HEALTH EDUCATION RESEARCH

Vol.23 no.4 2008 Pages 621632 Advance Access publication 25 September 2007

The role of life skills promotion in substance abuse prevention: a mediation analysis
Downloaded from http://her.oxfordjournals.org at University Library of Troms on September 1, 2010

Anneke Bu hler1*, Elke Schro der2 and Rainer K. Silbereisen2

Abstract
Research has shown that life skills programs are the most effective single activity in schoolbased substance abuse prevention. However, little is known about the processes through which they are effective. This study examines whether an evidence-based prevention program targeting general competence is effective through the promotion of knowledge about life skills and enhanced related behaviors. Based on a sample of 442 fth graders participating in a quasi-experimental prevention study, as expected, mediation analyses revealed that increased knowledge about life skills paralleled an increase in students distant attitudes toward alcohol and nicotine use. Unexpectedly, behaviors manifesting enhanced life skills were found not only among program participants who remained experimental/non-smokers or stopped smoking but also among smokers. In general, ndings suggest that favorable prevention outcomes may be inuenced through building knowledge about general life skills. The notion of uniform mechanisms of effectiveness in prevention programs is discussed.

Introduction
Although cigarette, alcohol and other substance use is a major risk factor for serious illnesses and mortality, they are behaviors practiced all over the world, and Germany is no exception. Among young people aged 1225 years, ;35% reported regular smoking and binge drinking during the last month prior to testing [1], and every fourth had tried illegal substances, mostly cannabis. As early regular substance use has been shown to be a risk factor for later substance-related problems of adjustment [2], these gures are alarming. Efforts to prevent substance abuse among adolescents can be effective, especially if socially interactive methods are used [3]. Interactive methods, which assign a facilitating role to the program mediator and stimulate exchange among program participants through work in small groups or roleplay, are used by the life skills approach. Life skills programs aim to prevent problem behavior by promoting abilities for adaptive and positive behavior, that enable us to deal effectively with the demands and challenges of everyday life [4, p. 1]. In this regard, they represent the notion of promoting positive development rather than relying on avoiding risks [5]. Most life skills programs consist of a general competence enhancement part (e.g. training in communication and interrelationship skills, critical and creative thinking, decision making and problem solving, self-awareness and empathy and coping with stress and emotions) and a part focusing on problem behavior (information about substances, value clarication, norm education, etc.).

r Therapieforschung, Parzivalstrasse 25, IFT Institut fu 80804 Munich, Germany and 2Department of Developmental Psychology, Friedrich Schiller University, Am Steiger 3, 07743 Jena, Germany hler. *Correspondence to: A. Bu E-mail: buehler@ift.de

The Author 2007. Published by Oxford University Press. All rights reserved. For permissions, please email: journals.permissions@oxfordjournals.org

doi:10.1093/her/cym039

A. Bu hler et al. The life skills approach has been found to be the most effective single activity in school-based substance abuse prevention, surpassed only by system-wide approaches [3]. Encouraged by the US experience, German researchers started to develop, implement and evaluate school-based life skills programs at the beginning of the 1990s (for a review, see [6]). In contrast to results found in North America, universal prevention through life skills programs affected substance use primarily in younger samples (Grades 16) rather than among older students. They resulted in a smaller percentage of students who smoked and a lower rate of drunkenness compared with the control condition [e.g. 79]. Knowing that a prevention program is effective leads to the question of why and how the program worked. With regard to universal prevention, metaanalysis of Tobler et al. [3] shows that programs using interactive methods, implemented in smaller projects, with a certain length (1130 sessions), with health professionals as program facilitators, targeting younger and older students (<7th grade or >10th grade) achieve higher effect sizes. Concerning selective prevention Springer et al. [10] identied ve aspects which were associated with positive outcomes: life skills-related content, connection-building methods, introspective methods, higher intensity (>3 h week) and coherent concept and implementation. Twenty-one studies that directly compare prevention programs with and without possible core ingredients of effective substance abuse prevention were reviewed by Cuijpers [11]. He concludes that (i) booster sessions may be benecial depending on other program characteristics, (ii) the leader (peerled or adult-led) may constitute one of several characteristics which determine effectiveness and (iii) effects of school programs can be increased when community components are added. Finally the mediation analysis approach tests whether change in mediating variables targeted by the intervention is due to program participation and whether this in turn can be related to change in substance use behavior [12]. Mediation analyses were rarely conducted in substance abuse prevention 622 [3, 11]. Cuijpers [11] found only seven studies examining this issue, meanwhile another four papers were published [1316]. The mediating variables identied were rather proximal and substancespecic risk factors such as, acceptability of drugs, knowledge about substances and prevalence of use, perceived peer approval of drug use, beliefs about substances and commitment or intention not to use. Only two distal, unspecic variables were found to be mediators of preventive effectsless tendency to risk-taking behavior in a school-based life skills program and improved parentchild communication in a family-oriented program (although results are inconsistent across studies [15]). Other substance-unspecic variables (e.g. interpersonal and decision-making skills, communication skills, problem solving, etc.) regarded as necessary ingredients of successful substance abuse prevention [17], were also promoted through program participation [1822]. Whether promotion of these life skills actually mediated the preventive effects, however, has not been proven up to now. The only two studies that attempted to conduct a mediational analysis including life skills measures were inconclusive. In these studies, Botvin et al. [23, 14] implemented their program with samples of students at high risk for substance use but could not test the mediating role of enhanced life skills because they found no program effects on the skills they measured. Taken together, prevention research has revealed several characteristics that are associated with effective substance abuse prevention outcomes. These characteristics concern implementation (length, intensity, size of project) and conceptualization (theoretical approach, interactive methods). The life skills approach is seen as a powerful tool in universal and selective prevention; although up to now, program effects have rather been related to change in proximal substance-specic variables than to generic life skills enhancement. Thus While the research conducted thus far examining the impact of the LST [life skills training] program on hypothesized mediators and attempts to identify specic mediating mechanisms are important rst steps, it is clear that additional research is needed in

