Background: Smoking is a serious health threat and identifying risk factors for smoking is thus of great importance. The aim of the study was to examine the effects of social-cognitive factors and school factors on lifetime smoking status among adolescents.

Methods: The study was based on cross-sectional data on 2,913 Danish adolescents in grade 7 attending 118 randomly selected public schools. Social-cognitive factors were examined with five measures: self-efficacy to resist pressure to smoke, social influence (norms), social influence (behavior), social influence (pressure), and attitude. We used multilevel analyses to estimate the associations between social-cognitive factors and lifetime smoking status as well as the group-level effects of school, school class, and gender group in the school class.

Results: Each social-cognitive factor was significantly associated with lifetime smoking status, even when several potential confounders and the effects of school, school class, and gender group were taken into account. Of the three group-level school factors, gender group in the school class had the strongest effect on smoking status.

Conclusion: We conclude that self-efficacy to resist pressure to smoke, attitude, and the three types of social influence are significantly associated with lifetime smoking status, even when the effects of group-level school factors are taken into account. The strong effect of gender group in school class on lifetime smoking status indicates that prevention actions should address the social context of adolescents. (Cancer Epidemiol Biomarkers Prev 2008;17(8):1862–71)

Smoking is the most important preventable cause of cancer in several countries (1). In Denmark, about 20% of deaths among adults ages >35 years are due to smoking (2). Smoking often starts in adolescence; in Denmark, 38% of boys and 32% of girls have tried smoking at age 13 years (3). About 80% of adolescents who smoke at age 15 years will continue smoking in adulthood and about 50% of those who smoke when they are adults will still smoke at age 60 years (4, 5). It is therefore important to identify risk factors for smoking or factors associated with smoking status among adolescents.

Several review studies (6, 7) and a dissertation review (8) have shown that social-cognitive factors such as believing that your friends have positive attitudes to smoking and believing that it is difficult to refuse smoking offers are associated with smoking status, and some of these factors have been found to be responsive to school-based smoking prevention programs (9). The attitude, social influence, and self-efficacy (ASE; ref. 10) model describes how cognitive processes influence the intention to smoke, which often leads to actual smoking (Fig. 1). According to this model, the three social-cognitive factors (ASE model) independently influence smoking initiation.

Figure 1.

Revised version of ASE model.

Figure 1.

Revised version of ASE model.

Close modal
  • Attitude is conceptualized as positive and negative evaluations of smoking (such as believing that smoking is bad for your health).

  • Social influence is measured in three dimensions: social norms (teachers' and friends' attitude toward smoking), perceived smoking behavior (among parents and friends), and direct pressure, which may overlap and may work in different directions (11).

  • Self-efficacy to resist pressure to smoke is defined as an individual's expectations of capability (12) and focus is on social refusal self-efficacy, which, in the context of smoking, is the expected ability to refrain from smoking in social situations. For both smokers and nonsmokers, the measure of self-efficacy is expected to tap into the belief in one's ability to resist group pressure and also tap into the intention to smoke. Persons who have not tried smoking and do not intend to try smoking may still doubt their ability to refuse. Persons who have not tried smoking may also still intend to try it.

Although many studies have suggested the importance of psychologic risk factors for smoking status, only a few have investigated the effect of the environment. Adolescents spend the majority of their time at school and this environment is thus likely to have a large effect on their lives. However, does that mean that you could have a higher risk of becoming a smoker at one school compared to another? Previous studies have shown that the school (13-16), the school class, and the gender group in the school class (17) may play important roles in smoking status. The school environment can be important because adolescents at the same school share attributes such as their physical environment and neighborhood. Adolescents in the same school class furthermore have the same teachers and spend up to 9 years together on school days. Adolescents in the same gender group (boys versus girls) in a school class also often share friends (18). These school factors have been found to be of particular importance in studies in which the population is sampled by school or school class, because persons in the same sampling unit often have more in common than with those in a different unit. This is referred to as a “clustering effect.” When sampling has been done by school, the school factors can be measured at a group level (instead of at the individual level); this way, one can take into account the variability between schools instead of only between adolescents.

It is important to account for psychological risk factors and environmental factors simultaneously, as these may interact or mediate each other (16) and because a clustering effect might distort the effects of individual psychological factors (19). We conducted a large study involving more than 2,900 adolescents representative of the general adolescent population in Denmark to examine the associations between social-cognitive factors and lifetime smoking, taking into account the effects of the school environment.

Materials

The study was based on data for the Danish Youth Cohort collected by the Danish National Institute of Public Health in 2004; data collection has been described previously (20). The data used in this study were baseline data (grade 7) from a longitudinal study of adolescents in six counties of Denmark followed up from grades 7 to 9. The mean age of the participants was 13 years. A total of 4,819 of 7,037 eligible pupils (68%) responded with informed consent from their parents, representing almost 9% of the total population of children in grade 7 at public schools in Denmark (n = 56,739; ref. 21). Of the 4,819 adolescents who agreed to participate, 29 were excluded because of nonresponse. In Denmark, the school system is organized around the school class, which is a group of children who have all their school hours together ideally throughout the first 9 years of primary and lower secondary school (22). In this study, adolescents were asked about which class they were in, not at baseline but in a follow-up study. As not all adolescents chose to participate in the follow-up study, information on class was not available for all persons. We decided to exclude all persons for whom information on school class was not available. This decision was made based on the importance of the effect from school class and gender group in school class found in a study by Rasmussen (17). The final population of the study consisted of 2,913 adolescents distributed across 118 schools, 230 classes, and 436 gender groups. There was an average of two classes per school, with an average of 13 pupils per class. In 24 of 230 classes, only one gender was represented.

