokers fell mostly within the “current smoker” category (95.9 and 89.7, respectively).
The strength of association between the taxonomy and the self-reported smoking identities was strongest for taxonomy 1 (Pearson C 0.72) and weakest for taxonomy 4 (Pearson C 0.64). There was only a difference of 0.01 in the strengths of association between taxonomy 1 and taxonomy 2.
The perceived addiction to tobacco by youths is an indicator of loss of autonomy over their smoking behaviour and of nicotine dependence early in the smoking initiation process.27 Although nicotine dependence symptoms are known to hinder cessation among youth,39 few studies have examined the relationship between the perceived addiction to tobacco and smoking identity in adolescents. The participants in our study significantly differed in their perceived levels of mental and physical addiction to tobacco by their smoking identities. These results provide evidence to support the validity of self-reported smoking identities as informative characterisations of youth smoking behaviours. This is consistent with other work that has found that ratings of perceived physical and mental addiction to tobacco are highly correlated with the emotional dependence (perceived physical addiction r 0.63; perceived mental addiction r 0.61) and nicotine dependence (perceived physical addiction r 0.69; perceived mental addiction r 0.66) dimensions of tobacco dependence, measured by the Dimensions of Tobacco Dependence Scale (DTDS).28 However, more research is required to determine whether youths can meaningfully differentiate between the various facets of their tobacco dependence (ie, the physical and psychological aspects of their emerging dependence).
The results of this study suggests that the shift from non-smoking identities to non-regular (ie, occasional, social, and irregular smokers) smoking identities marks an increasing vulnerability to tobacco dependence and decreasing autonomy over their tobacco use,26 placing such youth at increased risk of transitioning to regular use.16 Because these non-regular smoking identities corresponded with higher ratings of perceived physical and mental addiction to tobacco relative to non-smoker and ex-smoker identities (although lower than regular and heavy smoking identities), it is important that those designing tobacco control interventions for youth determine how emotional factors and underlying vulnerability to nicotine addiction may inform the smoking categories with which youth self-identify. This finding is strengthened by qualitative work suggesting that, during early experimentation with tobacco use, youths experience different dimensions of tobacco addiction (which include “psychological” and “physical” aspects) in relation to their need to use tobacco.32 Thus, in tailoring interventions for youth, it may be prudent to not only assess the smoking behaviour that adolescents exhibit (ie, “whether they smoke” or “the number of cigarettes smoked each day or month”), but to include questions regarding reasons why they believe they need to use tobacco.28
The inclusion of questions about youths’ specific attributions of their tobacco use may enhance our ability to identify and to develop tailored interventions for particular sub-groups of adolescent smokers40 who engage in tobacco use behaviour for different reasons (eg, smoking for social reasons, as a means of stress-reduction, or because they are unable to stop). Assessment questions (for eg, the DTDS) should determine the reasons attributed by adolescents for their tobacco use (eg, emotional, social, sensory, or nicotine dependence) and then interventions should specifically address those reasons. The effectiveness of the interventions can be tested by having adolescents successfully identify situations that may increase their desire to smoke, and also to determine which interventions (for eg, cognitive behavioural therapy, pharmacologic, or a combination of both) are effective to reduce their particular need to smoke. Such tailored interventions may address important barriers (see Mermelstein’s41 discussion about the challenges of intervening with adolescent smokers) that currently exist in smoking cessation interventions for adolescents.
The results of our analysis further suggest that although self-reported smoking identities have modest to strong associations with the smoking taxonomies frequently used by researchers (Pearson C 0.64–0.72), some of these classifications may not fully concur with or incorporate youths’ smoking identities. Although 98.7 of those classified as “non-smokers” based on taxonomy 4 also self-identified as non-smokers, and 95.9 of those classified as “current smokers” identified themselves as regular smokers, 20.1 of those that classified themselves as ex-smokers were categorised as “current smokers” by the taxonomy; 23.6, 34.6, and 25.0 of the participants that described themselves as irregular, occasional, and social smokers, respectively, were categorised as “non-smokers”, and 10.3 of the self-identified heavy smokers were categorised as “non-smokers”. These discrepancies reveal a significant potential for misclassification of some youths based on “the past 30 days of cigarette use” criterion, an approach widely used by researchers.20 21 37 38 Although few researchers have used more than one classification schema to categorise adolescents’ smoking behaviour, similar inconsistencies have been observed when comparing smoking in the past month with other classifications of smoking status (eg, smoking in the past month vs ever smoking).42 Such inconsistency can have important consequences when interpreting research outcomes.