Downloaded from http://her.oxfordjournals.org at University Library of Troms on September 1, 2010

Life skills promotion in substance abuse prevention this area [24, p. 226]. Against this backdrop, we wanted to test whether a program aimed at general competence enhancement is effective due to its promotion of life skills. Our hypotheses were as follows: (i) Participation in a substance abuse prevention program targeting general competence results in enhanced life skills knowledge and life skills behaviors as well as lower substance afnity and fewer cases of substance abuse, compared with controls. (ii) Enhancement of life skills knowledge and behaviors due to program participation mediates the program effects toward lower substance afnity and fewer cases of substance abuse among students. tions. The population of fth graders in Realschulen in the region we studied was 1505 [26]. Schools showed interest in school-based substance abuse prevention and agreed to participate in the study. Power analysis revealed that given an alpha level of 0.05, an anticipated effect size of 0.11 (small effect size), a desired statistical power level of 0.8 and a number of three predictors in a regression model, a minimum sample size of 103 is required (http:// statpages.org/index.html#Power). Data analysis is based on the reports of 448 students who completed questionnaires at all measurement points (treatment group n = 256, control group n = 192). The 195 missing cases were absent at one measurement point. This retention rate of 69.7% is similar to other German school-based evaluation studies [e.g. 9]. As shown in Table I, mean age of study participants was 10.8 years. Gender was distributed equally (49.8% girls) and the mean school mark reported was slightly >3 (marks range from 1, very good to 5, failed). Following the Solomon-four-groups-design [27], data of a comparable additional sample (n = 110) were used to investigate testing effects that could threaten internal validity. This random sample consisted of experimental and control students who had completed questionnaires at post-test only. Thus, comparing the outcome variable of all four groups, the factor pre-testing indicates whether it is testing rather than the intervention that has an effect and the interaction factor pre-testing 3 treatment indicates whether treatment in combination with a pretest has a different effect than without pre-test.

Downloaded from http://her.oxfordjournals.org at University Library of Troms on September 1, 2010

Methods Design and procedures

A pre-and post-test study with a treatment and control condition was conducted in schools. Classrooms were randomly assigned to the Solomonfour-groups design, i.e. to the experimental or control condition and to the pre-test or no pre-test condition. Passive informed consent of participation was given by parents. Teachers were trained during a 2-day workshop and were supervised throughout the intervention. The program was implemented mid-year in fth grade. Project staff administered questionnaires to students following a standardized protocol emphasizing condentiality of participant reports. Baseline and post-test data (12 months later) were collected during a regular 45-min classroom period. A bogus pipeline procedure [25] was used to increase the validity of selfreport substance use data.

Measures
In order to assess program implementation, after every program session teachers reported the number of program elements implemented and rated their quality of teaching on a ve-point scale. In addition, they reported the share of attentive and actively participating students. Substance use was measured using items that were developed by consensus in several German prevention research groups [8, 9, 28, 29]. Respondents were asked whether they had ever used tobacco or alcohol (response format: yes, several times, 623

Sample
The sample comprised 643 fth graders from 22 classes of seven Realschulen, a type of noncollege bound school typically leading students into professions not requiring university-level qualica-

A. Bu hler et al.
Table I. Sample and dropout characteristics Retention sample Treatment (n = 256) Gender: % girls (n) Age: years, M (SD) School achievement: M (SD) Substance use Nicotine abuse: % (n) Alcohol abuse: % (n) Tobacco distance: M (SD) Alcohol distance: M (SD) Life skills Knowledge skilled behavior: M (SD) Knowledge unskilled behavior: M (SD) Resources: M (SD) Decits: M (SD) 50.0 (128) 10.7 (0.61) 2.9 (0.71) 3.5 (9) 13.1 (33) 1.9 (0.71) 1.7 (0.83) 0.91 (0.15) 0.55 (0.24) 36.7 (11.5) 15.8 (5.1) Control (n = 192) 49.5 (95) 10.8 (0.65)b 2.7 (0.70)b 3.7 (7) 10.4 (20) 1.8 (0.64) 1.7 (0.81) 0.91 (0.15) 0.54 (0.25) 37.8 (11.5) 16.7 (5.1) Total (n = 448) 49.8 (223) 10.8 (0.63) 2.8 (0.68) 3.6 (16) 11.9 (53) 1.8 (0.68) 1.7 (0.82) 0.90 (0.15) 0.56 (0.24) 37.1 (11.5) 16.2 (5.1) Dropouts Total (n = 195) 38.9 (75)a 10.9 (0.62) 3.1 (0.84)a 4.8 (7) 14.5 (21) 1.8 (0.70) 1.7 (0.82) 0.89 (0.16) 0.50 (0.25)a 34.3 (12.2)a 14.9 (5.4)a