Measures

We were in this study interested in factors associated with lifetime smoking and thus with all types of smoking experiences among adolescents ages 13 years. Ideally, separate analyses should be conducted for several different stages of smoking [e.g., smoking a puff, previous smoking (monthly, weekly, and daily), current monthly smoking, and current regular smoking (weekly and daily)]. The number of persons in these categories were small in the current study (e.g., daily and weekly smokers; n = 59), and to obtain sufficient statistical power, smoking was measured as lifetime smoking, which implies everything from smoking a puff and previous smoking to current daily smoking. The information was based on answers to two questions (Table 1). The first question was on current smoking: “Do you smoke cigarettes?” The second question was on previous smoking: “Which sentence fits you the best? 1. I have stopped smoking, I smoked at least once a day,” etc. Adolescents were considered to be a lifetime smoker if they scored 1 to 5 on the first or second item. Thus, if a person answered “6. No, I do not smoke” in question 1 and “5. I have smoked a total of a couple of cigarettes at the most” in question 2, they would be considered a lifetime smoker. If an answer was missing to one question, the other response was used. If answers were missing to both questions, they were excluded from further analyses.

Table 1.

Distribution of answers to two questions about smoking from 2,913 Danish adolescents

Questionn (%)
Do you smoke cigarettes?  
    Missing 25 (1) 
    1. Yes at least once a day 32 (1) 
    2. Yes, at least once a week 27 (1) 
    3. Yes, at least once a month 24 (1) 
    4. Yes, but less than once a month 39 (1) 
    5. I have smoked a total of a couple of cigarettes at the most 200 (7) 
    6. No, I do not smoke 2,566 (88) 
Which sentence fits you the best?  
    Missing 347 (12) 
    1. I have stopped smoking, I smoked at least once a day 17 (1) 
    2. I have stopped smoking, I smoked at least once a week 14 (0) 
    3. I have stopped smoking, I smoked at least once a month 5 (0) 
    4. I have stopped smoking, I smoked less than once a month 16 (1) 
    5. I have smoked a total of a couple of cigarettes at the most 299 (10) 
    6. I have never tried smoking, not even a single puff 2,215 (76) 
Questionn (%)
Do you smoke cigarettes?  
    Missing 25 (1) 
    1. Yes at least once a day 32 (1) 
    2. Yes, at least once a week 27 (1) 
    3. Yes, at least once a month 24 (1) 
    4. Yes, but less than once a month 39 (1) 
    5. I have smoked a total of a couple of cigarettes at the most 200 (7) 
    6. No, I do not smoke 2,566 (88) 
Which sentence fits you the best?  
    Missing 347 (12) 
    1. I have stopped smoking, I smoked at least once a day 17 (1) 
    2. I have stopped smoking, I smoked at least once a week 14 (0) 
    3. I have stopped smoking, I smoked at least once a month 5 (0) 
    4. I have stopped smoking, I smoked less than once a month 16 (1) 
    5. I have smoked a total of a couple of cigarettes at the most 299 (10) 
    6. I have never tried smoking, not even a single puff 2,215 (76) 

The social-cognitive factors (Fig. 2) were measured based on items and indexes constructed and examined for their “criterion-related construct validity” in a previous study (20). Three of the indexes [social influence (norms; seven ordinal levels), social influence (pressure; four ordinal levels), and attitude (nine ordinal levels)] had satisfactory validity and were therefore used in the study. Two indexes [self-efficacy to resist pressure to smoke and social influence (behavior)] did not have satisfactory validity and had to be measured from individual items. We used one individual measure of self-efficacy to resist pressure to smoke (five ordinal levels) and three individual measures of social influence [behavior; mother's, father's, and best friend's smoking (each with three nominal values)].

Figure 2.

Social-cognitive factors and their item questions.

Figure 2.

Social-cognitive factors and their item questions.

Close modal

Several demographic and school well-being factors that have been shown to be potential confounders (17, 23) were included in the analyses (Table 2). Family affluence was assessed from three items: “How many cars do you have in your home?” none, one, two, more than two; “Do you have a room to yourself?” no, yes; and “Last year, how many times did you go on vacation with your family?” none, once, twice, more than twice. The item responses were dichotomized and summarized for a total score from 1 to 4. The family affluence scale has been found to be an appropriate proxy for socioeconomic status among adolescents (24-26). Besides the individual factors, three group levels were explored: school, school class, and gender group in the school class.

Table 2.

Lifetime smoking by potential confounders and gender among 2,913 Danish adolescents

Potential confoundern (%) Smokers
Girls (n = 1,487)Boys (n = 1,426)
Living with only one parent   
    Missing 12 (50) 14 (21) 
    No 1,055 (18) 1,033 (21) 
    Yes 420 (32) 379 (33) 
Family affluence scale   
    Missing 65 (20) 73 (22) 
    1 Very high 1,206 (21) 1,117 (24) 
    2 Somewhat high 182 (30) 194 (24) 
    3 Somewhat low 28 (29) 37 (30) 
    4 Very low 6 (17) 5 (20) 
Liking school   
    Missing 45 (16) 51 (20) 
    1 Very positive 333 (14) 260 (13) 
    2 Somewhat positive 646 (16) 587 (18) 
    3 Somewhat negative 379 (34) 415 (34) 
    4 Very negative 84 (57) 113 (47) 
Liking classmates   
    Missing 46 (17) 51 (20) 
    1 Very positive 722 (18) 673 (21) 
    2 Somewhat positive 514 (26) 575 (25) 
    3 Somewhat negative 170 (29) 113 (37) 
    4 Very negative 35 (31) 14 (50) 
Non-Danish ethnicity   
    Missing 10 (40) 10 (30) 
    No 1,298 (22) 1,225 (24) 
    Yes 179 (24) 191 (22) 
Family is religious   
    Missing 10 (60) 10 (30) 
    No 86 (22) 114 (29) 
    Yes 1,338 (22) 1,243 (23) 
    Do not know 53 (21) 59 (31) 
Potential confoundern (%) Smokers
Girls (n = 1,487)Boys (n = 1,426)
Living with only one parent   
    Missing 12 (50) 14 (21) 
    No 1,055 (18) 1,033 (21) 
    Yes 420 (32) 379 (33) 
Family affluence scale   
    Missing 65 (20) 73 (22) 
    1 Very high 1,206 (21) 1,117 (24) 
    2 Somewhat high 182 (30) 194 (24) 
    3 Somewhat low 28 (29) 37 (30) 
    4 Very low 6 (17) 5 (20) 
Liking school   
    Missing 45 (16) 51 (20) 
    1 Very positive 333 (14) 260 (13) 
    2 Somewhat positive 646 (16) 587 (18) 
    3 Somewhat negative 379 (34) 415 (34) 
    4 Very negative 84 (57) 113 (47) 
Liking classmates   
    Missing 46 (17) 51 (20) 
    1 Very positive 722 (18) 673 (21) 
    2 Somewhat positive 514 (26) 575 (25) 
    3 Somewhat negative 170 (29) 113 (37) 
    4 Very negative 35 (31) 14 (50) 
Non-Danish ethnicity   
    Missing 10 (40) 10 (30) 
    No 1,298 (22) 1,225 (24) 
    Yes 179 (24) 191 (22) 
Family is religious   
    Missing 10 (60) 10 (30) 
    No 86 (22) 114 (29) 
    Yes 1,338 (22) 1,243 (23) 
    Do not know 53 (21) 59 (31) 

