Health Education Research Advance Access originally published online on February 13, 2007
Health Education Research 2008 23(1):81-93; doi:10.1093/her/cym006
Factors related to adolescents' estimation of peer smoking prevalence
1 Department of Health Studies and Gerontology
2 Centre for Behavioural Research and Program Evaluation, University of Waterloo, 200 University Avenue West, Waterloo, Ontario, Canada N2L 3G1
3 Division of Preventive Oncology, Cancer Care Ontario, 620 University Avenue, Toronto, Ontario, Canada M5G 2L7
* Correspondence to: J. L. Reid. E-mail: jl3reid{at}uwaterloo.ca
 |
Abstract |
|---|
 Top Abstract IntroductionMethodsResultsDiscussionConflict of interest statementAcknowledgementsReferences |
|---|
Although adolescents who overestimate peer smoking prevalence are more likely to smoke, little research has focused on the factors associated with why the majority of adolescents overestimate peer smoking rate. The purpose of this study was to examine demographic, social, environmental and behavioural characteristics related to overestimation of peer smoking prevalence among secondary school students. The current study analysed data collected in two Canadian studies that used the Tobacco Module of the School Health Action, Planning and Evaluation System, a school-based questionnaire. One study surveyed 23 458 students (Grades 9–13) in 29 schools during 2001–02, and the other surveyed 25 452 students in 39 schools in 2003. Results of multiple logistic regression indicate that grade, gender, close friends’ smoking, seeing smoking at school, family members’ smoking, smoking in the home and smoking status have a clear association with overestimation; school smoking rate and susceptibility to smoking show a tentative relationship and warrant further study. Other factors may also be important for prevalence estimation, and further research is needed to identify these factors. Since adolescents tend to overestimate peer smoking prevalence and perceived prevalence is in turn linked to smoking behaviour, interventions should focus on creating realistic perceptions of smoking prevalence.
|
Introduction |
|---|
|
TopAbstractIntroductionMethodsResultsDiscussionConflict of interest statementAcknowledgementsReferences |
|---|
Smoking is the leading cause of preventable death in Canada [1], yet 19% of Canadian youth aged 15–19 smoke [2]. Since adolescence is a critical time for smoking initiation (85% of adult smokers started smoking in adolescence [3]), research related to youth smoking uptake is essential. Theoretical evidence suggests that environmental influences, particularly social influences, are related to youth smoking behaviour [4–6]. For instance, the Theory of Planned Behavior posits that behavioural intent (which determines behaviour) is influenced by attitudes, subjective norms and perceived control regarding the behaviour; application of this theory suggests that youths’ subjective norms about peer smoking would influence their behavioural intent and subsequent likelihood of smoking [4]. Although empirical research has demonstrated a relationship between youth smoking behaviour and perceptions of peer smoking prevalence [7–13], little research has examined what influences adolescents’ perceptions of peer smoking prevalence.
The research examining global perceptions has identified that adolescent populations tend to overestimate the prevalence of smoking [14–18]. This is cause for concern considering that beliefs about social norms for smoking are related to youth smoking behaviour [7–13], and perceived norms are actually more influential than actual smoking rates for predicting youth smoking [9–11]. However, if exaggerated perceptions of peer smoking prevalence can be corrected with appropriate interventions, it may be possible to influence smoking behaviour.
In order to develop such interventions, we need to identify factors related to youth perceptions of smoking prevalence; however, little research has focussed on this area. Our literature review identified only two empirical studies specifically examining the factors influencing youth prevalence estimates [17, 18]. Sussman et al. [17] found that an individual's own smoking status affected their prevalence estimates, with regular smokers giving the most inflated estimates and non-smokers providing the least inflated estimates. Students who made greater prevalence estimates also had more significant others (friends, family) who smoked, and were more likely to smoke in the next year. More recently, Unger and Rohrbach [18] identified that prevalence estimates among eighth-grade students are influenced by student characteristics, including friends’ smoking, female gender, actual school smoking prevalence, perceptions of smoking on television, perceived access to cigarettes, low academic performance, cigarette offers and ethnicity. Contrary to the findings of Sussman et al. [17] and Unger and Rohrbach [18] found that the respondents’ own smoking behaviour and susceptibility to smoking were not related to prevalence estimates when controlling for social influences (e.g. close friends). Additional research is required to better understand these relationships.