Downloaded from http://her.oxfordjournals.org at University Library of Troms on September 1, 2010

M = mean, SD = standard deviation. a Signicant difference between retention and dropout sample (P < 0.05). bSignicant difference between treatment and control group (P < 0.05).

yes, I tried and no) and whether they had used it recently (in the past 30 days). Current users reported frequency of use during the past 30 days and amount of use per occasion. The amount of alcohol use per occasion was reported in a free format. Subsequently, data were dichotomized into abuse and no use/experimental use. Any use beyond trying, i.e. consumed several times during lifetime or current use of more than one cigarette during the last week or more than a sip of alcohol per occasion, was classied as substance abuse. This denition is rooted in a developmentally oriented perspective on adolescent substance use that classies any use beyond experimenting in this age group as abuse [30, 31]. Afnity toward tobacco and alcohol was measured with two scales developed for prevention studies where their use predicted later substance use behavior (Tobacco and Alcohol Distance) [28, 29]. Scales consist of ve items each. Three items assess knowledge about tobacco/alcohol, e.g., Passive smoking is harmless. If two of three knowledge items are correctly answered by the respondent, a distance point is assigned. The fourth item assesses the affective aspect of afnity: Adolescent smokers/drinkers are posers. If this item 624

is supported, another distance point is assigned. Finally, the intentional aspect of afnity is measured by a fth item I do intend to smoke/drink regularly in the next 12 months. If respondents disagree with this item, a third possible distance point is assigned. Points are summed to classify high-distance adolescents (3 points), middledistance adolescents (2 points) and low-distance adolescents (0 and 1 point). Due to the lack of established life skills assessment scales, items capturing life skills behavior and knowledge were developed by the authors. The life skills behavior scales reect the ve main life skills areas dened by prevention experts [4]: self-awareness and empathy, creative and critical thinking, communication and interpersonal relationships, decision making and problem solving and coping with stress and emotions (see http:// www.ift.de/index.php?id=211 for full questionnaire). Thirty-three items were formulated following the World Health Organization denition of each skill [4, p. 2f]. Three additional items indicating refusal assertiveness were taken from Botvins questionnaire [32, pp. 91110]. Respondents reported how often they showed the behavior in question (1 = seldom, 2 = more often, 3 = most of the

Life skills promotion in substance abuse prevention time). Exploratory factor analysis with Varimax rotation resulted in a life skills resources factor (23 items, e.g. problem solving: If a problem occurs, I solve it step by step, assertive communication: Im not afraid to say my opinion, coping with emotions: I try to cheer myself up when Im down, self-awareness: I think about my strengths) and a life skills decits factor (13 items, e.g. interpersonal skills: I nd it hard to start a conversation with someone I dont know, problem solving: If a problem occurs, I dont look for alternatives but act the way that comes into my mind rst, critical thinking: Other people can easily persuade me, even if previously I had a different opinion). A sum score for resource items (Cronbachs a at T1: 0.86, T2: 0.86) and for decits items (Cronbachs a at T1: 0.70, T2: 0.73) was computed. Life skills knowledge items reect facts about life skills taught in the program. Sixteen items assessed knowledge about assertiveness (three items, e.g. Talking loud and clear is assertive), communication (six items, e.g. One should show a friend that one is listening by nodding or looking at him/her) and problem solving (seven items, e.g. One should always solve problems without the help of others, inverted). Exploratory factor analysis with Varimax rotation suggested a two-factor solution (variance explained at T1: 23.7% at T2: 33.7%). Two mean scores according to the factor solution were computed. Factors reected knowledge about problem solving and communication (six items, Cronbachs a at T1: 0.45; T2: 0.71) and knowledge about unskilled problem solving and communication (seven items, Cronbachs a at T1: 0.53; T2: 0.70). As assertiveness variables loaded on one factor at T1 and on the other at T2, they were not included in the sum scores. Table II summarizes the measurements of the outcome variables and potential mediators. With regard to validity of the developed measures, Table III shows cross-sectional and longitudinal correlations between potential mediators and other life skills relevant instruments used in our study, i.e. a short version of the Problem Questionnaire [33] and of the Coping Across Situations Questionnaire [33]. Due to other research questions in our project, both instruments were divided into two scales, namely, a scale concerning intrapersonal and a scale concerning interpersonal problems or situations. We expected life skills knowledge to be positively associated with resources in the life skill domain and negatively associated with decits in skilled behavior. Based on research on stress and coping [32], stressful interand intrapersonal problems were hypothesized to relate negatively to resources and knowledge about

Downloaded from http://her.oxfordjournals.org at University Library of Troms on September 1, 2010

Table II. Measurement of the outcome variables and potential mediators No. of items Outcome variables Substance use Nicotine abuse Alcohol abuse Tobacco afnity Alcohol afnity Mediator variables Life skills Knowledge skilled behavior Cronbachs a (T2) Score range

Abuser versus non-user/experimental user Abuser versus non-user/experimental user Low distance, middle distance, high distance Low distance, middle distance, high distance

