Health Education Research Advance Access originally published online on June 28, 2006
Health Education Research 2007 22(1):81-94; doi:10.1093/her/cyl050
Association of substance-use behaviours and their social-cognitive determinants in secondary school students
1 Research Group for Stress, Health, and Well-being, University of Leuven, Leuven, Belgium
2 Cooperative Liaison University of Leuven—Free Centres for Educational Guidance, Cooperative Liaison, Brussels, Belgium
* Correspondence to: A. Victoir. E-mail: An.Victoir{at}psy.kuleuven.be
 |
Abstract |
|---|
 Top Abstract IntroductionAim and hypothesesMethodsResultsDiscussionConflict of interest statementAppendixReferences |
|---|
In two samples of Flemish secondary school students, co-occurrence of different types of substance use was observed: smoking was associated with marijuana use in Sample 1 (n = 597) and alcohol consumption in Sample 2 (n = 403). It was investigated whether social-cognitive determinants of these behaviours were also associated. Low to medium correlations were observed. Confirmatory factor analyses showed that a model with general social-cognitive factors (across different substances) did not have adequate fit. Substance use was mainly associated with variables referring to the specific substance under consideration, with the exception of self-efficacy in buying and smoking cigarettes; this factor was linked not only to smoking but also to alcohol and marijuana use. Adolescents who regularly used two substances generally held positions on social-cognitive scales that were more unfavourable than those who only used one substance. In order to change determinants of use, substance-specific cognitions and skills may be important targets.
|
Introduction |
|---|
|
TopAbstractIntroductionAim and hypothesesMethodsResultsDiscussionConflict of interest statementAppendixReferences |
|---|
Health-related behaviours are often studied in isolation. For example, adolescents' smoking behaviour is linked to demographic, social-cognitive and environmental factors in order to construct models and plan interventions. However, in real life, behaviours do not occur in isolation. They tend to cluster: the number of people combining two or more behaviours exceeds the number expected from the prevalence of those behaviours. Exercise is linked to satisfactory sleep time [1], better diets and less smoking [2]. Alcohol consumption correlates with smoking [3] and illicit drug use [4, 5]. An increase in the use of one substance is often paralleled by higher use of other substances [6, 7].
Health-affecting behaviours appear to cluster in a health-enhancing and a health-threatening group [8, 9]. In both sets, further clusters may emerge. The debate on the number and nature of dimensions is ongoing. Both one- and multidimensional representations seem acceptable [10].
Use of multiple substances, sometimes termed poly-use, is perhaps the most consistent finding. There are several theories that can account for the clustering of substance-use behaviours. On a biochemical level, frequent use of one substance may alter the dopamine system, and hence the reinforcement value of substances [11]. On a psychological level, a positive evaluation of a used substance may generalize to other substances, including those not consumed before. Adolescents who smoke and drink regularly have more positive attitudes towards illicit drugs and higher odds of using them than non-smokers who do not drink [12]. Factors like family cohesion and friends condoning substance use seem to have similar influences on adolescents' careers in smoking, alcohol and marijuana use [6].
Different forms of substance use may also have a common function. Both alcohol and cannabis can counter feelings of depression, and help to manage the after-effects of other drugs [13]. Drinking, marijuana use and delinquent behaviours could all serve the function of ‘maturity landmarks’, or allow the adolescent to break societal norms [14, 15]. Sensation seekers derive positive consequences from new experiences. For some, high-risk sports provide excitement; others are inclined to experiment with psychotropic substances [16–18].
Finally, it is likely that substance-use behaviours co-occur because they occur in the same context. In many pubs, people drink alcohol and smoke.
Expectancy-value theories, on the other hand, state that the immediate determinants of use are substance specific. For example, according to the theory of planned behaviour (TPB) [19, 20] and related models, marijuana use is predicted by thoughts and feelings concerning marijuana use. If an adolescent happens to use a second substance (e.g. alcohol), intra- and interpersonal factors specific to alcohol use are the most important variables. From studies on the TPB, it was estimated that 
41% of the variance in intentions and 34% of the variance in behaviour are explained by behaviour-specific variables [21].
A focus on behaviour-specific determinants is not necessarily at odds with frameworks in which a common aspect of smoking, drinking and illicit drug use is the main explanatory factor. Personality factors and other distal variables are supposed to act through behaviour-specific cognitions [22].
However, substance-specific determinants will relate to each other differently depending on whether or not a superordinate factor is presupposed. If a ‘general’ factor underlies substance-specific cognitions, measures that operationalize the same construct (e.g. attitudes) but refer to different substances should correlate highly. If determinants are entirely substance specific, there should be—on average—no significant correlation between, for example, attitudes towards alcohol and attitudes towards smoking.