Wellman et al16 demonstrated that although smoking acquisition among youth is generally unidirectional, it is discontinuous, from intermittent use to more regular use, with periods of no smoking. In fact, in a comparison of the taxonomy recommended by Mayhew et al17 with youths’ perceptions of their smoking status, Leatherdale and McDonald24 found that the relationship between the taxonomy and youths’ perceptions were inconsistent, primarily among adolescents that self-identified as non-established smokers. In our analysis, more than one half of the youths that identified themselves as ex-smokers fell within the categorisation of “ex-experimental smoker” of taxonomy 1 and “experimental smoker” of taxonomy 2. This finding suggests that youth that identify themselves as ex-smokers may not have attained established smoking before making a decision to “quit”. Consequently, adolescents may not unambiguously fall within or identify with conventional intermittent or experimental smoking behaviour labels.43
It is important to note that many of the participants in our study selected multiple identities to describe their smoking status. Although we did not perform further analyses beyond obtaining frequencies of these multiple identities, due to small sub-sample sizes (table 2), these multiple identity responses provide a broader perspective of how adolescents may conceptualise their smoking identity. This is the first study, known to date, to show that adolescent smokers may assume multiple smoking identities. Notably, the multiple response selection seemed to capture youths in transition between non-smoking to irregular use and between irregular use towards more established tobacco use. In fact, of the 15 unique multiple identities selected by the participants (ie, 12 dual identities and 3 triple identities), 8 included the “social smoker” identity in addition to other smoking identities. A study of US college students44 found that social smokers (ie, individuals who smoked mainly with others rather than alone or equally by themselves) were more likely to be recent onset smokers, rather than having smoked for a long duration and exhibited less smoking frequency and intensity, less nicotine dependence, and less cessation intention and attempts. Thus, the selection of a “social smoker” identity, in combination with another identity, may be indicative of a transitional period between smoking stages. However, due to our limited data, such speculations are beyond the scope of this present study and may be considered for future explorations of adolescents’ smoking identities.
Several limitations should be considered when interpreting the results of this study. First, because the data were based on self-reported smoking behaviour, the validity of the responses may be in question. That is, the self reports may be informative, particularly in relation to how youths might respond to targeted messages, but may not be the most apt description of the youths’ smoking status. Second, in selecting the participants to include in this analysis, we did not include the participants that gave incongruent responses (eg, the participants that selected “I am an ex-smoker” and “I am regular smoker;” “I am an ex-smoker” and “I am heavy smoker;” “I am a heavy smoker” and “I am a non-smoker”). It is possible that these seemingly incongruent responses are valid smoking identities; however only 3.1 of the original sample was excluded because of such ambiguous responses. Third, because this analysis was based on cross-sectional data, causal relationships cannot be inferred and several factors (ie, culture, ethnicity, other substance use, and exposure to media messages about tobacco) may affect youths’ self-identities with respect to their tobacco use.
What this paper adds
Recent studies suggest that the conceptualisations of youth smoking identities may be inconsistent (to varying degrees) with classifications or taxonomies commonly used in research to operationally define their smoking behaviour. However, not much is known about the validity of using these smoking identities as measures of smoking behaviour, and how these smoking identities correspond with perceptions of their addiction status by youths.
Our study demonstrates that adolescents’ self-reported smoking identities may be important and valid measures that can be used to improve our understanding of their smoking behaviour. They are moderately associated with taxonomies commonly used by researchers to operationally define smoking behaviour and characterise different types of smokers based on their perceived addiction to tobacco.
In categorising the participants according to the taxonomy recommended by Mayhew et al17 (ie, taxonomy 2), 9.4 of the cases could not be properly classified. This was due, in part, to the fact that some individuals simply could not be defined by any category in taxonomy 2 (eg, participants that had smoked >100 cigarettes in their lifetime and not in the past month (n 304) could not be classified because they potentially fell into an intermediary category between the experimental and regular smoker categories. Additionally, some individuals were missing information in other variables used to reconstruct Mayhew et al’s17 categories, such as age of first smoking initiation (missing n 36), current age (missing n 17), and responses to the question, “How likely is it that you will ever smoke in the future?” among non-smokers (missing n 102). The intention of the categorisation developed by Mayhew et al17 was to identify stages through which adolescents progress in the development of their smoking behaviour. Thus, their taxonomy may be recommended for longitudinal analyses describing adolescents’ tobacco use as opposed to cross-sectional investigations. However, we were able to categorise >90 of our participants using this taxonomy.
In conclusion, with the potential discrepancies in the ways by which researchers define youths’ smoking and youths’ perceptions of their smoking identity, interventions to prevent initiation and even to promote cessation could potentially be mismatched with youths’ identities resulting in a reduction in efficacy. It is recommended, therefore, that whatever categorisations of youths’ smoking behaviour are used, researchers should also incorporate questions regarding self-reported smoking identity. It is further recommended that research designed to assess youths’ smoking behaviour be evaluated based on how well the participants’ smoking identities correspond with the categorisations employed. This will determine how well the conceptual definitions of smoking behaviour employed by researchers match youths’ conceptualisations. Future studies are needed to identify the factors that affect adolescents’ smoking identity, including the role of tobacco use, psychosocial issues, and tobacco industry advertising.
Funding: This research was supported by a Canadian Institutes of Health Research (CIHR) operating grant (grant #62980). CTCO was supported by a CIHR Strategic Training Program in Tobacco Research (STPTR) Post-doctoral Fellowship and a Michael Smith Foundation for Health Research (MSFHR) NEXUS Research Unit Traineeship. CGR was supported by CIHR and NEXUS Post-doctoral Fellowships. JLJ was supported by a CIHR Investigator Award. PAR is a Senior Scholar funded by the Michael Smith Foundation for Health Research.
Competing interests: None.
Ethics approval: This study was approved by the University of British Columbia Behavioural Research Ethics Board.
If you wish to reuse any or all of this article please use the link below which will take you to the Copyright Clearance Center’s RightsLink service. You will be able to get a quick price and instant permission to reuse the content in many different ways.