1 1 5 5

1/0 1/0 13 13

One should show a friend that one is listening by nodding or 6 looking at him Knowledge unskilled behavior One should talk until somebody interrupts 7 Resources If a problem occurs, I solve it step by step 23 Decits I nd it hard to start a conversation with someone I dont know 13

0.71 0.70 0.86 0.73

13 13 14 14

625

A. Bu hler et al.
Table III. Cross-sectional and longitudinal correlations between potential mediators and other life skills relevant measures Resources Time Resources Decits Knowledge skilled Knowledge unskilled Interpersonal stress Intrapersonal stress Interpersonal coping Intrapersonal coping
a

Decits 2 0.40*** 1 0.49** 0.12* 2 0.17** 0.49** 0.24**

Knowledge skilled 1 0.14** 0.02 0.02 0.06 2 0.02 0.25** 0.06 0.04 0.22**

Knowledge unskilled 1 0.12* 0.12* 0.09 0.04 0.20** 0.13* 2 0.04 (1)a 0.11* 0.03 () 0.14* 0.16** 0.20* 0.46** 0.05 () 0.09 0.07 0.11 0.01 (1) 0.21** 0.01 0.16**

1 2 1 2 1 2 1 2 1 2 1 2 1 2 1 2

Downloaded from http://her.oxfordjournals.org at University Library of Troms on September 1, 2010

0.07 0.04 0.01 0.08 0.00 0.21** 0.01 0.25**

0.07 () 0.04 0.00 0.06 0.03 (1) 0.22** 0.11* 0.24**

0.17** 0.09 0.12* 0.05 0.06 0.12* 0.00 0.07

0.12* (1) 0.13* 0.13* 0.19** 0.00 () 0.15** 0.07 0.08

0.08 0.16** 0.14** 0.19* 0.02 0.12* 0.01 0.14*

0.06 0.04 0.09 0.13* 0.01 0.20** 0.02 0.23**

0.02 0.13* 0.06 0.09 0.04 0.11* 0.01 0.14*

expected direction of correlation coefcients negative (italics) or positive (bold). *** p < .001, ** p < .01, * p < .05.

life skills and positively to reporting decits. In contrast, correlations between active and socially oriented coping behavior and resources or life skills knowledge should be positive whereas correlations between this coping style and decits should be negative. As shown, the coefcients are rather low and one of the expected relations is not significant at all (resources and stress). However, all signicant coefcients point to the direction expected.

uses interactive methods and emphasizes reference to the personal daily life of students. Control classrooms did not receive any systematic drug prevention activity. However, in some classrooms, social competence was trained as part of the routine school curriculum and drugs were sometimes discussed in cases where substance use had been observed among students. Thus, the control condition is considered a treatment-as-usual control group.

Intervention and control condition

Allgemeine Lebenskompetenzen und Fertigkeiten (ALF) (general life competencies and skills) is a life skills program for fth graders consisting of eight sessions on general life skills training (communication, interpersonal relationships, critical thinking, self-awareness, problem solving, coping with stress and emotions) and four sessions on substance use-related issues (information about immediate effects of nicotine and alcohol use and short-term and long-term negative consequences, normative education, discussion about motivation to smoke and drink, media and social inuences on use, resistance skills training) [34]. The program 626

Data analysis strategy

The process of mediation (see Fig. 1) is represented by the path from the program through the mediating variables (life skills) to the outcome. The mediator effect is equal to ab. Other effects in the model include the direct effect, s, and the total effect s = ab + s. Mediation regression analysis was conducted according to MacKinnons recommendations [12] using the statistical procedures described in MacKinnon and Dwyer [35]. Three types of regression models provided the parameter estimates and the standard errors needed to establish the presence or absence of mediation between one mediating

Life skills promotion in substance abuse prevention correspond to the unit of analysis (individual level), and consequently intraclass correlation may bias results of inferential statistics. To avoid this bias, inferential tests were computed with SUDAAN [37], a software program that takes intraclass correlation into account through adjustment of the standard estimate error of regression coefcients.
Fig. 1. Mediation model.

Downloaded from http://her.oxfordjournals.org at University Library of Troms on September 1, 2010

Results Preliminary analyses

variable and the outcome variable [36]. Depending on whether the criteria was assessed on interval level or was dichotomous, a linear or logistic regression was computed. The rst regression model tested and estimated the effect of the program on the outcome variables (s). The second regression model tested and estimated the effect of the program on the hypothesized mediating variable (a). If the program did not have at least a marginally signicant effect (P < 0.10) on the mediator hypothesized, no further analyses were conducted with that variable. The third regression model tested and estimated the effect of the mediating variable on the outcome variable after adjusting for the effect of the program (b). The magnitude of the mediated effect was calculated by multiplying the latter two regression coefcients (ab). The standard error of the mediated effect (rab) was derived using the multivariate 2 2 delta method [square root of (a2 a2 b +b aa )] [12]. The mediated effect for the dichotomous outcome variable (nicotine abuse) was calculated by comparing the total program effect and the direct program effect (s s) because of its categorical property (see [34]). In addition, a measure of the extent that is mediated, the percent of the total effect that is mediated [ab/(ab + s)] [12], was calculated for each mediator. Baseline measures of the outcome variables and the mediator variables were used as covariates in the analyses. The statistical test for signicance was calculated using a Z statistic. Due to the directed nature of the hypothesis, onetailed tests were employed to examine the signicance of the estimated parameters. Note that the unit of randomization (classroom level) does not