Studies of the relationships between determinants referring to different behaviours are scarce. From studies on stages of change for smoking and exercise, it appears that people who experience many benefits from smoking see more disadvantages to exercise. Those who think that smoking harms them are convinced that exercise would be beneficial. Self-efficacy to quit smoking is positively correlated with self-efficacy to be physically active [23, 24]. Attitudes, subjective norms, perceived behavioural control and intentions of energy balance behaviours correlate more strongly than the behaviours themselves [25]. In conclusion, correlations seem to occur both at the behavioural level and at the level of behaviour-specific determinants.
|
Aim and hypotheses |
|---|
|
TopAbstractIntroductionAim and hypothesesMethodsResultsDiscussionConflict of interest statementAppendixReferences |
|---|
The aim was to investigate associations on three levels: between substance-use behaviours, between determinants of these behaviours and between behaviours and determinants.
We expected that different types of substance use would show positive associations in samples of secondary school students. More specifically, smoking was expected to correlate with illicit drug use (Sample 1) and alcohol use (Sample 2).
Second, we wanted to investigate associations between determinants of these forms of substance use. If adolescents' cognitions are substance specific, as implied by the TPB and similar models, positive correlations between conceptually similar cognitions (e.g. attitudes) that refer to different substances (e.g. alcohol versus cigarettes) should only be observed in multi-users (using both substances) and non-users (using neither substance), not in selective users (using one substance only). If, on the other hand, high correlations are observed for selective users as well, it would mean that thoughts and feelings adolescents have with regard to a substance they have used tend to generalize to substances they have not tried before.
Finally, from the TPB, it can be expected that use of a substance would be explained by variables referring to that particular substance. Smoking should be explained by people's attitudes towards smoking, their ability to refuse a smoke, their intention to smoke, et cetera. Their attitudes towards drinking or their ability to refuse a drink would not be expected to explain their smoking behaviour. However, if determinants of use are not entirely substance specific (cf. Research Question 2), the use of one substance might be explained by variables referring to another substance. Again, the pattern of associations observed in selective users would indicate to what extent determinants are substance specific.
|
Methods |
|---|
|
TopAbstractIntroductionAim and hypothesesMethodsResultsDiscussionConflict of interest statementAppendixReferences |
|---|
Procedure
Data were collected from two independent samples of secondary school students, both providing data on two behaviours (smoking and illicit drug use in Sample 1 and smoking and alcohol use in Sample 2) via pencil-and-paper questionnaires. Respondents were recruited via convenience sampling. The data were gathered for the validation of a diagnostic instrument on health behaviours and their determinants, aimed at identifying priorities in health promotion planning [26]. Schools were invited to participate by health counsellors from student guidance centres, and received behavioural and determinant analyses reports in return for their cooperation. The school boards decided which questionnaires were to be distributed among which number of students. In the first school, which offers only technical and vocational courses, all students (Grades 1–7) received questionnaires on illicit drug use and smoking. In the second school, offering general, technical and vocational courses, students from the second, fourth and sixth grades received questionnaires on alcohol use and smoking. There were no schools in which all three types of substance-use behaviours were examined. Cooperation of students was voluntary; no consent was obtained from parents. Data were gathered in collective sessions supervised by the health counsellors. Teachers did not attend, unless their presence was required for practical reasons. Responses were collected anonymously; neither teachers nor counsellors had access to data. Filling out the questionnaires required
1 hour for participants >15 years of age and an hour and a half for younger students.
Participants
In total, 1011 pairs of questionnaires were collected. Response rates were estimated by health counsellors and school personnel to exceed 90%. No information was available concerning non-responders.
Drug use/smoking sample
Data on determinants of illicit drug use and smoking were collected from 604 respondents (464 girls and 130 boys; 10 respondents did not indicate the gender). Ages ranged from 13 to 22 years (M = 16.3 years, SD = 1.99). Respondents were enrolled in the same school, either in technical (n = 412) or in vocational (n = 183) secondary education; two respondents did not indicate what type of education they were enrolled in.
Alcohol use/smoking sample
In the second school, 407 respondents provided data on determinants of alcohol consumption and smoking (129 boys and 274 girls). Ages ranged from 13 to 21 years (M = 16.1 years, SD = 1.79). Respondents were enrolled in general (n = 204), technical (n = 121) or vocational (n = 76) education; two respondents did not indicate the type of education.