The present study builds on the existing research by examining additional social and environmental influences and by using a more nuanced outcome measure for perceived prevalence (i.e. accuracy of estimation is calculated on an individual basis by comparing it to the within-grade, within-school smoking rate, rather than using the mean/median perceived prevalence for a group). The purpose of this study was to examine how student characteristics (demographic and behavioural) and the social and environmental influences surrounding youth were related to their overestimation of peer smoking prevalence.
|
Methods |
|---|
|
TopAbstractIntroductionMethodsResultsDiscussionConflict of interest statementAcknowledgementsReferences |
|---|
Design
This secondary data analysis used cross-sectional data from two different host studies that used the Tobacco Module of the School Health Action, Planning and Evaluation System (SHAPES) [19]. The Tobacco Module is a previously validated and reliable [20] four-page (
20-min) questionnaire designed to measure youth smoking behaviour, attitudes about smoking and contextual influences on tobacco use (refer to www.shapes.uwaterloo.ca for additional details). The primary data set (Study A) was used to develop the predictive model, and a second data set (Study B) was used to validate the model. In Study A, data were collected from 23 458 students (Grades 9–13) in 29 secondary schools in the province of Ontario, Canada during the 2001–02 academic year. Only three of the 17 school boards approached declined participation, and at the school level, 88% of secondary school students participated.
These data were collected at school during class time, supervised by teachers and research staff. Active information-passive parental consent with active student assent was used to reduce demands on schools and to increase student participation rates. Information letters were mailed to parents, who could call a toll-free number (accessible 24 h a day) to withdraw permission for their son/daughter. The University of Waterloo Office of Research Ethics and appropriate school board and public health ethics committees approved all procedures, including passive consent.
In Study B, the same questionnaire and procedures were used to collect data from 25 452 students (Grades 9–13) in 39 secondary schools in one public health district of Ontario, Canada. In accordance with the needs of the health district, Study B data collection occurred over two waves in 2003 (27 schools in Spring, 12 schools in Fall). Two-thirds of the schools in the selected area were approached and all participated. At the school level, 81% of students participated.
Measures
Demographic variables included grade and gender; social models reflected the influence of people closest to the adolescent, including family members and close friends who smoke; environmental influences considered smoking in the home, smoking in the workplace, seeing smoking at school and actual school smoking rate; behavioural characteristics included smoking status and susceptibility to smoking (Fig. 1).
|
Outcome (dependent) variables
Two outcome variables were calculated using individual responses to the question ‘How many people your age, in your school, do you think smoke cigarettes?’ (0–10%, 11–20%, 21–30%, 31–40%, 41–50%, 5–60%, 61–70%, 71–80%, 81–90%, 91–100%). First, a simple estimation score was created by comparing a student's response with the actual within-grade, within-school smoking rate based on the number of increments (10-percentage-point ranges) the student's response was above or below the correct range: accurate (0) if in the correct range, a category of overestimation (+1 for one range above to +9 for 9 ranges above) if the response was above the correct range or a category of underestimation (-1 for one range below to -9 for 9 ranges below) if the response was below the actual rate. Second, in order to better characterize youth who clearly overestimated smoking prevalence (i.e. so we know where to best target interventions), the simple estimation scores were grouped into two categories to create the dichotomous estimation variable [respondents who clearly overestimated (+2 or higher on the estimation score) = 1, respondents who did not clearly overestimate (0 or lower on the estimation score) = 0 and estimation scores of +1 were excluded so students who marginally overestimated would not be misclassified].
Explanatory (independent) variables
Individual level variables.
Individual level variables included demographics, social models, environmental influences and behavioural characteristics. Demographic variables included grade (9–13) and gender (female = 1, male = 0). Social models included a measure of family smoking [number of family members (parents or older siblings) who smoke (0–4)] and a measure of friends smoking [how many of the respondent's five closest friends smoke (0–5)]. Environmental influences included a measure of the number of people smoking in the home everyday or almost everyday (0–4+), exposure to smoking in the workplace (yes = 1, no = 0) and seeing others smoking at school [as in previous research [21], measured by responses to two items: (i) ‘I often see others smoking near this school’ and (ii) ‘How many students at this school smoke where they are not allowed to?’].