Attrition was more prominent among boys and for students with lower school achievement (see Table I). In addition, dropouts had less life skills knowledge at baseline and reported less life skills resources than the remaining respondents. In contrast, dropouts also reported fewer decits in life skills. Unexpectedly, there were no signicant differences in substance abuse or substance afnity between the retained and the dropout sample. Signicant baseline differences between experimental and control students in the study were limited to age and school achievement: the mean age was slightly higher and the mean mark slightly better in the control than in the experimental group (see Table I). Implementation check results showed that the mean percentage of implemented elements per program session was ;90% (range 80100%). The mean number of sessions in which a student participated was 10.7 out of 12 sessions. Teachers reported ;60 to 85% of students were attentive during the sessions and that ;50 to 75% participated actively. Concerning the quality of implementation, teachers rated their own behavior between 2 and 3 on a scale of 15 (5 = very bad). In sum, program implementation was deemed satisfying. Concerning possible testing effects [27], there were no differences in substance-related and life skills outcome variables between the samples with and without a pre-test, i.e. in one-way analyses of variance (ANOVAs) for continuous outcome variables and in logistic regression for dichotomous outcome variables the factor pre-testing and the factor pre-testing 3 treatment was not signicant. 627

A. Bu hler et al. Thus, testing as a threat to internal validity can be ruled out. and use, e.g., in effective problem solving and communicating with others, in contrast to the controls. No program effects were found concerning knowledge about unskilled behavior and life skill decits, and consequently, no mediation analyses were conducted on these variables.

Program effects
Elsewhere we reported signicant effects of the prevention program concerning delaying the onset of smoking and heavy drinking [8]. This study replicated signicant effects on all outcome variables of interest except alcohol abuse. Nicotine abuse decreased in the intervention group from 3.5 to 2.4%, whereas in the control group, nicotine abuse increased from 3.7 to 6.3% (P < 0.05). Concerning tobacco and alcohol afnity, students participating in the intervention program developed a more critical view against smoking and alcohol consumption after the program (P < 0.001). In the control group, students afnity to smoking did not change; their afnity to drinking alcohol became more critical but did not reach the distant level of the intervention group (see Table IV). As hypothesized, there were statistically signicant program effects on the mediator variables knowledge about skilled behavior (P < 0.01) and life skill resources (P < 0.001). Participants gained from the program by enhancing their knowledge

Downloaded from http://her.oxfordjournals.org at University Library of Troms on September 1, 2010

Mediating effects
The mediating effects of the program on abuse and afnity toward smoking and alcohol drinking are presented in Table V. Knowledge about skilled behavior signicantly mediated the effect of the program ALF on all outcome variables, including nicotine abuse (P < 0.001), tobacco afnity (P < 0.01) and alcohol afnity (P < 0.01). As expected, enhanced life skills paralleled less nicotine abuse and an increasingly distant attitude toward legal substances. In contrast, life skills resources, did not signicantly mediate program effects on tobacco and alcohol afnity, but there was evidence for an inconsistent mediating effect [38] on nicotine abuse (P < 0.10). Although intervention students were less likely to smoke and were more likely to use life skills (see Table IV), enhanced life skills

Table IV. Program effects on substance use and life skills (n = 448) T1 ALF Substance use Nicotine abuse % (n) Alcohol abuse % (n) Tobacco afnity M (SD) Alcohol afnity M (SD) Life skills Knowledge skilled behavior M (SD) Knowledge unskilled behavior M (SD) Resources M (SD) Decits M (SD) Control T2 ALF Control Program effects s rs Z value

3.5 (9) 13.1 (33) 1.85 (0.71) 1.80 (0.70) ALF 0.91 (0.15) 0.54 (0.25) 36.7 (11.5) 15.8 (5.1)

3.7 (7) 10.4 (20) 1.8 (0.64) 1.65 (0.67) Control 0.90 (0.15) 0.56 (0.24) 37.8 (11.5) 16.7 (5.1)

2.4 (6) 10.8 (27) 2.14 (0.69) 1.98 (0.74) ALF 0.93 (0.16) 0.59 (0.30) 37.7 (12.8) 15.5 (5.4)

6.3 (12) 10.5 (20) 1.83 (0.65) 1.81 (0.70) Control 0.88 (0.21) 0.59 (0.28) 34.9 (11.6) 16.0 (5.2)

1.03 0.194 0.313 a 0.05

0.511 0.05 0.05 ra 0.02

2.01* n.s. 3.88*** 6.26*** Z value 2.50** n.s.

3.5

1.06

3.30*** n.s.

M = mean, SD = standard deviation. All models are adjusted for baseline substance use and baseline measures of the mediators. *P < 0.05, **P < 0.01, ***P < 0.001.

628

Life skills promotion in substance abuse prevention do not seem to represent the cause for their nonsmoking behavior. On the contrary, the negative score of the mediating effect indicates that enhanced life skills parallel onset of nicotine abuse. To clarify this issue further, in Fig. 2, the inconsistent mediating effect of life skills resources on nicotine abuse is shown. Changes in life skills resources are shown for three different groups of ALF students: (i) students who started heavier smoking between T1 and T2 (onset of nicotine abuse), (ii) students who remained non-smokers at T1 and T2 and (iii) students who changed smoking status from smoker at T1 to non-smoker at T2. The inconsistent mediating effect arises because a higher increase of life skills resources can be observed among smokers than among remaining experimental or non-smokers. However, the highest increase of life skills resources was reported by Alf students who stopped smoking. Mean differences between the three groups are signicant as an ANOVA revealed [F (2, 201) = 3,75, P < 0.05].