Representativeness of the samples
In Flanders, most adolescents (98.3%) are enrolled in general (relative proportion of 40.1%), technical (32.1%) or vocational (27.8%) courses, with the remaining 1.7% following language courses for non-native speakers or art classes [27]. Students from technical courses (53.3% over both samples) are over-represented in this study, while students from general courses (20.4%) are under-represented; the proportion of students from vocational courses (25.9%) approaches the proportion observed in the Flemish Community.
Respondents aged 19 years or older accounted for 15.1% (n = 76) of the drug use/smoking sample and 8.9% (n = 35) of the alcohol use/smoking sample. These respondents were mainly recruited in technical and vocational courses, some of which last for 7 years (as opposed to the usual 6). Approximately 10% of the students in these courses have doubled one or more years [28]. This explains the high percentage of students >18 in our samples.
In both schools, technical and vocational courses prepared students for professions in health care, fashion, bodily care and office management, which may explain the over-representation of female students.
All respondents lived in the small provincial towns where the schools were located or in the surrounding countryside.
Measures
Substance use
Current smoking was measured by one item. Respondents indicated if they considered themselves non-smokers or monthly (smokes at least once a month, but not every week), weekly (smokes at least once a week, but not daily) or daily smokers. Alcohol consumption was measured by means of 12 items: number of week days (one item) and weekend days (one) on which the respondent usually consumed alcohol, number of alcohol drinks consumed on week (five) and weekend days (five). Illicit drug use was measured by means of five items, referring to marijuana, heroin/cocaine, LSD, XTC/amphetamines and other drugs. Respondents indicated whether they had used the substance in the past (but not in the last month), or in the last month. It should be noted that in Flanders, alcohol consumption and smoking is legal for adolescents >16. Marijuana possession is illegal, but is not prosecuted.
Social-cognitive measures
The determinants of use were measured with a questionnaire conceptually based on the TPB, including measures of attitudes, prescriptive norms and intentions. Questions on self-efficacy, i.e. ease/difficulty of performing actions, were included, rather than direct measures of perceived control. Ease/difficulty assessments are more closely linked to behaviours and intentions possibly because respondents see control as a binary (on/off) state [29]. In accordance with the ASE model [30], descriptive norms were also included.
For attitudinal beliefs and prescriptive norms, weighing factors (importance and motivation to comply, respectively) were included in order to compose multiplicative measures. However, data inspection revealed that 17% of the respondents in the smoking/drug-use sample did not fill in the ‘importance’ measure. Using the multiplicative composites would result in a loss of power. Therefore, item scores on perceived consequences were totalled. Similarly, raw scores for norms were used. As unweighted scores may predict intentions as successfully as composites [31, 32], this operationalization was considered to be the optimal one under the given circumstances.
Non-discriminative items were eliminated and principal component analysis was performed to assess construct validity. The pattern of factor loadings led to the construction of multiple scales for some concepts (e.g. three dimensions for attitude towards smoking) and aggregated scales for related concepts if factor loadings suggested a single dimension (e.g. descriptive and prescriptive norms regarding drug use were aggregated in a ‘social influence’ scale). All scales had acceptable internal consistency and are summarized in the Appendix (Table AI).
Analyses
Associations between behaviours
Associations between smoking and illicit drug use and smoking and alcohol use were assessed with chi-square analyses.
Associations between determinants of substance-use behaviours
These associations were examined via correlational and confirmatory factor analyses. For the latter, determinant measures were modelled as indicators of four general latent factors: attitudes, self-efficacy, social influences and intentions. Correlations between intentions and the other three factors were allowed, as implied by the TPB. If this model proved unsatisfactory, it was investigated whether the loadings of the measures on the latent factors were higher in multi- and non-users than in selective users, by computing the relative fit increment of a model with free estimates for all parameters over a model with equal factor loadings for all groups.
Associations between determinants and behaviours
These associations were examined via logistic regression. In the first model (I-SE model), substance use was linked to intentions and self-efficacy and in the second model (AT-SOC-SE model) to attitudes, social influences and self-efficacy. These models reflect the two tiers of determinants in the TPB, with intentions and actual control (for which self-efficacy is a proxy) as most proximal to behaviour. In all models, determinants referring to two substances were used to predict the observed substance use. For example, in the AT-SOC-SE model for alcohol use and smoking, eight predictors referred to alcohol use (health concerns regarding alcohol consumption, sensory appeal of alcoholic drinks, social benefits of drinking, norms of friends and parents regarding alcohol consumption, parents' and friends' alcohol consumption and self-efficacy in refusing drinks) and seven predictors referred to smoking (health concerns regarding smoking, social benefits of smoking, social penalties, parents' and friends' influences on smoking, self-efficacy in refusing a smoke and self-efficacy in buying/smoking cigarettes). These predictors were used to calculate the odds of belonging to a particular user group (multi-user or selective user), rather than to the non-user group. In order for these tests to have sufficient power, missing values on social-cognitive items were substituted by means on the conditions that (i) the respondent indicated a valid answer on at least 70% of the questions and (ii) at least 90% of the respondents indicated an answer. From the original number of 407 alcohol/smoking and 604 drugs/smoking questionnaires, four (0.9%) and seven (1.2%), respectively, were discarded due to insufficient data. No missing value substitutions were made for items measuring substance use. All predictors were standardized.