Behavioural characteristics include smoking status and susceptibility. Consistent with existing research [22–24], student smoking status was defined as never smoker (has never smoked a cigarette, not even a puff), tried smoking (has tried smoking but not again in the last 30 days), experimental smoker (smoked more than once in the last 30 days but does not smoke everyday or almost everyday), regular smoker (has smoked everyday or almost everyday in the last 30 days) and former smoker (has smoked >100 cigarettes in the past but has not smoked in the last 2 weeks and considers his/herself as having quit). Consistent with the validated algorithm of Pierce et al. [25], susceptibility to smoking among never smokers was determined by responses to three questions: (i) ‘Do you think in the future you might try smoking cigarettes?’ (ii) ‘If one of your best friends were to offer you a cigarette, would you smoke it?’ and (iii) ‘At any time during the next year do you think you will smoke a cigarette?’. Students responded to these questions on a four-point Likert scale. Those who selected ‘definitely not’ for all three questions were considered not susceptible (0); all others were considered susceptible (1).
Organizational level variables.
Organizational level variables included school smoking rate and within-grade smoking rate. School smoking rate (across grades) was calculated as the total number of smokers (sum of experimental and regular) as a proportion of all students classified for smoking status, then grouped into categories representing schools with low (below the first quartile), moderate (between the first and third quartiles) or high (above the third quartile) smoking rates. Within-grade smoking rate was calculated as the total number of smokers (sum of experimental and regular) in a specific grade at a specific school as a proportion of all students classified for smoking status in the same grade at the same school. The 10-percentage-point range (corresponding to the response options for the estimation question) that included the actual within-grade smoking rate was assigned to each student based on their grade and school, and was used for comparison to calculate the estimation score.
Analyses
The analyses for this study occurred in two steps. In Step 1, data from Study A were used to develop the predictive model. Chi-square tests for categorical variables and Spearman rank order correlations for ordinal and continuous variables were calculated. Only variables with a significant chi-square or correlation with the response variable of 0.10 or higher were retained. These explanatory variables were then included in a logistic regression model to determine their relationship with the dichotomous estimation variable (students who clearly overestimated versus students who did not clearly overestimate). In Step 2, we used the data from Study B and the same modelling procedures as Step 1 to validate our initial model. All analyses were conducted using SAS 8.2 [26].
All eligible respondents were included in the samples; however, those with missing or inappropriate grade level (n = 237 for Study A, n = 88 for Study B) were excluded from further analyses, and some individuals were excluded from specific analyses on a case-wise basis as a result of missing values on the items used. Since the logistic regression procedure required a complete set of variables for each respondent, the sample size for this step was reduced to 62% (61%) of the total sample for Study A (Study B). Given the substantial sample size, we are still confident that these data fairly represent the general student populations surveyed.
|
Results |
|---|
|
TopAbstractIntroductionMethodsResultsDiscussionConflict of interest statementAcknowledgementsReferences |
|---|
Demographic characteristics of Studies A and B are presented in Table I. The gender distribution was similar across both studies (49.8% male in Study A, 50.3% male in Study B). In each study, the mean age of students was 15.8. Table I also shows the characteristics of students retained for the logistic regression analyses compared with those who were excluded.
|
Estimation of peer smoking
The distribution of peer smoking estimates is presented in Table II. In Study A, 81.1% of students overestimated, 10.3% were accurate and 8.6% underestimated the within-grade smoking rate at their school. When simple estimation scores are grouped into two categories of estimation [and those who barely overestimated (+1) excluded], 78.1% of those included clearly overestimated and 21.9% were accurate or underestimated. Study B data showed a similar distribution, but students were slightly more ‘accurate’.
|
Correlations
The correlations between the independent variables (grade, school smoking rate, family members' smoking, friends' smoking, smoking in the home, seeing smoking at school, gender, smoking in the workplace, smoking status and susceptibility) and the dichotomous estimation variable are presented in Table III. Small but significant effect sizes were found. In Study A, Spearman correlations for ordinal variables ranged from 0.10 to 0.20, and all were significant (P < 0.001). Spearman correlations for Study B ranged from 0.03 to 0.23, and all were significant (P < 0.001). For categorical variables in both data sets, all chi-square tests were significant (P < 0.001).