Statistical Analysis

Multilevel logistic regression models (GLIMMIX procedure, SAS 9.1 statistical package, SAS Institute) were used to estimate the (fixed) effects on an odds ratio (OR) scale of the social-cognitive factors and the (random) effects of school, school class, and gender group in school class. Factors that are included conventionally in multilevel logistic regression are called “fixed effects” to distinguish them from random effects. Like conventional (multivariate) logistic regression models, multilevel logistic regression models relate the log odds of the event (e.g., smoking) to a linear combination of risk factors. Although, in conventional logistic regression, estimates of variables for each level of the factors are of interest, in multilevel logistic regression analysis, levels of selected factors are assumed to be described by a Gaussian distribution and are included in the logistic model as such. We included school, school class, and gender group as nested random effects in the model. This implies that that these factors were modeled in a way that takes into account the clustering between persons within the different structures. As estimates of random variance components are often difficult to interpret, they were converted to median OR (27) and thus translated into an OR scale. Median OR is theoretically the increased median risk resulting from a move to, for example, another school with a higher risk for smoking. It is directly comparable with the OR of individual variables (19). Median OR was computed from the formula (19):

where VA is the corresponding group-level factor variance and 0.6745 is the 75th percentile of the cumulative standard Gaussian distribution.

To investigate the effects of the group-level factors and the mediating abilities of social-cognitive factors, the ASE model was constructed by examining an increasing portfolio of models. First, models that included only the group-level factors added individually were examined to assess the raw context effects on smoking. Then, each social-cognitive factor was added individually to the model to assess the association with the individual factor. This was repeated in the presence of demographic and school well-being factors as potential confounders. Finally, a model that included all the social-cognitive factors and confounders was used to estimate if the social-cognitive factors had individual effects on smoking status.

In a previous study, we examined the criterion-related construct validity of the social-cognitive indexes and found that some had positive local dependence and differential item functioning (20). Positive local dependence implies that the dependence among certain items is stronger than their relation to the assumed common latent trait; differential item functioning implies that item scores depend not only on the assumed latent trait but also on exogenous variables such as gender and ethnicity (28). The presence of local dependence in the indexes used in the current study means that the OR should be interpreted with caution. The overall association of these indexes with smoking was therefore additionally examined with F tests. The presence of differential item functioning from gender and liking school on attitude and from gender on social influence (norms) was accounted for by including corresponding interaction terms. For each of the continuous predictors, self-efficacy to resist pressure to smoke, social influence (norms), social influence (pressure), and attitude, we tested whether the association with smoking status was linear on a logit scale. Linear trends were seen for social influence (norms) and social influence (pressure), whereas a quadratic term was added for attitude, and a separate estimate for score level 5 was added for self-efficacy to resist pressure to smoke to account for nonlinearity. The three measures of social influence (behavior) were included as categorical variables.

In a subanalysis, we furthermore examined the mutually and confounder adjusted fixed effects of the social-cognitive factors on lifetime smoking while excluding persons who were current weekly or daily smokers (n = 59).

At baseline in 2004, the mean age was 13 years. Slightly more boys (24%) than girls (22%) had tried smoking. Few adolescents were regular smokers: 23% had ever tried smoking, 17% had tried a couple of cigarettes at the most, and 6% had tried more than a couple of cigarettes. The results of descriptive statistics showed that self-efficacy to resist pressure to smoke, social influence (norms), social influence (behavior), social influence (pressure), and attitude were associated with smoking status (Table 3). Of the potential confounders, family structure (living with one parent), not liking school, negative relationships with classmates, and lack of family religiousness among boys appeared to be associated with smoking (Table 2).

Table 3.

Lifetime smoking by social-cognitive factors and gender among 2,913 Danish adolescents