Profiles of user groups
Differences between respondents with no, selective or combined substance use were assessed with chi-square tests and analyses of variance (ANOVAs).
|
Results |
|---|
|
TopAbstractIntroductionAim and hypothesesMethodsResultsDiscussionConflict of interest statementAppendixReferences |
|---|
Association of substance-use behaviours
Smoking and marijuana use
In the first sample, 123 respondents (21%) indicated they had used marijuana, with 47 respondents (7.9%) using this substance in the last month. Few respondents reported past use of cocaine/heroin (n = 14, 2.4%), LSD (n = 12, 2.1%), XTC (n = 26, 4.5%) or other drugs (n = 18, 3.1%). Recent use of cocaine/heroin and LSD was reported by 3 respondents (0.5%), recent XTC use by 13 respondents (2.2%) and recent use of other drugs by 8 respondents (1.4%). All but six users of these drugs were also using marijuana. Because of this redundancy and the low prevalence of illicit drug use other than marijuana, it was decided to consider only marijuana use in further analyses. Current daily smoking was reported by 142 respondents (23.8%), weekly smoking by 53 respondents (8.9%) and monthly smoking by 62 respondents (10.4%); 338 respondents indicated they were non-smokers at the time of the study (56.6%).
Smoking and marijuana use were associated (
2(6, N = 586) = 145.91, P < 0.001). The strength of association is partly explained by the large proportion of respondents using neither substance (Table I, upper part). However, cross-tabulation of smoking and marijuana use for respondents who did use at least one substance also revealed a significant association (2(6, N = 217) = 209.83, P < 0.001).
|
Respondents were allocated to one of four groups: non-users (non-smokers who had never used marijuana), selective smokers (respondents who smoked but had never used marijuana), selective marijuana users (non-smokers who used marijuana in the previous month) and multi-users (weekly/daily smokers who had used marijuana in the previous month).
Univariate ANOVA showed that non-users (M = 15.8 years, SD = 2.04) were younger than selective smokers (M = 16.5 years, SD = 1.83) and multi-users (M = 17.0 years, SD = 1.51). The proportion of multi-users in vocational education was almost twice the proportion of multi-users in technical education (11.9 versus 6.4%, 2(3, N = 527) = 10.37, P < 0.05). There was also a gender difference (2(3, N = 526) = 8.83, P < 0.05), with a larger proportion of boys than girls in the multi-user (12.9 versus 6.8%) and selective marijuana user (6.9 versus 3.4%) groups.
Smoking and alcohol use
In the second sample, 121 respondents reported never drinking alcohol (30.0%). The majority of pupils who did drink did so once (n = 136, 33.7%) or twice (n = 77, 19.1%) a week. On average, respondents had 4.9 drinks (SD = 3.35) on days they consumed alcohol, with those drinking on more days reporting more drinks per occasion (r = 0.35, n = 277, P < 0.001). Daily smoking was reported by 78 respondents (19.3%), weekly smoking by 26 respondents (6.4%), and monthly smoking by 36 respondents (8.9%); 263 respondents (65.3%) were non-smokers.
Smoking and frequency of alcohol consumption were associated (2(21, N = 401) = 168.65, P < 0.001; Table I, lower part). Respondents were allocated to one of four groups: non-users (non-smokers who never drank alcohol), selective alcohol users (non-smokers who consumed alcohol at least once a week), selective smokers (weekly/daily smokers who never drank) and multi-users (weekly/daily smokers who drank alcohol at least once a week). Multi-users (M = 16.9 years, SD = 1.73) were older than selective alcohol users (M = 16.2 years, SD = 1.68), who were in turn older than non-users (M = 15.1 years, SD = 1.58). The proportion of boys and girls was similar in the four groups (2(3, N = 368) = 0.80, P = 0.87). The proportion of multi-users was lower in general (19.8%) than in technical (31.5%) and vocational (34.7%) classes (2(6, N = 367) = 12.60, P < 0.05).