|
Logistic regression
The results of the logistic regression analyses are presented in Table IV. The logistic regression model explained 74.3% (c = 0.743) of the variance in the outcome variable, estimation of smoking prevalence. Students were more likely to overestimate if they were younger (Grade 9–11), were female, had more friends who smoke, saw smoking at school, had family members who smoke, went to a school with a low smoking rate or were exposed to select levels of smoking in the home. The students' own smoking status was significantly associated only at certain levels: triers and former smokers were more likely to overestimate, while regular smokers were less likely.
|
When validating the multiple logistic regression model using Study B data, smoking in the workplace and school smoking rate were not included in the regression, due to an omitted survey item and low correlation, respectively. Odds ratios and significance levels for Study B were similar to Study A for most variables, including grade, gender, family members' smoking and friends' smoking. Smoking in the home and seeing smoking at school showed a stronger relationship in Study B. Susceptibility, while not included in the Study A regression due to non-significance, reached significance in the model for Study B, with susceptible students slightly more likely to overestimate. Smoking status showed a weaker relationship in Study B, and was significant only for regular smokers, who were less likely to overestimate.
|
Discussion |
|---|
|
TopAbstractIntroductionMethodsResultsDiscussionConflict of interest statementAcknowledgementsReferences |
|---|
Consistent with other studies documenting the tendency of adolescents to overestimate the prevalence of peer smoking [14–16], we identified that the majority of students (over three-quarters) overestimated the within-grade smoking rate at their school. This is cause for concern as the perceived prevalence of peer smoking is associated with the smoking behaviour of adolescents even more so than actual prevalence [9, 10]. Since behavioural theory posits that altering this exaggerated normative perception of peer smoking prevalence may in turn lead to a reduced influence on smoking behaviour [4], interventions should focus on creating realistic perceptions of the prevalence of smoking. Interventions that include a specific, directed effort to correct misperceptions may be effective in reducing smoking prevalence. This approach has been effective in interventions for alcohol abuse prevention [27] and has been included in some ‘social influences’ approaches to smoking prevention, where it has been viewed as necessary but not sufficient on its own as a smoking prevention tool [22]. More research is needed to determine the individual effects of this component and its utility in smoking prevention.
The finding that most students overestimate peer smoking prevalence may be partly due to ambiguity in the question, since it simply reads ‘How many people your age, in your school, do you think smoke cigarettes?’ without qualifying the meaning of ‘smoke’ (with frequency, for example), so students make their own assumptions about what constitutes a smoker. Sussman et al. [17] obtained prevalence estimates by asking students to estimate the number of their peers who have tried smoking and who smoke weekly, and found that estimates for trying smoking were fairly accurate, while weekly smoking was highly overestimated. Perhaps some students in the current study are assuming a definition of smoking that reflects behaviour of trying smoking rather than more regular smoking, which may help to explain inflated estimates of smoking—these estimates may be accurate for trying behaviour of their peers, not for regular smoking. Overall, students seem to have the idea that ‘everyone is doing it’, as evidenced by their prevalence estimates.
Correlations between the dependent variable (estimation) and the independent variables were low in both samples. However, when combined in the model, they explained 74% of the variance observed in the overestimation variable. Both the multiple logistic regression model and subsequent validation revealed that grade, close friends’ smoking, gender, seeing smoking at school, family members’ smoking, smoking in the home and smoking status were significantly associated with overestimation. School smoking rate and susceptibility were significant in only one model, showing a more tentative association. Smoking in the workplace was not significantly associated in the Study A model and was not included in the Study B data set.
Demographics such as grade and gender had strong associations with overestimation. Students in lower grades and females were more likely to overestimate, a finding consistent with the literature [7, 14, 18]. This insight could be used to target interventions to youth populations that are more likely to perceive an inflated prevalence of peer smoking (i.e. younger students and females); evaluation of this type of targeted approach would be required.
Social models, especially close friends, were also important for overestimation. Students with a greater number of close friends who smoked were increasingly more likely to overestimate (compared with those with no close friends who smoked). This is consistent with the observation of Sussman et al. [17] that students reporting more friends who smoked made higher prevalence estimates and Unger and Rohrbach's [18] finding that best friends’ smoking accounted for the largest proportion of variance in peer smoking prevalence estimates. Having more family members who smoke was also associated with an increased likelihood of overestimation compared with having no family members who smoked. Family members may be particularly influential in this model given that only parents and older siblings, both of whom may serve as role models, were included. Theoretical evidence [4–6] supports the influence of social models on normative beliefs about smoking; the more important others (i.e. friends and family) one has who smoke, the more apt one is to believe that smoking is a normative behaviour.
