Subjects
Originally, the present sample is based on a cohort of N = 1,964 pupils aged 6 to 17 who were living in the Canton of Zurich, Switzerland in 1994. The cohort was a stratified randomized sample representing the 12 counties of the canton, the school grades, and the types of school and formed the basis of the Zurich Epidemiological Study of Child and Adolescent Psychopathology (ZESCAP). A full description of details of the sampling procedure was given in a previous article [22]. The preadolescents and adolescents (aged 11 – 17 years) of the ZESCAP sample (N = 1110) provided the basic cohort of the longitudinal Zurich Adolescent Psychology and Psychopathology Study (ZAPPS).
This cohort of 1110 subjects was studied longitudinally at three times, namely, in 1994 (time 1), 1997 (time 2) and 2001 (time 3). At each time, a multidimensional screening based on various questionnaires was performed. Subsequently, structured psychiatric interviews were used with those subjects who scored above the cut-off scores and with a certain number of controls scoring below the cut-off score on each screening instrument. Due to the screening procedure the sample was reduced at each time. In addition, subjects dropped out from the sample (e.g. after leaving school) on both the screening and the interview level. At the interview stage compared to the screening stage, a larger proportion of subjects and, particularly, more males than females dropped out. In order to work with a full data set including all questionnaires and interviews based on a sample that still was representative for local census data, the final longitudinal cohort with three waves of assessment was reduced to N = 593. Mean ages of this longitudinal cohort at the three times of assessment were 13.6 (SD = 1.6), 16.6 (SD = 1.6), and 20.2 (SD = 1.7) years. The sample was composed of 284 (47.9%) males and 309 (52.1%) females. These 593 subjects were representative for the census population with regard to gender (Chi2 = 2.14, df = 1, p = n.s.) and biannual age distribution of 17 – 22 years olds (Chi2 = 2.67, df = 2, p = n. s.).
Subjects with problematic substance use in this sample had to fulfil the following criteria: weekly or daily consumption of tobacco, daily alcohol consumption or heavy or problem drinking according to the definition that is given below in the description of the substance use questionnaire, and three or more times use of cannabis during the last month. At time 1, a total of N = 30 subjects were identified who met at least one of these criteria. There were 13 (43%) males and 17 (57%) females in this sub-sample. At time 2, based on the same criteria a total of N = 155 participants were identified who were considered to be subjects. There were 79 (50.1%) males and 76 (49.9%) females in this sub-sample. At time 3, a third PSU group based on the same criteria was defined. This sub-sample comprised a total of N = 290 subjects including 160 (55.2%) males and 130 (44.8%) females. At each time, the rest of the cohort without problematic substance use served as control group. The PSU group was significantly older than the controls at time 1 (15.1 vs.13.5, t = 7.99, df = 36.9, p < .001) and time 2 (17.0 vs.16.4, t = 4.32, df = 299, p < .001) but not at time 3 (20.2 vs. 20.2, df = 591, p = n.s.).
Measures
The ZAPPS is based on a theoretical model in order to study conditions and processes that are essential to the mental health of growing young people as well as to the development of mental problems and disorders. A broadband questionnaire was chosen in order to obtain information on relevant behavioural and emotional problems of adolescents. Furthermore, various questionnaires dealing with depression, abnormal eating behaviour, and substance abuse were also included. In order to analyze potential risk, compensatory, vulnerability, and protective factors of psychopathology [23], life events were hypothetically seen as stressors, and various psychosocial variables including coping, self-related cognitions, and features of the social network were regarded as moderating factors with regard to behavioural and emotional problems. Questionnaires were filled out confidentially by the subjects during school hours in 1994 and had to be mailed in 1997 and 2001. All questionnaires reflect raw scores and are positively keyed, i.e. high scores represent high expression of the content of the scale.
Substance Use Questionaire (SUQ)
The questionnaire was designed by Müller and Abnet [24] in collaboration with the World Health Organization for a nationwide Swiss survey. It covers 22 items that deal both with the consumption of legal drugs and illegal drugs. Nicotine use of both the respondent and his parents is covered by five items. A further eight items deal with alcohol use by the respondent. The response format varied for the different items. The introductory question for nicotine use inquired whether or not the respondent had ever smoked at least one cigarette. For frequency of current nicotine use the scale ranged from "does not apply = 0" to "daily = 3". Alcohol use was assessed via a general introductory question (0 = no consumption, 1 = only a sip, 2 = an entire glass or more) and a detailed list of various alcoholic beverages with a response format ranging from 0 (no consumption) to 5 (daily consumption). Various subgroups were identified and a typology of adolescent alcohol use was validated [21] Two types are relevant for the present study. Heavy drinkers were defined by two positive responses to the following items: I drink until I feel high/until I get drunk. Problem drinkers had to respond positively to the following two items: I drink when I feel lonely/when I feel bad and having a problem.