Associations of behavioural determinants across substances
Smoking and marijuana use
Correlations between responses that measured the same concept for different substances ranged from 0.01 to 0.53 in the total sample (Table II). Only one large size correlation (r > 0.5 [33]) was observed, between friends' influence on drug use and their influence on smoking. A confirmatory factor analysis suggested that items referring to smoking and those referring to illicit drug use could not be considered indicators of general attitudes, social influences, self-efficacy or intentions (2(51)= 915.08, P < 0.001, Adjusted Goodness of Fit Index (AGFI) = 0.690, Root Mean Square Error of Approximation (RMSEA) = 0.168).
|
When the sample was split in multi/non-users versus selective users, smoking intentions were more strongly related to drug-use intentions in the multi/non-user group (r = 0.78) than in the selective user group (r = –0.11). Other correlations did not differ in magnitude between groups. A confirmatory factor model in which factor loadings for selective users were allowed to differ from those for multi/non-users was inspected. It fit the data better than a constrained model with identical factor loadings across the groups (
(102) = 722.44, P < 0.001, RMSEA = 0.151; 
(114) = 801.51, P < 0.001, RMSEA = 0.143; 
(12) = 79.07, P < 0.001), but neither model had optimal fit.
Smoking and alcohol use
In the second sample, correlations were again small to moderate for the total sample (Table III). The confirmatory factor analysis revealed that smoking items and alcohol items were unlikely to refer to common latent factors (2(116) = 924.33, P < 0.001, AGFI = 0.697, RMSEA = 0.132).
|
In this sample, the bivariate correlations varied more across groups. Attitudes, social influences, self-efficacy and intentions referring to smoking correlated more strongly with their counterparts referring to alcohol use in multi/non-users than in selective users. Again, a confirmatory model that allowed for group-specific loadings had a better fit than the one in which loadings were constrained to be equal for both groups (
(234) = 1017.61, P < 0.001, RMSEA = 0.132; (249) = 1100.42, P < 0.001, RMSEA = 0.141; (15) = 82.43, P < 0.001).
Association of substance-use behaviour and substance-specific determinants
Smoking and marijuana use
User group was regressed on social-cognitive responses in two separate models, reflecting the two tiers of determinants in the TPB (Table IV). Low values represent ‘healthy’ responses on all scales. For example, a permissive norm on illegal substance use received a high score and a norm favouring abstinence a low score. Main effects of age, educational type and gender were included in all analyses. The results described here refer to a model containing all demographic and social-cognitive predictors. Only significant associations are described in detail.
|
Selective smokers had high smoking intentions, believed that buying/smoking cigarettes is easy, saw many social benefits and few penalties to smoking and had friends who favoured smoking. One drug-related variable was associated with selective smoking: respondents with a favourable attitude towards drugs were ‘less’ likely to be selective smokers. Selective marijuana users had high drug-use intentions, positive attitudes towards drugs and perceived social influences favouring drug use. Two smoking-related variables (seeing few health-adverse consequences to smoking and believing that buying/smoking cigarettes is easy) were also related to selective marijuana use. Multi-use was predicted by both drugs- and smoking-related responses.
Smoking and alcohol use
Selective smokers had high smoking intentions. They also believed that drinking alcohol has social benefits. Selective alcohol users indicated that alcoholic drinks have high sensory appeal. They found it difficult to refuse a drink, and had high intentions to drink alcohol. Two smoking-specific variables were significant for this group. Respondents who believed that refusing a smoke is difficult were less likely to be selective alcohol users. Second, selective alcohol users believed that buying/smoking cigarettes is easy. Both smoking- and alcohol-related variables were significant in multi-users (Table V).
|
Due to the small number of selective smokers (n = 10), regression weights for this group may be less reliable. However, an analysis with a three-level dependent variable (multi-users, selective alcohol users and the reference group of non-users) revealed identical results, indicating that the estimates for these groups were not affected by the inclusion of the selective smokers.