n (%) Smokers
Girls (n = 1,487)Boys (n = 1,426)
Self-efficacy   
    Missing 17 (0) 19 (11) 
    1 Definitely refrain from smoking 994 (13) 1,015 (17) 
    2 292 (29) 188 (37) 
    3 58 (41) 81 (44) 
    4 77 (75) 46 (72) 
    5 Definitely not refrain from smoking 49 (57) 77 (34) 
Social influence (norms)   
    Missing 25 (12) 24 (17) 
    1 Negative 936 (15) 873 (17) 
    2 215 (22) 175 (26) 
    3 161 (44) 156 (35) 
    4 119 (43) 169 (39) 
    5 22 (68) 18 (89) 
    6 6 (67) 11 (91) 
    7 Positive 3 (67) 0 (0) 
Social influence (behavior)   
Mother smokes   
    Missing 53 (23) 37 (14) 
    No 894 (17) 875 (19) 
    Do not know 14 (29) 20 (30) 
    Yes 526 (30) 494 (33) 
Father smokes   
    Missing 77 (22) 60 (23) 
    No 832 (18) 804 (21) 
    Do not know 27 (41) 36 (31) 
    Yes 551 (27) 526 (29) 
Best friend smokes   
    Missing 19 (0) 19 (16) 
    No 1,206 (16) 1,135 (18) 
    Do not know 104 (35) 148 (33) 
    Yes 158 (66) 124 (70) 
Social influence (pressure)   
    Missing 26 (12) 26 (19) 
    1 Low pressure 1,272 (17) 1,245 (20) 
    2 76 (39) 77 (49) 
    3 80 (66) 58 (57) 
    4 High pressure 33 (73) 20 (80) 
Attitude toward smoking   
    Missing 16 (6) 17 (6) 
    1 Negative 1,088 (15) 1,061 (18) 
    2 148 (31) 113 (35) 
    3 103 (35) 85 (29) 
    4 54 (56) 52 (62) 
    5 53 (66) 66 (42) 
    6 10 (50) 14 (50) 
    7 9 (33) 9 (89) 
    8 6 (100) 5 (100) 
    9 Positive 0 (0) 4 (75) 
n (%) Smokers
Girls (n = 1,487)Boys (n = 1,426)
Self-efficacy   
    Missing 17 (0) 19 (11) 
    1 Definitely refrain from smoking 994 (13) 1,015 (17) 
    2 292 (29) 188 (37) 
    3 58 (41) 81 (44) 
    4 77 (75) 46 (72) 
    5 Definitely not refrain from smoking 49 (57) 77 (34) 
Social influence (norms)   
    Missing 25 (12) 24 (17) 
    1 Negative 936 (15) 873 (17) 
    2 215 (22) 175 (26) 
    3 161 (44) 156 (35) 
    4 119 (43) 169 (39) 
    5 22 (68) 18 (89) 
    6 6 (67) 11 (91) 
    7 Positive 3 (67) 0 (0) 
Social influence (behavior)   
Mother smokes   
    Missing 53 (23) 37 (14) 
    No 894 (17) 875 (19) 
    Do not know 14 (29) 20 (30) 
    Yes 526 (30) 494 (33) 
Father smokes   
    Missing 77 (22) 60 (23) 
    No 832 (18) 804 (21) 
    Do not know 27 (41) 36 (31) 
    Yes 551 (27) 526 (29) 
Best friend smokes   
    Missing 19 (0) 19 (16) 
    No 1,206 (16) 1,135 (18) 
    Do not know 104 (35) 148 (33) 
    Yes 158 (66) 124 (70) 
Social influence (pressure)   
    Missing 26 (12) 26 (19) 
    1 Low pressure 1,272 (17) 1,245 (20) 
    2 76 (39) 77 (49) 
    3 80 (66) 58 (57) 
    4 High pressure 33 (73) 20 (80) 
Attitude toward smoking   
    Missing 16 (6) 17 (6) 
    1 Negative 1,088 (15) 1,061 (18) 
    2 148 (31) 113 (35) 
    3 103 (35) 85 (29) 
    4 54 (56) 52 (62) 
    5 53 (66) 66 (42) 
    6 10 (50) 14 (50) 
    7 9 (33) 9 (89) 
    8 6 (100) 5 (100) 
    9 Positive 0 (0) 4 (75) 

Social-Cognitive Factors

The multilevel analyses showed that the social-cognitive factors were significantly associated with smoking status after adjustment for the random effects of school, school class, and gender group in the school class (Table 4; Fig. 3). In addition, we generally observed that the estimated associations became weaker after adjustment for potential confounders and were weakened slightly further after mutual adjustment for all social-cognitive factors. Nevertheless, all social-cognitive factors remained significant in the adjusted models (Table 4). F tests in the final model confirmed that each of the indexes was significantly associated with smoking status.

Table 4.

Multilevel analysis of lifetime smoking: unadjusted and mutually adjusted fixed effects of social-cognitive factors among 2,913 Danish adolescents

Individual levelUnadjusted OR (95% CI)Adjusted OR* (95% CI)Adjusted OR (95% CI), (n = 2,626)
Self-efficacy (n = 2,873) (n = 2,732)  
Linear (score 1-4) per score 2.2 (2.0-2.5) 2.0 (1.8-2.3) 1.7 (1.5-1.9) 
    Score 5 vs 4 0.4 (0.3-0.6) 0.4 (0.3-0.7) 0.5 (0.3-0.9) 
Social influence (norms) (n = 2,859) (n = 2,727)  
    Per score among boys 1.6 (1.4-1.7) 1.5 (1.4-1.7) 1.3 (1.2-1.5) 
    Per score among girls 1.6 (1.5-1.8) 1.5 (1.4-1.7) 1.3 (1.1-1.4) 
Social influence (behavior)    
    Mother smokes (n = 2,821) (n = 2,688)  
        No 1.0 (—) 1.0 (—) 1.0 (—) 
        Do not know 1.6 (0.8-3.3) 1.2 (0.6-2.7) 1.7 (0.7-3.8) 
        Yes 2.0 (1.7-2.4) 1.7 (1.4-2.0) 1.7 (1.4-2.1) 
    Father smokes (n = 2,774) (n = 2,651)  
        No 1.0 (—) 1.0 (—) 1.0 (—) 
        Do not know 2.1 (1.3-3.5) 1.6 (0.9-2.8) 1.7 (1.0-3.1) 
        Yes 1.8 (1.5-2.1) 1.5 (1.2-1.8) 1.3 (1.1-1.6) 
    Best friend smokes (n = 2,872) (n = 2,729)  
        No 1.0 (—) 1.0 (—) 1.0 (—) 
        Do not know 2.3 (1.8-2.9) 1.8 (1.4-2.3) 1.2 (0.9-1.6) 
        Yes 8.6 (6.7-11.1) 7.1 (5.4-9.4) 3.4 (2.5-4.6) 
Social influence (pressure) per score (n = 2,856) (n = 2,728)  
 2.5 (2.1-2.8) 2.3 (2.0-2.6) 2.1 (1.8-2.4) 
Attitude toward smoking§ (n = 2,808) (n = 2,730)  
 1.0 (0.9-1.0) 1.8 (1.5-2.1) 1.4 (1.2-1.7) 
Individual levelUnadjusted OR (95% CI)Adjusted OR* (95% CI)Adjusted OR (95% CI), (n = 2,626)
Self-efficacy (n = 2,873) (n = 2,732)  
Linear (score 1-4) per score 2.2 (2.0-2.5) 2.0 (1.8-2.3) 1.7 (1.5-1.9) 
    Score 5 vs 4 0.4 (0.3-0.6) 0.4 (0.3-0.7) 0.5 (0.3-0.9) 
Social influence (norms) (n = 2,859) (n = 2,727)  
    Per score among boys 1.6 (1.4-1.7) 1.5 (1.4-1.7) 1.3 (1.2-1.5) 
    Per score among girls 1.6 (1.5-1.8) 1.5 (1.4-1.7) 1.3 (1.1-1.4) 
Social influence (behavior)    
    Mother smokes (n = 2,821) (n = 2,688)  
        No 1.0 (—) 1.0 (—) 1.0 (—) 
        Do not know 1.6 (0.8-3.3) 1.2 (0.6-2.7) 1.7 (0.7-3.8) 
        Yes 2.0 (1.7-2.4) 1.7 (1.4-2.0) 1.7 (1.4-2.1) 
    Father smokes (n = 2,774) (n = 2,651)  
        No 1.0 (—) 1.0 (—) 1.0 (—) 
        Do not know 2.1 (1.3-3.5) 1.6 (0.9-2.8) 1.7 (1.0-3.1) 
        Yes 1.8 (1.5-2.1) 1.5 (1.2-1.8) 1.3 (1.1-1.6) 
    Best friend smokes (n = 2,872) (n = 2,729)  
        No 1.0 (—) 1.0 (—) 1.0 (—) 
        Do not know 2.3 (1.8-2.9) 1.8 (1.4-2.3) 1.2 (0.9-1.6) 
        Yes 8.6 (6.7-11.1) 7.1 (5.4-9.4) 3.4 (2.5-4.6) 
Social influence (pressure) per score (n = 2,856) (n = 2,728)  
 2.5 (2.1-2.8) 2.3 (2.0-2.6) 2.1 (1.8-2.4) 
Attitude toward smoking§ (n = 2,808) (n = 2,730)  
 1.0 (0.9-1.0) 1.8 (1.5-2.1) 1.4 (1.2-1.7) 
*