Downloaded from http://her.oxfordjournals.org at University Library of Troms on September 1, 2010

Table V. Mediation effects of the program on substance use Mediated effect (ab) or (s s)a Nicotine abuse (s s) Z value Percent mediated (ab/ab + s)

Discussion
So far, preventionists have not fully understood the mechanisms through which life skills programs are effective. The ndings of the current mediation analyses suggest that the prevention program ALF achieved its results of lowering the rate of nicotine abuse and fostering a critical attitude toward substance use among students in Grade 5 by increasing knowledge and use of general life skills. Learning techniques to manage the demands and challenges of everyday life successfully (e.g. problem solving, communicating) helped students to develop a critical view on substance use and to reduce nicotine abuse. In addition, process evaluation revealed that the program elements were successfully implemented in the intervention classes and a testing

Knowledge skilled 0.230 (0.060) 3.83*** 13.6 behavior Resources 0.074 (0.055) 1.34+ Tobacco afnity Knowledge skilled behavior Resources Alcohol afnity Knowledge skilled behavior Resources (ab) 0.015 (0.034) 2.30** n.s. (ab) 0.019 (0.049) 2.51** n.s. 14.8 5.0

All models are adjusted for baseline substance use and baseline measures of the mediators. a Parameter estimation depending on whether outcome variable is continous or dichotomous (see [28]). ** P < 0.01, ***P < 0.001, +P < 0.1.

Fig. 2. Inconsistent mediating effect.

629

A. Bu hler et al. effect [27] could be ruled out through the analysis of an additional sample. Knowing why an activity is preventive as well as knowing the limits of its effectiveness is helpful in optimizing substance abuse prevention activities [12]. The ALF program can be regarded as effective in the sense that enhanced protective factors led to a reduction of substance abuse. The fact that life skills resources and knowledge about skilled behavior were inuenced rather than life skills deficits and knowledge about unskilled behavior may reect the conceptual basis of the prevention program implemented: ALF was designed to focus on protective factors rather than on risk factors [39]. The program emphasized building up a skilled behavioral repertoire rather than correcting maladjusted behavior or restructuring maladjusted cognitions. In comparison to other prevention approaches like resistance skills programs or substance-specic information sessions, life skills programs are characterized by general competence enhancement. This study indicates the importance of substanceunspecic life skills promotion in preventing substance abuse, an issue that had not been proven up to now. Three aspects may explain why mediating effects were evident in our study. First, on the conceptual basis of general competence enhancement, we chose a global measurement with which to analyze the mediation role of general life skills in substance prevention. Other researchers [23, 14] tried to explain the underlying mechanisms of effective life skills programs by testing the mediating role of isolated life skills (e.g. decision making). Our ndings conrm that life skills promotion is related to favorable prevention outcomes and as such represents a promising way in substance abuse prevention. Secondly, the population examined in our study is a general population sample, the US studies targeted a minority sample [23, 14]. It may be that life skills promotion is a mediating condition in a non-select population rather than in at risk groups. Finally, teachers in our study reported to have implemented ;90% of program elements. The mean number of program points covered in one US study as observed by project staff was 630 48.2% [14]. Although observation and self-report data may be not comparable with regard to validity, this difference may point to the fact that life skills training needs to be comprehensively implemented with a high proportion of attentive and active students in order to inuence favorable prevention outcomes. We would like to emphasize that in contrast to other evaluation studies, our outcome variable tobacco use goes beyond experimental smoking behavior and indicates what from a developmental perspective [28, 29] can be classied as nicotine abuse. Thus, a more conservative and theoretically sound effectiveness criterion was applied, although, obviously, the nal goal of substance abuse prevention, i.e. a reduction of substance abuse in adolescence and adulthood, could not be tested in this study design. Our results show that increasing knowledge about life skills (communication, problem solving) was followed by a more distant attitude toward tobacco and alcohol and fewer cases of nicotine abuse after the intervention. Thus, cognitive changes concerning life skills presumably led to cognitive and behavioral changes in the substance use domain. The importance of cognitive changes in substance-specic constructs for successful prevention (e.g. acceptability of drugs, knowledge about the prevalence of use, perceived peer approval of drug use, intention not to use) has been demonstrated [11]. Our results point to the fact that changes in cognitions about general competence skills are also an important agent in successful substance use prevention. Unlike tobacco or alcohol afnity, behavioral changes do not seem to play a role in the development of substance-related cognitions. Our ndings concerning the mediating effect of life skills resources on nicotine abuse are to be interpreted as an inconsistent mediating effect [38]. Contrary to our expectations, ALF students who smoked after the intervention did so because they improved in life skills. This may be seen as an alarming message for preventionists favoring the life skills prevention approach. A closer look at Fig. 2, however, reveals that the situation is more complex. Indeed, ALF

Downloaded from http://her.oxfordjournals.org at University Library of Troms on September 1, 2010