User profiles of behavioural determinants
Smoking and marijuana use
A multivariate ANOVA with gender, education type and user group as independent variables; age as covariate and social-cognitive responses as dependent variables revealed significant main effects for gender (Wilks' lambda = 0.937, F(12, 383) = 2.13, P < 0.05) and user group (Wilks' lambda = 0.282, F(36, 1132.34) = 16.80, P < 0.001) and an interaction between education type and user group (Wilks' lambda = 0.862, F(36, 1132.34) = 1.61, P < 0.05). Because the number of respondents per cell was low for some combinations of the independent variables, least squares means for user groups (adjusted for age) were inspected. The groups differed on all determinants of smoking and illicit drug use, except on parents' influences on smoking (Table VI). In general, multi-users had the most ‘unfavourable’ profile.
|
Smoking and alcohol use
Significant effects of age (Wilks' lambda = 0.647, F(17, 271) = 8.67, P < 0.001), education type (Wilks' lambda = 0.795, F(34, 542) = 1.94, P < 0.01) and user group (Wilks' lambda = 0.171, F(51, 807.61) = 12.80, P < 0.001) were observed. No interaction effects were tested because of the limited size of the selective smoker group. Again, multi-users generally had the most unfavourable scores (Table VII).
|
|
Discussion |
|---|
|
TopAbstractIntroductionAim and hypothesesMethodsResultsDiscussionConflict of interest statementAppendixReferences |
|---|
Smoking was related to marijuana and alcohol use in two samples of secondary school students. Correlations between proximal determinants of these behaviours were also observed. These could point to shared latent factors, such as a ‘general attitude towards substances’ or a common function of different forms of substance use. However, the magnitude of the observed correlations suggested that shared latent factors would not account for all the variance in the data. Indeed, confirmatory factor analyses showed that a model with ‘substance-free’ superordinate factors did not fit the observed data pattern.
Therefore, it was tested whether a ‘substance-specific’ model would hold. In such a model, adolescents' thoughts and feelings regarding illicit drug use would reveal nothing about the way they thought and felt about smoking cigarettes, for example. Adolescents may have a favourable or unfavourable view of both substances, in which case one would expect them to use both or neither, respectively. There may, however, also be a considerable number of adolescents who have a favourable view of one substance and not of the other. In that case, one would expect them to limit their use to the substance favoured. This pattern, however, was only partially confirmed. Correlations between conceptually related determinants tended to be higher for adolescents who used both or neither substance and lower for selective users, but the magnitude of the differences was insufficient to conclude that determinants of different types of substance use are generally independent.
Nevertheless, the user groups revealed interesting patterns. Determinants referring to a particular substance were associated with the use of that substance, after controlling for other variables (including those referring to different drugs). This suggests that substance-specific cognitions are important, even if some variability in use is explained by general factors. There were exceptions to the substance-specific match of cognitions and behaviour. Self-efficacy in smoking and buying cigarettes predicted smoking, but it also predicted selective marijuana and alcohol use. On the basis of the present data, it is difficult to explain why this determinant seemed to have an overall effect. Possibly, self-proclaimed skills in smoking and buying cigarettes, even in the face of negative consequences, could be seen as norm breaking or even sensation seeking. That would explain its pervasive influence. However, as we have no independent indication that these adolescents want to break with societal norms or were sensation seekers, this explanation is tentative.
When looking at mean responses on the social-cognitive variables, it was apparent that multi-users held the most ‘unhealthy’ position, whereas non-users gave the ‘healthiest’ answers. For whatever reason (biological, psychological, learning effects), multi-use does seem to exacerbate users' unfavourable position on substance-specific continua.
To summarize, the co-occurrence of substance-use behaviours, i.e. the ‘larger-than-expected’ proportion of multi-users, might be explained by substance-specific determinants, next to factors that influence all kinds of substance use. Further research can elucidate whether the latter are mediated by the former, as is the basic tenet of the theory of triadic influence [34].
The results suggest that improving adolescents' general' skills (e.g. refusal skills) or cognitions (e.g. evaluations of ‘drugs’) in different contexts (type of substance, location, and time frame [35]) might be an effective route to prevention. What an adolescent has learned regarding one substance may not easily transfer to another. It also seems advisable to target different user groups: smokers may derive little benefit from sessions that mainly deal with illicit drug or alcohol use. Multi-users may be the most difficult group to approach. Their cognitions are most unfavourable. Their health risk might not decrease significantly, unless strategies aimed at decreasing different types of substance use are part and parcel of the intervention. Initiating cessation of more than one form of substance use might be difficult [36], so phasing might be necessary.
The present study had limitations. Within-group variability may be large. The definitions of selective and multiple use were basic; no distinctions were made between regular and problematic use. Users may have been on the verge of quitting. Alternatively, adolescents who did not use a substance may have contemplated its use. As we had no indication of the stage of change occupied by the respondents, it was impossible to make finer within-group discriminations. It must also be noted that the respondents' age range was wide and that girls were over-represented in both samples. The influence of intra- and interpersonal determinants of use may wax and wane with adolescents' developmental status [37], so stratification by age group might reveal different patterns of association. A re-analysis of our data, excluding adolescents >18 (results were not reported) did not lead to appreciable differences, but the data set did not allow for more detailed inspection of different age groups. Because girls use substances less frequently than boys do [38], the proportion of non-users in our samples is likely to be larger than the Flemish proportion.