Univariate estimates adjusted for living with only one parent, family affluence scale, liking school, liking classmates, non-Danish ethnicity, and family is religious.

Mutually adjusted estimates adjusted for all potential confounders and all social-cognitive factors.

An interaction term was added for social influence (norms) × gender because a previous validation study found differential item functioning from gender on this index.

§

Based on analyses without an interaction term for liking school and gender. See results for attitude when interaction terms are added for liking school and gender in Fig. 3.

Figure 3.

Effect of attitude on lifetime smoking by liking school and gender. A three-factor interaction term was included for attitude, attitude × liking school × gender, because a previous validation study found differential item functioning from these factors on attitude. Tests indicated that the attitude variable should ideally be used as a class variable, but we used the variable as a continuous scale, allowing for nonlinearity by adding a quadratic term.

Figure 3.

Effect of attitude on lifetime smoking by liking school and gender. A three-factor interaction term was included for attitude, attitude × liking school × gender, because a previous validation study found differential item functioning from these factors on attitude. Tests indicated that the attitude variable should ideally be used as a class variable, but we used the variable as a continuous scale, allowing for nonlinearity by adding a quadratic term.

Close modal

In the fully adjusted model, self-efficacy to resist pressure to smoke was strongly associated with smoking status [OR, 1.7; 95% confidence interval (95% CI), 1.5-1.9; Table 4]. For social influence (norms), no difference in effect was seen between boys (OR, 1.3; 95% CI, 1.2-1.5) and girls (OR, 1.3; 95% CI, 1.1-1.4). Best friend's smoking (OR, 3.4; 95% CI, 2.5-4.6) and social influence (pressure; OR, 2.1; 95% CI, 1.8-2.4) also both showed strong associations. The effect of attitude was slightly greater among girls than boys (Fig. 3) and varied only slightly with the level of liking school, the middle level showing the strongest effect. The results of the subanalysis examining the mutually and confounder adjusted fixed effects of the social-cognitive factors on lifetime smoking while excluding persons who were current weekly or daily smokers (n = 59) showed only small differences compared with when these persons were included. All social-cognitive factors had a significant effect on lifetime smoking: self-efficacy [OR, 1.6; 95% CI, 1.5-1.8], social influence (norms) same results for boys and girls [OR, 1.2; 95% CI, 1.1-1.4], mother smokes [OR, 1.6; 95% CI, 1.3-2.0], father smokes [OR, 1.3; 95% CI, 1.0-1.6], best friend smokes [OR, 3.1; 95% CI, 2.2-4.2], social influence (pressure) [OR, 2.1; 95% CI, 1.8-2.4] and attitude [OR, 1.5; 95% CI, 1.2-1.8].

Group-Level School Factors

We found wide variation in smoking prevalence among schools, classes, and gender groups in the school class. The results of null models showed that the effect of school decreased when school class and gender group in the school class were introduced and that the effect of school class vanished when gender group was introduced. The random effect of school and the effect of gender group in the school class increased when confounders and social-cognitive factors were added to the model (Table 5). The largest random effect in the adjusted model was that of gender group in the school class (median OR, 2.45).

Table 5.

Estimated random effects of null models and of adjusted models with the contextual levels of school, school class, and gender group in school class among 2,913 Danish adolescents

Null model* variance component (median OR)Adjusted model variance component (median OR)
School 0.27 (1.64) 0.32 (1.71) 
School class 0 (—) 0 (—) 
Gender group 0.35 (1.75) 0.89 (2.45) 
Null model* variance component (median OR)Adjusted model variance component (median OR)
School 0.27 (1.64) 0.32 (1.71) 
School class 0 (—) 0 (—) 
Gender group 0.35 (1.75) 0.89 (2.45) 
*

Included school, school class, and gender group in the school class.

Median OR is theoretically the increased median risk due to moving to, for example, another school with a higher risk for smoking and is directly comparable with the OR of an individual variable.

Included school, school class, and gender group in the school class with adjustment for all potential confounders and all social-cognitive factors.

The results of this study show that all the social-cognitive factors examined were significantly associated with lifetime smoking status. The association between self-efficacy to resist pressure to smoke and smoking status suggests that a belief in being able to refrain from smoking is important for whether or not a person smokes, but the direction of the causal pathway should be confirmed in longitudinal studies. Our results are in accordance with those of previous studies, including a high-quality review (8), which found that self-efficacy is associated with smoking status, and a Dutch study of 85 adolescents, which found that this factor makes a unique contribution among the ASE factors (29).

