Results

As expected, analysis of smoking prevalence rates indicated that smoking levels increased substantially from baseline (sixth and seventh graders) to the two year follow up (eighth and ninth graders). At baseline, 15.4% of participants reported that they had ever smoked, 10.6% smoked in the past year, 3.3% smoked in past month, and 1.4% smoked in the past week. At follow up, 30.4% of eighth and ninth graders reported that they had ever smoked, 22.0% reported smoking in the past year, 10.6% reported smoking in the past month, and 5.0% reported smoking in the past week.

ATTRITION ANALYSES

Analyses were conducted to examine differences on the four smoking measures between the panel sample and dropout students at the two year follow up. Dropouts were more likely than the panel sample to smoke in the past month (7.5% v 3.3%, χ² (1) = 21.30, p < 0.001), were more likely to smoke in the past week (3.6% v 1.4%, χ² (1) = 3.05, p < 0.001), smoked more frequently (mean = 1.72 v mean = 1.35, F(1,2374) = 49.33, p < 0.001), and intended to smoke more (mean = 1.60 v mean = 1.41, F(1,2293) = 27.95, p < 0.001). The loss of high end smokers would be expected to attenuate the estimation of the proposed relationships, making significant associations more difficult to identify. Thus, the impact of sample attrition on the findings of this study would be to make them more conservative. Dropouts had lower composite personal efficacy scores (mean = 19.51) than panel students (mean = 20.16; F(1,1933) = 15.11, p < 0.001) but no significant differences in decision making skills. It is common in longitudinal aetiologic studies of smoking behaviour for dropouts to be more likely to be smokers and be more deviant prone, as was found in the work of Chassin and colleagues, for example.42 43 In the current study, we acknowledge that caution is warranted in generalising any findings, in particular to those who were heavier smokers and less efficacious.

CONFIRMATORY FACTOR ANALYSIS

Four latent factors were specified in the CFA or measurement model, each of which contained from three to five indicator items (fig1). The baseline smoking factor had loadings ranging from 0.44 to 0.88, the decision making latent (baseline) factor had loadings ranging from 0.62 to 0.70, and the personal efficacy latent factor (baseline) had loadings ranging from 0.50 to 0.74. The refusal assertiveness latent factor measured one year later had loadings ranging from 0.45 to 0.87. The two year follow up smoking latent factor had loadings ranging from 0.62 to 0.96. Factor loadings for all latent constructs were highly significant (p < 0.001) and in the expected direction. Thus the CFA analysis indicated that the measurement model was properly specified and that each factor was statistically reliable based on the hypothesised model.

• Download figure
• Open in new tab
• Download powerpoint

Figure 1

Confirmatory factor analysis model. Large circles represent latent constructs, rectangles are measured variables5, and small circles with numbers are residual variances.

According to the criteria to evaluate the overall fit of the CFA, the model was a good to excellent fit (χ² (179, n = 1435) = 679, p < 0.001; χ²/df = 3.8; SRMR = 0.04; NFI = 0.93, NNFI = 0.94, CFI = 0.95). Although the χ² p value was significant (which indicates that additional models could be fit to the data) this is not uncommon with large sample sizes.44 45

As shown in table 1, several of the latent factors from the CFA model were significantly intercorrelated. The strongest relationships were between refusal assertiveness and two year smoking (r = −0.32, p < 0.001), while baseline and two year smoking were highly correlated as expected (r = 0.37, p < 0.001). Furthermore, decision making and personal efficacy were moderately associated (r = 0.31, p < 0.001). Thus, the CFA analysis showed that the measurement model was excellent, with high factor loadings for all indicator variables and fit indices in the good to excellent range.

STRUCTURAL EQUATIONS MODELLING

As recommended by MacCallum and his colleagues,46 the first step involved testing a saturated model that includes all possible paths from the exogenous latent factors to the construct of refusal assertiveness and to the two year smoking latent factor, as well as the path from the indirect factor to the outcome. The three exogenous latent factors in the model were baseline smoking, decision making, and personal efficacy, and the covariances among these factors were estimated in the saturated model. The SEM model also included the latent factor of refusal assertiveness, and the outcome latent factor of two year smoking.

Following a test of the saturated model, all non-significant paths were trimmed and the model was retested. The results of testing this final model are illustrated in fig 2. Each of the three baseline exogenous latent factors directly predicted refusal assertiveness: decision making (β = 0.15, p < 0.001) and personal efficacy (β = 0.09, p < 0.01) predicted higher refusal assertiveness, while baseline smoking predicted lower refusal assertiveness (β = −0.20, p < 0.001). Refusal assertiveness predicted less two year smoking (β = −0.26, p < 0.001). Baseline smoking also predicted two year smoking (β = 0.27, p < 0.001), as expected. In terms of goodness-of-fit indices, there was an excellent fit of the model to the data (χ² (178, n = 1435) = 537, p < 0.001; χ²/df = 3.0; SRMR = 0.04; NFI = 0.95, NNFI = 0.96, CFI = 0.96). In summary, the findings shown in fig 2indicate that each of the three exogenous latent factors significantly predicted refusal assertiveness, which in turn predicted two year smoking.