Some research has demonstrated that students attending schools with higher smoking rates make higher prevalence estimates, likely due to seeing or hearing about more smokers at school [15]. However, others have found that students on campuses with lower tobacco use rates made more exaggerated prevalence estimates [16]. We found that a higher actual school smoking rate was not related to overestimation; in fact, students in schools with a low or moderate smoking rate were more than twice as likely to overestimate peer smoking when compared with schools that had a high smoking rate. These findings may be due to a ceiling effect, where the range of the response options limits the degree to which a student can overestimate (creates a ceiling), regardless of whether they may have estimated a higher prevalence rate due to higher actual smoking in the school. Conversely, adolescents may tend to make high estimates regardless of actual rates, with the margin of overestimation (the difference between the two) then determined more by the actual rate than the estimation score. Research is required to test these assumptions.
One way that students may form these prevalence estimates is through observing the smoking behaviour of their classmates. Students in both samples who reported seeing ‘lots’ of smoking at school were more than twice as likely to overestimate smoking prevalence, and seeing ‘some’ was significant in the Study B sample, but seeing ‘few’ students smoking was not significant. It appears that in the school environment, seeing smoking is more important for forming prevalence estimates than the actual proportion of smokers. As the actual prevalence of smokers decreases (as in Study B), smokers may become more visible and may have a greater effect on estimation [5]. A previous study found that noticing teens smoking was associated with smoking behaviour [28], but no studies have examined the effect of seeing smoking on estimation.
Environmental influences outside of school were not as strong. Smoking in the home showed associations only at some levels in Study A, but at all levels in Study B. The influence of smoking in the home on prevalence estimates may vary depending on who those smokers are or the number of family members smoking. Interventions targeting restrictions on smoking in the home may be somewhat helpful, but interventions targeting parents and families may have a greater impact given the importance of social influences. The environmental influence of smoking in the workplace was not significant and not included on the survey version used in Study B. There was little variation in responses to this item, since only about half of students had jobs, students typically do not spend many hours at part-time jobs and many workplaces have smoking bans.
Behavioural factors also showed weaker associations with estimation of smoking. When compared with never smokers, those who had tried smoking or were former smokers were somewhat more likely to overestimate (in Study A only), while regular smokers were actually less likely to overestimate, and no significant effect was found for experimental smokers. In Study B, the only significant association was for regular smokers, who were less likely to overestimate. This contradicts findings by Sussman et al. [17] that regular smokers are more likely to overestimate due to a false consensus effect; however, possibly confounding peer influence variables were not controlled for. Susceptibility to smoking was not significant in the Study A regression analysis, which agrees with a previous study that used comparable questions to determine susceptibility, and found that although susceptible students gave high estimates, susceptibility was not significant in a regression model [18]. In Study B, however, susceptibility was significantly associated, but students susceptible to smoking were only slightly more likely to overestimate. The lower smoking rate of Study B may explain the reduced importance of smoking status and increased importance of susceptibility.
Limitations and strengths
The main limitation of this study is its cross-sectional design. Relationships are therefore correlational and causality cannot be inferred. Further research is needed to determine how prevalence estimates are formed and the factors that contribute to these estimates. Longitudinal studies to assess the variables associated with prevalence estimates over time may be informative in this regard.
While the use of convenience sampling may limit the representativeness of the sample and generalizability of the results, for this study, the relationships between variables are more important than representativeness. However, school response rates were high, and these data are representative of the general student population in Canada [24]. The overall smoking rate of the Study A sample was 31.3%, compared with the Canadian Tobacco Use Monitoring Survey (CTUMS) national sample reported smoking rate of 22.5% for those aged 15–19 for the same year [29]. The Study B sample had a smoking rate of 20.7%, compared with the CTUMS national rate of 18% for those aged 15–19 for the same year [30]. Although the smoking rates as determined by SHAPES are somewhat higher than the CTUMS figures, the survey methods are different (telephone versus pencil and paper), the definitions for determination of smoking status differ and participation rates were higher in our sample (non-participants are more likely to smoke [31]). In addition, the SHAPES samples each only include one area of Ontario, whereas CTUMS is national, so each may be approximately representative of the population sampled.