Our results also showed that all three types of social influence were significantly associated with smoking status. The association between social influence (norms) and smoking status suggests that the norms of best friends, friends, and teachers toward smoking are important for smoking status. These results are consistent with those of a review, which supported associations between smoking status and the norms of peers (6). Also, a cross-sectional study (n = 6,274) on social norms showed a strong association between smoking status and adults disapproving of adolescent smoking (30).

We found strong associations between friends' smoking, parental smoking, and smoking status. Previous studies have had diverging results with regard to parental smoking. A large longitudinal study of a cohort of 5,520 families showed that the probability that having two parents who smoked would influence a child to try smoking was 53% (31), but two earlier reviews concluded that, although peer smoking has consistently been found to influence smoking status, the results for parental smoking have been inconsistent (8, 32).

Social influence (pressure) was strongly associated with smoking status, with an estimated doubling of risk per score. This suggests that direct pressure from best friends, friends, and other adolescents may play an important role in lifetime smoking. It should be noted that adolescents might be influenced to smoke by their peers, but they might also select their friends based on their smoking status. A longitudinal study of 20,747 adolescents in the United States, in which structural equation models were used, showed that both peer influence and peer selection were present, peer influence being the most salient (33), but a review of previous qualitative studies in this area concluded that peer selection plays an important role (34).

In accordance with previous studies that found an effect of attitude, we found a significant association between attitude and smoking status. A review pointed out that positive attitude is especially important (8), whereas a Hungarian cross-sectional study of 261 randomly selected adolescents, in which factor analysis was used, found that an antismoking attitude was important (35). We found a positive effect of a positive attitude on smoking status, indicating that considering that smoking is not stupid or bad for the health makes an adolescent more likely to try smoking. The effect of attitude was slightly greater among girls and among pupils at middle levels of liking school. It is possible that there is a stronger concordance between attitude and smoking behavior among girls than boys and that attitude plays a more important role among pupils with a neutral perception of going to school than among those with a positive perception of school.

In summary, the associations found in this study between smoking status and social-cognitive factors reaffirm the place of these factors in the ASE model.

Concerning the group-level effects, the median OR for gender group in the school class, after considering the effect of school and school class, was 2.45, indicating that if an adolescent moves from one randomly chosen gender group to another with a higher risk for smoking, the risk is increased by a median of 2.45 times. The finding that gender group in the school class is the strongest group-level factor suggests that it is a more refined way of measuring a class effect because it takes into account differences between gender groups in smoking prevalence and that in some school classes, for example, most of the boys are smoking, whereas none or few of the girls are smoking or the other way around. Our results on gender group in the school class suggest that the social influence of peers does not operate across the gender boundary. This implies that the girls are influencing each other to smoke and the boys are influencing each other to smoke. This is possibly due to the fact that close friendships are often gender specific in this age group and that it is especially within close friendships adolescents influence and model each other. Previous studies of the role of gender in smoking initiation include a qualitative study with in-depth interviews of 85 adolescents in Northern Ireland, which suggested that girls often link smoking with romance and femininity, whereas boys link smoking with being cool, hard, and masculine (36).

Although the data support was weak, a review reported some gender differences in psychosocial determinants of smoking, for example, that female smoking was associated with self-confidence, rebellion, and social experience, whereas male smoking was associated with social insecurity (37). Our results showed a strong effect of gender group in the school class, even after accounting for individual measures of social influence, suggesting that the social influence might derive from numerous sources simultaneously. The mechanisms underlying the effect of gender group in a school class have not been articulated previously but could, as described in a discussion of theoretical and empirical findings on peer and adolescent smoking (38), be associated with adolescents' decisions to start smoking in order to fit in or to affect their popularity, individuality, status, and thus their own social position in relation to the dynamics of the gender group.

The importance of gender group in the school class also suggests that the group with the highest proximity is the most important for smoking status. Consistent with our findings, previous studies have found that close peer groups are more important than more peripheral (17, 39, 40). A longitudinal study of 6,527 adolescents in the United States showed that school-level smoking was not associated with smoking initiation, whereas peer smoking was (39). A longitudinal study of 2,883 adolescents in Sweden showed that class-related rather than school-related factors were associated with smoking initiation (40). If the three group-level factors are seen from a social learning perspective, they can be incorporated into the ASE model (Fig. 1). The three hierarchical group-level factors could represent three levels of social proximity: the school, with low proximity, defined in social learning theory as the “imposed environment” (41); the school class, with medium proximity; and the gender group in the school class, with the highest proximity, similar to the “selected environment” in social learning theory (41). Further studies are needed to explore the effect of gender group in the school class to identify specific factors that explain this effect.

One of the main advantages of this study is the large, gender-balanced, population-based sample of more than 2,900 adolescents at 118 schools, which were randomly selected from the public school system. Small samples have sometimes been used in studies in this area (29, 35), but the advantage of a larger sample is that it allows use of large statistical models that include many relevant variables. Inclusion of adolescents in six counties meant that those in both rural and urban areas were represented and that the study population was representative of the general population of the same age group with regard to several exogenous variables, such as ethnicity and family structure (42, 43). Another advantage is that we were able to take into account results from a previous validation study that provided evidence of deviations from criterion-related construct validity for the social-cognitive indexes. Items are often collapsed into scales without further psychometric investigation (30), but ignoring these issues can distort results. In addition, because of the hierarchical structure of the data, multilevel analyses could be done with three nested group levels. Instead of measuring gender group in the school class as a group-level effect, one could measure an interaction between school class and gender, but this would not take into account the hierarchical structure of adolescents within the specific school and school class (17). To our knowledge, this is the first study to include group-level factors in multilevel analyses for exploring the ASE model factors. Finally, this study included adjustment for several important confounders that are rarely included, such as liking school, liking classmates, and family structure.