Life skills promotion in substance abuse prevention students who started to smoke showed a higher increase in life skills resources than ALF students who remained experimental or non-smoker. This is reminiscent of research ndings showing that only substance-specic assertiveness was inversely associated with substance use, whereas dimensions of social assertiveness and dating assertiveness were positively associated with substance use [40]. Nevertheless, the highest increase of life skills resources was observed among those ALF students who stopped abuse. Thus, our results suggest that in this age group, enhanced general life skills may be less important for maintaining ones status as experimental or non-smoker than for giving up already developed heavier smoking behavior. Validity of these results is limited, however, because of small group size. Although speculative, this differential role of life skills promotion may support the notion of different pathways of substance use behavior [41, 28] responding to different elements of prevention. Students who started smoking heavily before fth grade may be early starters, a group of adolescents with more persistent problem behavior from childhood to adulthood [41]. They may have beneted from general life skills training in order to stop nicotine abuse. In contrast, for the majority of non-problematic adolescents training in general life skills was not the core mediating process in order to remain nonsmokers. Change in other mediating variables (e.g. problem-specic coping skills, substancespecic variables) may have been the agent in this group. Thus, although still needed to be replicated, our results point to the possibility that rather differential than uniform mechanisms of effectiveness operate in prevention programs. The present study has several strengths and weaknesses. Strengths include the random assignment, a sound theoretical denition of substance abuse, established data collection protocols, implementation check, control of testing effect and statistical analyses that control for intraclass correlations. Limitations of the study include the fact that the data are constrained to assessments 1 year after baseline measurement. Follow-up data are needed to determine whether the ultimate goal of substance use preventionless substance abuse in adolescence and adulthoodcan be accomplished through the prevention program. Finally, the fact that we focused on life skills promotion does not rule out the possibility of alternative mediating processes of effectiveness such as group processes on the classroom level or substancespecic cognitive mechanisms. Future investigation considering more than one plausible mediating mechanism would help to decide what works and thus help to optimize established school-based prevention programs.

Downloaded from http://her.oxfordjournals.org at University Library of Troms on September 1, 2010

Funding
The German Science Foundation (grant DFG 296/ 31-1) and by a stipend of the German National Academic Foundation. Acknowledgements We gratefully thank all participating schools.

Conicts of interest statement

None declared. References
1. Federal Centre for Health Education. Drug Afnity of Adolescents in the Federal Republic of Germany 2004. Cologne, Germany: Federal Centre for Health Education, 2005. 2. Kraus L, Bloomeld K, Augustin R et al. Prevalence of alcohol use and the association between onset of use and alcohol-related problems in a general population sample in Germany. Addiction 2000; 95: 1389401. 3. Tobler N, Roona M, Ochshorn P et al. School-based adolescent drug prevention programs: 1998 meta-analysis. J Prim Prev 2000; 20: 275336. 4. World Health Organiszation (WHO). Life Skills Education in Schools. Geneva, Switzerland: WHO, 1994. 5. Lerner RM, Taylor CS, von Eye A. Pathways to Positive Development among Diverse Youth. San Francisco, CA: Jossey-Bass, 2002. 6. Bu hler A, Heppekausen K. Gesundheitsfo rderung durch Lebenskompetenzprogramme in Deutschland. [Health Promotion through Life Skills Programs in Germany.]. Ko ln, Germany: Bundeszentrale fu r gesundheitliche Aufkla rung, 2005.