The difference in smoking rate between the two samples may have several reasons, such as the timing of data collection, the population studied and environmental factors. The data collection for Study A took place in 2001–02 and for Study B in 2003; the national youth smoking rate dropped during this time. Also, a number of French schools were included in Study B, and French-speaking persons outside of Quebec have reported lower smoking rates than their English-speaking counterparts [32]. Lastly, the municipality examined in Study B introduced a no-smoking bylaw shortly before data collection; this reduces visibility of smoking in public places, and may have affected smoking rates or perceptions in the area. Other population differences may also exist.
Several factors contributed to the loss of respondents from the total sample to the final analysis. The logistic regression procedure decreased the sample size included in this analysis to less than two-thirds of the total. However, a similar study retained 73% of the sample for regression analysis due to missing variables [18], so this reduction in sample size is not unique. Although some differences exist between the students retained and excluded, they are not large enough to be meaningful. The classification of smokers resulted in the loss of >3000 respondents from each sample, although the definitions used are from a classification system that reduced the number of respondents lost in this process [22–24]. However, even after losing respondents with missing variables, the sample size was very large, strengthening the results. The sample size in Study B was slightly larger, giving it more power—this may account for greater significance found for several variables.
One should consider that the data are based on the self-report of adolescents. Self-report is generally quite accurate in school-based surveys [18], and validation of smoking status was done using carbon monoxide testing in a portion of students [20]. The difference between estimated and actual prevalence could be exaggerated if students underreport their own smoking behaviour while also overestimating others’ smoking, but that is unlikely in this study since the reported smoking rate is high for both samples when compared with the documented national rate [29, 30]. For responses describing the behaviour of others (such as family and friends’ smoking), the student may be providing their perception of the actual state rather than the true actual state, which may be a confounding factor since the study examines perceived prevalence.
The survey instrument used to collect the data has been tested and shown to be valid and reliable [20], strengthening conclusions based on it. The multivariate analysis also strengthens the findings, since the contribution of each individual factor is modelled in the context of the other variables included.
Implications
The insight provided regarding the influence of both non-modifiable (grade, gender) and modifiable risk factors (close friends’ and family members’ smoking, smoking at school and in the home and smoking status) could be used to target interventions to populations exhibiting the non-modifiable risk factors or tailor interventions to address characteristics which are amenable to modification via intervention. Additional research would be required to evaluate the benefits of a targeted or tailored approach to intervention programming.
In addition, more research is needed to investigate why adolescents overestimate their peers’ smoking prevalence. Further studies in this area should include other variables that may be important for estimation of smoking prevalence. For example, it makes intuitive sense that psychological factors may play a role, since estimation is a cognitive construct, and that social factors may also contribute to perceptions of peer behaviour. Other variables that influence smoking may also be important for estimation; for example, Unger and Rohrbach [18] included such additional variables as smoking prevalence among similar others, perceived access to cigarettes, cigarette offers and perceptions of smoking on television—all of which were significant except for similar others. One variable in particular that seems worth exploring is the effect of the media on students’ perceptions of smoking, given the connection between exposure to smoking in the media and youth smoking behaviour [33, 34]. In addition, qualitative study in this area may prove useful in revealing the thought processes through which students form their prevalence estimates and factors that may affect them.
Conclusions
The majority of adolescents overestimate the prevalence of peer smoking. Multiple logistic regression and subsequent validation of the model with a second data set indicated that grade, close friends' smoking, gender, seeing smoking at school, family members' smoking, smoking in the home and smoking status have a clear association with overestimation; school smoking rate and susceptibility to smoking show a tentative relationship and warrant further study. Other factors may also be important for prevalence estimation, and further research is needed to identify these factors. Since adolescents tend to overestimate peer smoking prevalence and perceived prevalence is in turn linked to smoking behaviour, interventions should focus on creating realistic perceptions of the prevalence of smoking. These interventions should be targeted towards younger students and females, as they represent demographic groups that are more likely to have inflated prevalence estimates.