One limitation of the study was that it was based on cross-sectional data, which provide weak evidence for the direction of a cause-effect relation. Despite its cross-sectional design, the results of the study pointed to important associations, which should be explored in longitudinal studies. In this study, we examined lifetime smoking including both adolescents who, for example, smoke daily and who have smoked only a puff. Previous studies have shown that different smoking stages may have different cognitive characteristics (44). Ideally, separate analyses should thus be done for different stages of smoking [e.g., smoking a puff, previous smoking (monthly, weekly, daily), current monthly smoking, and current regular smoking (weekly and daily)], but because this study included adolescents ages 13 years (an age group with few smokers in each of these stages), this would have resulted in lack of sufficient statistical power.

As seen in Table 1, the majority of adolescents who in this study were classified as smokers had smoked a couple of cigarettes at the most and smoking status in this study thus mainly implies experimental smoking. In addition, the number of persons who were current regular smokers was low (n = 59). The results of the subanalysis of the social-cognitive factors on lifetime smoking while excluding persons who were current weekly or daily smokers (n = 59) only showed small differences compared with when including these persons. Thus, we have no indications of the inclusion of current regular smokers distorting the results. The current study is cross-sectional and it may be less complicated to investigate these issues in longitudinal studies because there is more control of the timing of social-cognitive factors and smoking status.

Another potential limitation lies in the co-linearity among the social-cognitive factors. Mutually adjusted analyses estimates individual effects of the social-cognitive factors, but due to the co-linearity the estimates may not be interpreted straightforwardly. We conducted unadjusted, confounder adjusted, and mutually adjusted analyses to clarify the influence of this co-linearity. Finally, selection bias might be a potential problem. The fact that 56% of the invited schools did not participate might have led to selection bias, as the nonparticipating schools might have differed from the other schools, for example, with higher smoking rates; however, we were unable to investigate this hypothesis, as we had no information on the nonparticipating schools. As the study population appeared to be representative of the general population in respect of several variables, however, there was no indication of selection bias.

The public health message of this study with regard to prevention of smoking is that social-cognitive factors if their effect is confirmed in longitudinal studies could be targeted, for example, in school-based programs (suggested by the importance of peers), by teaching social skills (to target self-efficacy to resist pressure to smoke) and facts about smoking (to target attitude), and by involving parents in the program (suggested by the importance of parents' behavior). The results for the school factors suggest that the school class environment and gender are important, but more knowledge is required about the underlying mechanisms. It might be relevant to perform screening and then to target prevention initiatives to, for example, the boys in specific classes, at specific schools.

The results suggest that the individual-level factors (social influence, attitude, and self-efficacy to resist pressure to smoke) are important for smoking status, even when the effect of school, school class, and gender group in the school class is taken into account. Social-cognitive factors constitute a complex web, each influencing the other as well as smoking behavior. The results support use of the ASE model as a theoretical framework but emphasize the importance of viewing individuals in their social context, where school and gender play important roles.

No potential conflicts of interest were disclosed.

Grant support: Pharmaceutical Foundation of 1991 (Apoterkerfonden af 1991), Danish Lung Association's Research Foundation (Danmarks Lungeforenings Forskningsfond), and Foundation of 17.12.1981 (Fonden af 17.12.1981).

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.

We thank the Danish National Institute of Public Health for giving us access to data.