• Download figure
• Open in new tab
• Download powerpoint

Figure 2

Structural equation model. Large circles represent latent factors and small circles with numbers reflect residual variances.

Discussion

The current study was designed to increase our understanding of the aetiology of cigarette smoking among inner city youth and provide additional information to guide efforts to refine prevention approaches. The findings from this longitudinal study highlight the importance of both generic and smoking specific aetiologic factors. These findings suggest that smoking prevention programmes should include components to enhance broader competence in addition to providing training in refusal skills. This is because youth with better personal efficacy and decision making skills are better able to implement smoking refusal strategies, which in turn reduce smoking behaviour. This is an important finding because research that only examined direct effects of sound decision making and high personal efficacy as deterrents to adolescent smoking have been mixed with some showing a relationship7-9 and others showing no relationship.10 11 Yet in the current tested model general competence, as represented by decision making and personal efficacy, served as a critical foundation for refusal skills. Although refusal skills training requires a briefer investment of time, this study implies that the additional time and effort to teach general competence skills in addition to refusal skills may be very worthwhile.

According to these findings, adolescents whose decision making skills were deficient were less likely to use refusal skills and more likely to succumb to pro-smoking social influences. These adolescents might not have been able to weigh adequately the consequences of peer pressure to smoke or compromise by being assertive about not smoking without offending their friends. For example, they may have decided to just smoke one cigarette to fit in with their friends without evaluating the long term consequences (for example, their friends will expect them to smoke again next time). Such faulty decision making meant less frequent use of refusal skills and a greater tendency to smoke. In contrast, adolescents with adequate decision making skills may have been more likely able to weigh the consequences of succumbing to peer pressure and realise they could most appropriately respond by utilising assertive refusal skills. Thinking through the consequences and making use of other sound decision making practices would facilitate frequent use of refusal skills decreasing subsequent smoking. In terms of personal efficacy, adolescents who believed in their own ability to perform behaviours appeared to be more prepared to engage in refusal skills that minimised later smoking. Knowledge of refusal skills coupled with the belief in being able to actually use these skills seemed to be a fundamental combination in diminishing subsequent smoking.

Several limitations of the current study should be noted. Since only a limited number of variables were included, the tested model is not the only possible good fitting model of smoking behaviour among inner city minority adolescents. Even so, the model is parsimonious and consistent with prior prevention research conducted with inner city minority adolescents.22-25 Because adolescents participating in this study were students, these results cannot be generalised to non-student populations. Yet, the middle school period has low dropout rates and absentees were pursued on two return data collections, lessening the detrimental impact relative to conducting such a study in high school and without pursuit of absentees. As the study was only conducted with inner city students, it is possible that different results might be obtained with suburban or rural populations. Finally, the loss of high end smokers from the panel sample may have attenuated the estimation of variable relationships and may mean the findings cannot be generalised to adolescents who are heavier smokers.

Although smoking prevention programmes that focused on refusal skills training have proven effective,47-51 this prevention approach can be further improved through the addition of other personal and social skills training. Specifically, inclusion of broader competence skills to enhance personal efficacy and teach decision making and problem solving skills appears warranted by these findings. The current study showed the importance of more general competence factors in enhancing refusal assertiveness and deterring adolescent smoking. Consequently, the tested model provided support for the competence enhancement approach to smoking prevention over the refusal skills training only approach.

Refusal training approaches teach adolescents information, norms, and refusal skills with a problem specific focus. But this type of approach assumes that adolescents are naturally motivated to resist pro-smoking influences. In contrast, the competence enhancement approach to smoking prevention recognises that training in a broader array of personal and social skills helps reduce the motivation to smoke and does not make the assumption that adolescents want to avoid social influences to smoke. Competence enhancement approaches include both aspects of a refusal skills programme (that is, training in handling social influences to smoke) and a more comprehensive personal and social skills training programme (setting and achieving goals, decision making and problem solving, coping strategies for stress and anxiety, communicating effectively, developing friendships, and general assertiveness). Learning this comprehensive array of skills is designed to boost personal efficacy. Moreover, these skills can be applied more generally to a variety of challenges that adolescents face including, but not limited to, smoking. The link between each general competence measure with refusal assertiveness in the present study suggests that decision making and personal efficacy motivated the use of refusal skills.

Reviews of the smoking prevention literature indicate that the competence enhancement approaches have proven long term effectiveness, as well as short term effectiveness, and generalisability to ethnic minority groups.3 12 13 For example, a competence enhancement programme delivered during junior high school reduced pack-a-day smoking at the end of high school by 25% compared to a control group according to a large, long term study.16This body of research combined with the current study provides empirical support for a smoking prevention approach that can be implemented in a school setting and has a dual emphasis on resistance skills training and more general personal and social skills training. Moreover, the results of the current study underscores the utility of such an approach for inner city, minority adolescents. Future research is needed to determine the effectiveness of this prevention approach with other populations as well as to determine the relative importance of the various personal and social skills taught in this type of prevention approach.