All measures were concurrent self-reports obtained via questionnaires. Measures across substances were conceptually identical, but operationalized differently. This could have lowered correlations, but it did protect against artificially large correlations based solely on word format. Finally, distal variables were not included. A longitudinal design incorporating substance-specific and general determinants of use would clarify the prospective role of both types of influence.
|
Conflict of interest statement |
|---|
|
TopAbstractIntroductionAim and hypothesesMethodsResultsDiscussionConflict of interest statementAppendixReferences |
|---|
None declared.
|
Appendix |
|---|
|
TopAbstractIntroductionAim and hypothesesMethodsResultsDiscussionConflict of interest statementAppendixReferences |
|---|
|
|
References |
|---|
|
TopAbstractIntroductionAim and hypothesesMethodsResultsDiscussionConflict of interest statementAppendixReferences |
|---|
1. Steptoe A, Wardle J, Fuller R, et al. Leisure-time physical exercise: prevalence, attitudinal correlates, and behavioral correlates among young Europeans from 21 countries. Prev Med 1997 26:845–54.[CrossRef][Web of Science][Medline]
2. Coulson NS, Eiser C, Eiser JR. Diet, smoking and exercise: interrelationships between adolescent health behaviours. Child Care Health Dev 1997 23:207–16.[Web of Science][Medline]
3. Schuit AJ, van Loon JM, Tijhuis M, et al. Clustering of lifestyle risk factors in a general adult population. Prev Med 2002 35:219–24.[CrossRef][Web of Science][Medline]
4. Sutherland I and Willner P. Patterns of alcohol, cigarette and illegal drug use in English adolescents. Addiction 1998 93:1199–208.[CrossRef][Web of Science][Medline]
5. Lewinsohn PM, Rohde P, Brown RA. Level of current and past adolescent cigarette smoking as predictors of future substance use disorders in young adulthood. Addiction 1999 94:913–21.[CrossRef][Web of Science][Medline]
6. Duncan TE, Tildesley E, Duncan SC, et al. The consistency of family and peer influences on the development of substance use in adolescence. Addiction 1995 90:1647–60.[CrossRef][Web of Science][Medline]
7. Everett SA, Giovino GA, Warren CW, et al. Other substance use among high school students who use tobacco. J Adolesc Health 1998 23:289–96.[CrossRef][Web of Science][Medline]
8. Aarø LE, Laberg JC, Wold B. Health behaviours among adolescents: towards a hypothesis of two dimensions. Health Educ Res 1995 10:83–93.
9. Ingledew DK, Hardy L, Cooper CL. Latent class analysis applied to health behaviours. Pers Individ Dif 1995 19:13–20.
10. Røysamb E, Rise J, Kraft P. On the structure and dimensionality of health related behaviour in adolescents. Psychol Health 1997 12:437–52.
11. Koob GF and Le Moal M. Drug abuse: hedonic homeostatic dysregulation. Science 1997 278:52–8.
12. Best D, Rawaf S, Rowley J, et al. Drinking and smoking as concurrent predictors of illicit drug use and positive drug attitudes in adolescents. Drug Alcohol Depend 2000 60:319–21.[CrossRef][Web of Science][Medline]
13. Boys A, Marsden J, Strang J. Understanding reasons for drug use amongst young people: a functional perspective. Health Educ Res 2001 16:457–69.
14. Jessor R and Jessor SL. Problem Behavior and Psychosocial Development: A Longitudinal Study of Youth.New York: Academic Press 1977.