|
Conflict of interest statement |
|---|
|
TopAbstractIntroductionMethodsResultsDiscussionConflict of interest statementAcknowledgementsReferences |
|---|
None declared.
|
Acknowledgements |
|---|
|
TopAbstractIntroductionMethodsResultsDiscussionConflict of interest statementAcknowledgementsReferences |
|---|
We would like to thank the data analysts at the Population Health Research Group at the University of Waterloo for consulting on statistics and SAS programming. This research was supported by Ontario Tobacco Research Unit Graduate Studentship for Research in Tobacco Control awarded to the first author. The host studies were supported by the Social Sciences and Humanities Research Council and the National Cancer Institute of Canada with funds from the Canadian Cancer Society.
|
References |
|---|
|
TopAbstractIntroductionMethodsResultsDiscussionConflict of interest statementAcknowledgementsReferences |
|---|
1. Makomaski Illing EM, Kaiserman MJ. Mortality attributable to tobacco use in Canada and its regions, 1998. Can J Public Health (2004) 95:38–44.[Web of Science][Medline]
2. Health Canada. Summary of Annual Results for 2005. Canadian Tobacco Use Monitoring Survey (CTUMS), 2005. Available at: http://www.hc-sc.gc.ca/hl-vs/tobac-tabac/research-recherche/stat/ctums-esutc/2005/ann_summary-sommaire_e.html. Accessed: 29 January 2007.
3. Health Canada. Smoking in Canada: An Overview. Canadian Tobacco Use Monitoring Survey (CTUMS), 2001. Available at: http://www.hc-sc.gc.ca/hl-vs/tobac-tabac/research-recherche/stat/ctums-esutc/fs-if/2001/2001overview_e.html. Accessed: 29 January 2007.
4. Ajzen I. The Theory of Planned Behavior. Organ Behav Hum Decis Process (1991) 50:179–211.[CrossRef][Web of Science]
5. Bandura A. Social Foundations of Thought and Action: A Social Cognitive Theory (1986) Englewood Cliffs, NJ: Prentice Hall.
6. Flay BR, Petraitis J, Hu FB. Psychosocial risk and protective factors for adolescent tobacco use. Nicotine Tob Res. (1999) 1:S59–65.[Abstract]
7. Botvin GJ, Botvin EM, Baker E, et al. The false consensus effect: predicting adolescents’ tobacco use from normative expectations. Psychol Rep (1992) 70:171–8.[CrossRef][Web of Science][Medline]
8. Chassin L, Presson C, Sherman S, et al. Predicting the onset of cigarette smoking in adolescents: a longitudinal study. J Appl Soc Psychol (1984) 14:224–43.[CrossRef][Web of Science]
9. Ellickson PL, Bird CE, Orlando M, et al. Social context and adolescent health behavior: does school-level smoking prevalence affect students' subsequent smoking behavior? J Health Soc Behav. (2003) 44:525–35.[CrossRef][Web of Science][Medline]
10. Iannotti RJ, Bush PJ, Weinfurt KP. Perception of friends’ use of alcohol, cigarettes, and marijuana among urban schoolchildren: a longitudinal analysis. Addict Behav (1996) 21:615–32.[CrossRef][Web of Science][Medline]
11. Iannotti RJ, Bush PJ. Perceived vs. actual friends’ use of alcohol, cigarettes, marijuana, and cocaine: which has the most influence? J Youth Adolesc (1992) 21:375–89.[CrossRef][Web of Science]
12. Jackson C. Initial and experimental stages of tobacco and alcohol abuse during late childhood: relation to peer, parent, and personal risk factors. Addict Behav (1997) 22:685–98.[CrossRef][Web of Science][Medline]
13. Simons-Morton B, Crump AD, Haynie DL, et al. Psychosocial, school, and parent factors associated with recent smoking among early-adolescent boys and girls. Prev Med (1999) 28:138–48.[CrossRef][Web of Science][Medline]
14. Page RM. College students' distorted perception of the prevalence of smoking. Psychol Rep (1998) 82:474.[Web of Science][Medline]
15. Page RM, Hammermeister J, Roland M. Are high school students accurate or clueless in estimating substance use among peers? Adolescence (2002) 37:567–73.[Web of Science][Medline]
16. Perkins HW, Meilman PW, Leichliter JS, et al. Misperceptions of the norms for the frequency of alcohol and other drug use on college campuses. J Am Coll Health (1999) 47:253–8.[Web of Science][Medline]
17. Sussman S, Dent CW, Mestel-Rauch J, et al. Adolescent nonsmokers, triers, and regular smokers' estimates of cigarette smoking prevalence: when do overestimations occur and by whom? J Appl Soc Psychol (1988) 18:537–51.[CrossRef][Web of Science]
18. Unger JB, Rohrbach LA. Why do adolescents overestimate their peers’ smoking prevalence? Correlates of prevalence estimates among California 8th-grade students. J Youth Adolesc (2002) 31:147–53.[CrossRef][Web of Science]
19. Cameron R, Manske SR, Brown KS, et al. Integrating public health policy, practice, evaluation, surveillance, and research using local data collection and feedback systems: the example of the school health action planning and evaluation system. Am J Public Health. in press.