631

A. Bu hler et al.
7. Bo lcskei P, Ho rmann A, Hollederer A et al. Suchtpra vention an SchulenBesonderheiten des Nikotinabusus. [Substance abuse prevention at schoolsthe special case of nicotine abuse.]. Pra v Rehabil 1997; 9: 828. 8. Kro ger C, Reese A. Schulische Suchtpra vention nach dem LebenskompetenzkonzeptErgebnisse einer vierja hrigen Interventionsstudie. [School-based prevention of substance abuse through life skills trainingresults of a four year intervention study.]. Sucht 2000; 46: 20917. 9. Petermann H, Fischer V. Wie effektiv ist schulische Suchtpra vention?Ergebnisse der Leipziger Pra ventionsstudie. [How effective is school-based substance abuse prevention?]. In: Leppin A, Hurrelmann K, Petermann H (eds). Jugendliche und Alltagsdrogen. Konsum und Perspektiven der Pra vention. Berlin, Germany: Luchterhand, 2000, 14161. 10. Springer JF, Sale E, Hermann J et al. Characteristics of effective substance abuse prevention programs for high-risk youth. J Prim Prev 2004; 25: 17194. 11. Cuijpers P. Effective ingredients of school-based drug prevention programs. A systematic review. Addict Behav 2002; 27: 100923. 12. MacKinnon D. Analysis of mediating variables in prevention and intervention research. In: NIDA. Scientic Methods for Prevention Intervention, Research Monograph (Bd. 139). Washington, DC: NIH, 1994, 12753. 13. Botvin G, Grifn K, Diaz T et al. Smoking initiation and escalation in early adolescent girls: one-year follow-up of a school-based prevention intervention for minority youth. J Am Med Womens Assoc 1999; 54: 13943. 14. Botvin G, Grifn K, Diaz T et al. Drug abuse prevention among minority adolescents: posttest and one-year followup of a school-based preventive intervention. Prev Sci 1999; 2: 113. 15. Ennett S, Baumann K, Pemberton M et al. Mediation in a family-directed program for prevention of adolescent tobacco and alcohol use. Prev Med 2001; 33: 33346. 16. Orlando M, Ellickson P, McCaffrey D et al. Mediation analysis of a school-based drug prevention program: effects of Projekt ALERT. Prev Sci 2005; 6: 3546. 17. Dusenbury L, Falco M. Eleven components of effective drug abuse prevention curricula. J Sch Health 1995; 65: 4205. 18. Schinke S, Blythe B. Cognitive-behavioral prevention of childrens smoking. Child Behav Ther 1981; 3: 2542. 19. OHearn T, Gatz M. Evaluating a psychosocial competence program for urban adolescents. J Prim Prev 1999; 20: 11944. 20. Caplan M, Weissberg R, Grober J et al. Social competence promotion with inner-city and suburban young adolescents: effects on social adjustment and alcohol use. J Consult Clin Psychol 1992; 60: 5663. 21. Botvin G, Baker E, Dusenbury L et al. Preventing adolescent drug abuse through a multimodal cognitive-behavioral approach: results of a 3-year-study. J Consult Clin Psychol 1990; 58: 43746. 22. Botvin G. Preventing drug abuse in schools: social and competence enhancement approaches targeting individual-level etiologic factors. Addict Behav 2000; 25: 88797. 23. Botvin G, Dusenbury L, Baker E et al. Smoking prevention among urban minority youth: assessing effects on outcome and mediating variables. Health Psychol 1992; 11: 2909. 24. Botvin G, Grifn K. Life skills training: empirical ndings and future directions. J Prim Prev 2004; 25: 21132. 25. Baumann K, Koch G, Bryan E. Validity of self-reports of adolescent cigarette smoking. Int J Addict 1988; 17: 11316. 26. Landesamt fu r Datenverarbeitung und Statistik NRW. (2007). [Electronic data tape]. Available at: https://www. landesdatenbank-nrw.de. Accessed: 21 April 2007. 27. Cook T, Campbell D. Quasi-Experimentation. Chicago, IL: Rand McNally College Publishing, 1979. 28. Kersch B. Tabakdistanzein Evaluationskriterium unterrichtlicher Suchtpra ventionsmanahmen bei 13- bis 16ja hrigen Schu lerinnen und Schu lern Ergebnisse einer Leipziger La ngsschnittstudie. [Tobacco distanceevaluation criterion of substance abuse prevention programs for students age 13 to 16. Results of the Leipzig Longitudinal Study.]. Sucht 1998; 44: 1524. 29. Kersch B, Petermann H, Fischer V. AlkoholdistanzEin Evaluationskriterium schulischer Sucht- und Drogenpra vention. [Alcohol distanceevaluation criterion for schoolbased substance abuse prevention.]. Kindheit Entwickl 1998; 7: 24451. 30. Silbereisen R, Reese A. Alkohol und illegale Drogen. [Alcohol and illegal drugs.]. In: Raithel J, (ed.). Risikoverhaltensweisen Jugendlicher. Opladen: Leske+Budrich, 2001, 13153. 31. Newcomb M, Bentler P. Substance use and abuse among children and teenagers. Am Psychol 1989; 44: 2428. 32. Rhodes J, Jason L. Preventing Substance Abuse among Children and Adolescents. New York: Pergamon Press, 1988. 33. Seiffge-Krenke I. Stress, Coping, and Relationships in Adolescence. Research Monographs in Adolescence. New York: Erlbaum, 1995. 34. Walden K, Kro ger C, Kutza R et al. ALFAllgemeine Lebenskompetenzen und Fertigkeiten. Programm fu r Schu ler und Schu lerinnen der 5. Klasse mit Informationen zu Nikotin und Alkohol. [General Life Competencies and Skills. A Life Skills Program for 5th Graders.]. Hohengehren: Schneider Verlag, 1998. 35. MacKinnon D, Dwyer J. Estimating mediating effects in prevention studies. Eval Rev 1993; 17: 14458. 36. Judd CM, Kenny DA. Process analysis: estimating mediation in treatment evaluations. Eval Rev 1981; 5: 60219. 37. Shah B, Branwell B, Bieler G. SUDAAN Users Manual. Release 7.5. Research Triangle Park, NC: RTI, 1997. 38. MacKinnon D, Krull J, Lockwood C. Equivalence of the mediation, confounding and suppression effect. Prev Sci 2000; 1: 17381. 39. Kro ger C, Reese A, Walden K et al. Pra vention des Substanzmibrauchs an Schulen durch das Lebenskompetenzprogramm ALF. [School-based substance abuse prevention through the life skills program ALF.] IFT Berichte (Bd. 108). Mu nchen: IFT. 1999. 40. Wills T, Baker E, Botvin G. Dimensions of assertiveness: differential relationships to substance use in early adolescence. J Consult Clin Psychol 1989; 57: 4738. 41. Moftt T. Adolescence-limited and life-course-persistent antisocial behavior: a developmental taxonomy. Psychol Rev 1993; 100: 674701. Received on September 20, 2006; accepted on June 20, 2007

Downloaded from http://her.oxfordjournals.org at University Library of Troms on September 1, 2010

632