1
Ezzati M, Lopez AD. Estimates of global mortality attributable to smoking in 2000.
Lancet
2003
;
362
:
847
–52.
2
British Heart Foundation. Total numbers of deaths and numbers of deaths due to smoking by country, by cause, adults aged 35 and over, by sex, 2000. Europe: British Heart Foundation Statistics; 2006. Available from: http://www.heartstats.org/temp/ESspTabsp4.1spweb05.xls. Accessed 2006 1 May.
3
Holstein B, Due P. The school children study 2002 [in Danish]. Copenhagen: University of Copenhagen, Institute for Public Health; 2003. Available from: http://www.hbsc.dk/rapp-2002.htm. Accessed 2005 Dec 19.
4
Britton J, Bates C, Channer K, Cuthbertson L, Godfrey C, McNeill A. Nicotine addiction in Britain; a report of the Tobacco Advisory Group of the Royal College of Physicians. Royal College of Physicians of London; 2000.
5
Molyneux A, Lewis S, Antoniak M, et al. Prospective study of the effect of exposure to other smokers in high school tutor groups on the risk of incident smoking in adolescence.
Am J Epidemiol
2004
;
159
:
127
–32.
6
Tyas S, Pederson LL. Psychosocial factors related to adolescent smoking: a critical review of the literature.
Tob Control
1998
;
7
:
409
–20.
7
Conrad KM, Flay BR, Hill D. Why children start smoking cigarettes: predictors of onset.
Br J Addict
1992
;
87
:
1711
–24.
8
Kremers S. On your marks. Revising, testing, and integrating stage models of smoking initiation [dissertation]. Maastricht: University of Maastricht; 2002.
9
de Vries H, Kok G. From determinants of smoking behaviour to the implications for a prevention programme.
Health Educ Res
1986
;
1
:
85
–94.
10
de Vries H, Mudde A. Predicting stage transitions for smoking cessation applying the attitude-social influence-efficacy model.
Psychol Health
1998
;
13
:
369
–85.
11
de Vries H, Backbier E, Kok G, Dijkstra M. The impact of social influences in the context of attitude, self-efficacy, intention, and previous behavior as predictors of smoking onset.
J Appl Soc Psychol
1995
;
25
:
237
–57.
12
Bandura A. Self-efficacy. The exercise of control. New York: W.H. Freeman & Co.; 1997.
13
Sellstrom E, Bremberg S. Is there a “school effect” on pupil outcomes? A review of multilevel studies.
J Epidemiol Community Health
2006
;
60
:
149
–55.
14
Aveyard P, Markham WA, Cheng KK. A methodological and substantive review of the evidence that schools cause pupils to smoke.
Soc Sci Med
2004
;
58
:
2253
–65.
15
Johansen A, Rasmussen S, Madsen M. Health behaviour among adolescents in Denmark: influence of school class and individual risk factors.
Scand J Public Health
2006
;
34
:
32
–40.
16
Poulsen LH. Adolescents' smoking: the role of smoking restriction policies and the school and family environments [dissertation]. Copenhagen: Institute of Public Health, Department of Social Medicine, University of Copenhagen; 2005.
17
Rasmussen M. Smoking behaviour of adolescent boys and girls-the influence of school related factors [dissertation]. Copenhagen: Institute of Public Health, University of Copenhagen; 2005.
18
Erwin P. Friendship among children and adolescents [in Danish]. Copenhagen, Denmark: Hans Reitzels Forlag; 2000.
19
Merlo J, Chaix B, Ohlsson H, et al. A brief conceptual tutorial of multilevel analysis in social epidemiology: using measures of clustering in multilevel logistic regression to investigate contextual phenomena.
J Epidemiol Community Health
2006
;
60
:
290
–7.
20
Hansen PE, Siersma V, Ross L, et al. Psychometric properties of brief indexes designed to measure social-cognitive predictors of smoking initiation.
Drug Alcohol Depend
2007
;
88
:
64
–74.
21
Danish Ministry of Education. Statistics on schools in Denmark [in Danish]. Copenhagen: Danish Ministry of Education; 2005. Available from: http://www.uddannelsesstatistik.dk/pls/www_ndb/ndb. Accessed 2005 Dec 19.
22
Anderson S. In a class of its own [in Danish]. Copenhagen: Gyldendal Uddannelse; 2000.
23
Ausems M, Mesters I, van Breukelen G, de Vries H. Do Dutch 11-12 years olds who never smoke, smoke experimentally or smoke regularly have different demographic backgrounds and perceptions of smoking?
Eur J Public Health
2003
;
13
:
160
–7.
24
Currie CE, Elton RA, Todd J, Platt S. Indicators of socioeconomic status for adolescents: the WHO health behaviour in school-aged children survey.
Health Educ Res
1997
;
12
:
385
–97.
25
Mullan E, Currie C, Boyce W, Morgan A, Kalnins I, Holstein B. Focus area rationale: social inequality. In: Currie C, Samdal O, Boyce W, Smith R, editors. Health behaviour in school-aged children: a World Health Organisation cross-national study: HBSC Protocol for 2001/02 survey. Edinburgh: University of Edinburgh, Child and Adolescent Health Research Unit; 2001. p. 175–99.
26
Torsheim T, Currie C, Boyce W, Kalnins I, Overpeck M, Haugland S. Material deprivation and self-rated health: a multilevel study of adolescents from 22 European and North American countries.
Soc Sci Med
2004
;
59
:
1
–12.
27
Larsen K, Petersen JH, Budtz-Jorgensen E, Endahl L. Interpreting parameters in the logistic regression model with random effects.
Biometrics
2000
;
56
:
909
–14.
28
Groenvold M, Petersen MA. The role and use of differential item functioning (DIF) analysis of quality of life data from clinical trials. In: Fayers P, Hays R, editors. Assessing quality of life in clinical trials. Methods and practice. New York: Oxford University Press; 2005. p. 195–208.
29
de Vries H, Dijkstra M, Kuhlman P. Self-efficacy: the third factor besides attitude and subjective norm as a predictor of behavioral intentions.
Health Educ Res
1988
;
3
:
273
–82.
30
Eisenberg ME, Forster JL. Adolescent smoking behavior: measures of social norms.
Am J Prev Med
2003
;
25
:
122
–8.
31
Bricker JB, Peterson AV, Jr., Leroux BG, Andersen MR, Rajan KB, Sarason IG. Prospective prediction of children's smoking transitions: role of parents' and older siblings' smoking.
Addiction
2006
;
101
:
128
–36.
32
Avenevoli S, Merikangas KR. Familial influences on adolescent smoking.
Addiction
2003
;
98
Suppl 1:
1
–20.
33
Hoffman BR, Monge PR, Chou CP, Valente TW. Perceived peer influence and peer selection on adolescent smoking.
Addict Behav
2007
;
32
:
1546
–54.
34
Arnett JJ. The myth of peer influence in adolescent smoking initiation.
Health Educ Behav
2007
;
34
:
594
–607.
35
Piko B. Smoking in adolescence do attitudes matter?
Addict Behav
2001
;
26
:
201
–17.
36
Rugkasa J, Stewart-Knox B, Sittlington J, Abaunza PS, Treacy MP. Hard boys, attractive girls: expressions of gender in young people's conversations on smoking in Northern Ireland.
Health Promot Int
2003
;
18
:
307
–14.
37
Clayton S. Gender differences in psychosocial determinants of adolescent smoking.
J Sch Health
1991
;
61
:
115
–20.
38
Kobus K. Peers and adolescent smoking.
Addiction
2003
;
98
Suppl 1:
S37
–55.
39
Ellickson PL, Bird CE, Orlando M, Klein DJ, McCaffrey DF. Social context and adolescent health behavior: does school-level smoking prevalence affect students' subsequent smoking behavior?
J Health Soc Behav
2003
;
44
:
525
–35.
40
Rosendahl KI, Galanti MR, Gilljam H, Bremberg S, Ahlbom A. School and class environments are differently linked to future smoking among preadolescents.
Prev Med
2002
;
34
:
649
–54.
41
Bandura A. Social cognitive theory of personality. In: Pervin LA, John OP, eds. Handbook of personality: theory and research. New York: Guilford Press; 1999. p. 154–96.
42
Statistics Denmark. News from Statistics Denmark. Subject: population and election. Immigrants, descendants, and foreign citizens [in Danish]. Copenhagen: Statistics Denmark;2005 Jan 1. Available from: http://www.dst.dk/Statistik/Nyt/Emneopdelt.aspx?si=6&msi=2. Accessed 2005 Dec 19.
43
Statistics Denmark. News from Statistics Denmark. Subject: population and election. Households and families [In Danish]. Copenhagen: Statistics Denmark; 2004 24 Feb. Available from: http://www.dst.dk/Statistik/Nyt/Emneopdelt.aspx?si=9&msi=2. Accessed 2005 Dec 19.
44
Kremers SP, Mudde AN, de Vries H. Model of unplanned smoking initiation of children and adolescents: an integrated stage model of smoking behavior.
Prev Med
2004
;
38
:
642
–50.