20. The School Smoking Profile. a tool for evaluating interventions. In: Presentation Given at the Third National Conference on Tobacco or Health, Science in Policy and Action (2002) Ottawa, Canada.
21. Reitsma AH, Manske SR. Smoking in Ontario schools. Does policy make a difference? Can J Public Health (2004) 95:214–8.[Web of Science][Medline]
22. Cameron R, Brown KS, Best JA, et al. Effectiveness of a social influences smoking prevention program as a function of provider type, training method, and school risk. Am J Public Health (1999) 89:1827–31.
23. Leatherdale S, Cameron R, Brown S, et al. Senior student smoking at school, student characteristics, and smoking onset among junior students: a multi-level analysis. Prev Med (2005) 40:853–9.[CrossRef][Web of Science][Medline]
24. Leatherdale S, McDonald P, Cameron R, et al. A multi-level analysis examining the relationship between social influences for smoking and smoking onset. Am J Health Behav (2005) 29:520–30.[Web of Science][Medline]
25. Pierce JP, Choi WS, Gilpin EA, et al. Validation of susceptibility as a predictor of which adolescents take up smoking in the United States. Health Psychol (1996) 15:355–61.[CrossRef][Web of Science][Medline]
26. SAS Institute Inc. The SAS System for Windows (2001) Cary, NC: SAS Institute Inc. Version 8.2.
27. Steffian G. Correction of normative misperceptions: an alcohol abuse prevention program. J Drug Educ (1999) 29:115–38.[CrossRef][Web of Science][Medline]
28. Eisenberg ME, Forster JL. Adolescent smoking behavior: measures of social norms. Am J Prev Med (2003) 25:122–8.[Web of Science][Medline]
29. Health Canada. Summary of Results for 2001. Canadian Tobacco Use Monitoring Survey (CTUMS), 2001. Available at: http://www.hc-sc.gc.ca/hl-vs/tobac-tabac/research-recherche/stat/ctums-esutc/2001/ar-rpa-2001_e.html. Accessed: 29 January 2007.
30. Health Canada. Summary of Results for 2003. Canadian Tobacco Use Monitoring Survey (CTUMS), 2003. Available at: http://www.hc-sc.gc.ca/hl-vs/tobac-tabac/research-recherche/stat/ctums-esutc/2003/ann_summary-sommaire_e.html. Accessed: 29 January 2007.
31. Best JA, Thomson SJ, Santi SM, et al. Preventing cigarette smoking among school children. Annu Rev Public Health (1988) 9:161–201.[Web of Science][Medline]
32. Smith P, Begley L, O'Loughlin JL, et al. Chapter 3: youth smoking behaviour. In: 2002 Youth Smoking Survey Technical Report—Health Canada, ed. (2005) Ottawa, Canada: Minister of Supply and Services Canada. Cat. H46–1/44-2002E.
33. Charlesworth A, Glantz SA. Smoking in the movies increases adolescent smoking: a review. Pediatrics (2005) 116:1516–28.
34. McCool JP, Cameron LD, Petrie KJ. The influence of smoking imagery on the smoking intentions of young people: testing a media interpretation model. J Adolesc Health (2005) 36:475–85.[CrossRef][Web of Science][Medline]
Received on November 7, 2005; accepted on November 21, 2006
CiteULike 
Connotea 
Del.icio.us What's this?
| ||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||