15. Jessor R. Problem-behavior theory, psychosocial development, and adolescent problem drinking. Br J Addict 1987 82:331–42.[CrossRef][Web of Science][Medline]
16. Greene K, Krcmar M, Walters LH, et al. Targeting adolescent risk-taking behaviors: the contributions of egocentrism and sensation-seeking. J Adolesc 2000 23:439–61.[CrossRef][Web of Science][Medline]
17. Kopstein AN, Crum RM, Celentano DD, et al. Sensation seeking needs among 8th and 11th graders: characteristics associated with cigarette and marijuana use. Drug Alcohol Depend 2001 62:195–203.[CrossRef][Web of Science][Medline]
18. Pedersen W. Mental health, sensation seeking and drug use patterns: a longitudinal study. Br J Addict 1991 86:195–204.[CrossRef][Web of Science][Medline]
19. Ajzen I. The theory of planned behavior. Organ Behav Hum Decis Process 1991 50:179–211.[CrossRef][Web of Science]
20. Madden TJ, Scholder Ellen P, Ajzen I. A comparison of the theory of planned behavior and the theory of reasoned action. Pers Soc Psychol Bull 1992 18:3–9.[Abstract]
21. Godin G and Kok G. The theory of planned behavior: a review of its applications to health-related behaviors. Am J Health Promot 1996 11:87–98.[Web of Science][Medline]
22. Sutton S. Social-psychological approaches to understanding addictive behaviours: attitude-behaviour and decision-making models. Br J Addict 1987 82:355–70.[CrossRef][Web of Science][Medline]
23. Boudreaux ED, Francis JL, Carmack Taylor CL, et al. Changing multiple health behaviors: smoking and exercise. Prev Med 2003 36:471–8.[CrossRef][Web of Science][Medline]
24. King TK, Marcus BH, Pinto BM, et al. Cognitive-behavioral mediators of changing multiple behaviors: smoking and a sedentary lifestyle. Prev Med 1996 25:684–91.[CrossRef][Web of Science][Medline]
25. Kremers SPJ, De Bruijn G-J, Schaalma H, et al. Clustering of energy balance-related behaviours and their intrapersonal determinants. Psychol Health 2004 19:595–606.[CrossRef]
26. Victoir A, Van den Broucke S, Van den Bergh O. DIGG: Handleiding Voor Het Diagnose-Instrument Voor Gezondheid En Gedrag.Brussels, Belgium: VCLB Service 2002.
27. Ministry of the Flemish Community, Education Department. Schoolbevolking in het gewoon secundair onderwijs [School Population in Regular Secondary Education]. Available at: http://www.ond.vlaanderen.be/onderwijsstatistieken/2004-2005/zakboekje/secundair.pdf. Accessed: 21 September 2005.
28. Flanders Education Inspectorate. Inspectieverslag 1999–2000: Secundair onderwijs [Report of the Inspectorate 1999–2000: Secondary Education]. Available at: http://www.ond.vlaanderen.be/schooldirect/inspverslag/SO-stat.htm. Accessed: 20 May 2005.
29. Trafimow D, Sheeran P, Conner M, et al. Evidence that perceived behavioural control is a multidimensional construct: perceived control and perceived difficulty. Br J Soc Psychol 2002 41:101–21.[CrossRef][Web of Science][Medline]
30. de Vries H, Backbier E, Kok G, et al. The impact of social influences in the context of attitude, self-efficacy, intention, and previous behavior as predictors of smoking onset. J Appl Psychol 1995 25:237–57.[Medline]
31. Trafimow D and Finlay KA. The prediction of attitudes from beliefs and evaluations: the logic of the double negative. Br J Soc Psychol 2002 41:77–86.[CrossRef][Web of Science][Medline]
32. Bogers RP, Brug J, van Assema P, et al. Explaining fruit and vegetable consumption: the theory of planned behaviour and misconception of personal intake levels. Appetite 2004 42:157–66.[CrossRef][Web of Science][Medline]
33. Cohen J. Statistical Power Analysis for the Behavioral Sciences. 2nd edn. Mahwah, NJ: Lawrence Erlbaum 1988.
34. Flay BR and Petraitis J. The theory of triadic influence: a new theory of health behavior with implications for preventive interventions. In Albrecht GS (Ed.). Advances in Medical Sociology.Greenwich, CN: JAI Press 1994 vol. IV: pp. 19–44.
35. Sutton S. Some suggestions for studying situational factors within the framework of attitude-behaviour models. Psychol Health 1996 11:635–39.
36. Stotts AL, Schmitz JM, Grabowski J. Concurrent treatment for alcohol and tobacco dependence: are patients ready to quit both? Drug Alcohol Depend 2003 69:1–7.[CrossRef][Web of Science][Medline]
37. Avenevoli S and Merikangas KR. Familial influences on adolescent smoking. Addiction 2003 98:1–20.[Web of Science][Medline]
38. Vereecken C and Maes L. Voorlopig rapport jongeren en gezondheid 1990–2002 [Prelimnary Report Youth and Health 1990–2002]. Available at: http://users.ugent.be/cvereeck/hbsc/. Accessed: 1 February 2006.
Received on June 30, 2005; accepted on May 9, 2006
CiteULike 
Connotea 
Del.icio.us What's this?
| ||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||