Youth Self Report (YSR)
The problem behaviour section of the YSR [25] and its Swiss adaptation [26] consists of the following primary subscales: socially withdrawn, somatic complaints, anxious/depressed, social problems, thought problems, attention problems, delinquent behaviour, and aggressive behaviour. Two second-order scales reflecting internalizing and externalizing can be calculated. Only these two dimensions were considered in the present analyses. Alpha coefficients of internal consistency for the two scales at the two times ranged from .81 to .87.
Young Adult Self Report (YASR)
With the exception of the subscale measuring social problems and the inclusion of the subscale measuring intrusiveness the YASR [27] consists of the same primary and secondary dimension as the YSR [28]. The YASR was used at time 3 (2001) and only the internalizing and externalizing problem scores were considered in the present analyses. The Alpha coefficients amounted to .89 and .80, respectively.
Life Event Scale (LES)
A total of 36 items were chosen from pre-existing questionnaires on life events. The time frame was defined as the twelve months prior to filling out the questionnaire. Beside frequencies of life events, a total impact score was calculated. This was based on a scale attached to each item ranging from -2 to +2 and indicating how unpleasant or pleasant the respective event was [29]. The alpha coefficients of internal consistency for the total number of life events ranged from .71 to .73 and for the total impact score from 0.71 to 0.84.
Coping Capacities (CC)
Our modified version of the German Coping Across Situations Questionnaire [30] addresses coping in four problem areas with school, parents, peers, and the opposite sex. Factor analysis resulted in two scales measuring active coping and avoidant behaviour. The CC was used at times 1 and 2 and the alpha coefficients of internal consistency for the two scales ranged from .56 to .70.
Self-Related Cognitions (SRC)
The ten-item scale for the measurement of self-esteem by Rosenberg [31] and items from a German questionnaire assessing self-awareness [32] were further included into the questionnaire. The latter scale assesses introspective capacities for one's feelings, actions, and past. Alpha coefficients for the two scales across the three assessments ranged from 0.77 to 0.89. The SRC was used at all three times.
Perceived Parental Behaviour (PPB)
Based on pre-existing literature, we developed an inventory that consisted of 32 items [33]. Factor analysis resulted in 3 factors explaining 34% of the variance for mothers and 35% of the variance for the fathers. The 3 scales were labelled "acceptance" (e. g., "my mother/father praises me when I do something good"), "rejection" (e. g. "my mother/father easily becomes upset if I don't do what she/he says") and "control" (e. g. "my mother/father has clear rules for my behaviour"). These scales were used only at time 1 and time 2. Alpha coefficients of internal consistency ranged between 0.68 and 0.89.
Perceived School Environment (PSES)
These scales were derived from a German project on development in adolescence [34] and consist of 32 items that deal with the perceived psychosocial qualities of the school environment. Our own factorial analyses re-identified the 5 factors labelled "competition among students" (e. g. "in our class, each student tries to be more successful than the other"), "control by the teacher" (e. g. "many of our teachers treat us like small children"), "performance stress" (e. g. "we hardly manage our homework"), "possibility to participate " (e. g. "our teachers ask for our opinion before deciding"), and "peer acceptance" (e. g. "I consider myself to be one of the most accepted students in our class"). These scales were used only at time 1 and time 2. The resulting scales had Alpha coefficients of between .65 and .79 at the two times of assessment.
Social Network (SN)
These newly developed scales cover six situations in which emotional or instrumental support is required. For each situation, the questionnaire asks whether or not 9 close individuals (family members, relatives, friends, and teachers) provide support. In addition, the efficiency of each of these individuals is also rated. Factor analyses across situations revealed 2 stable dimensions, namely size and efficiency of the social network with alpha coefficients ranging from .70 to .87 across the three times of assessment.
Statistical analyses
All questionnaire scores represent raw scores. Data were analysed by use of the 14th version of the SPSS (2006) program. Continuity of substance abuse at risk groups was tested by use of the McNemar Test. Comparisons between risk groups and controls were based on univariate and multivariate analyses of covariance (ANCOVA and MANCOVA) with sex and age as the controlled covariates. Logistic regression analyses were performed stepwise with forward selection in order to identify those variables that allowed the best prediction of cross-sectional and longitudinal risk status.