Pilot Evaluation Study of the Life Skills Program REBOUND: Effects on Substance Use, Knowledge About Substances, and Risk Perception

Abstract

The main aim of the study is pilot evaluation of the life skills program REBOUND in a school context focusing on substance use, risk perception, and knowledge about psychoactive substances ( = 723 students in five schools and 46 classes, Mage = 14.8, range 14-18) for the total sample and in the subgroups gender, age, and school type. Main goal of the study is collecting evidence for program optimization. A controlled study was carried out with repeated measurement before and after the intervention (4-6 months). Multilevel analyses, ANCOVA, and logistic regression analyses were applied to measure the effects. Overall, significantly lower incidence rates of drunkenness (odds ratio [OR] = .55; p = .033), improved knowledge about psychoactive substances (p = .006), lower personal (p = .013) and general tobacco risk perception among users (p = .002), and lower general tobacco (p = .018) and cannabis (p = .000) risk perception in non-users were found in the total intervention group. In subgroups, significantly lower rates for the incidence of drunkenness can be shown for males (p = .008) and for younger participants (p = .004). Students at academic high school (German Gymnasium) showed a decrease in 30-day prevalence for alcohol (p = .017) and cannabis (p = .014), and they improved in their knowledge about psychoactive substances (p = .000). In vocational high school classes (German Realschule), there was an increase in the relative alcohol risk perception of the students (p = .019). REBOUND contributes to a controlled use of alcohol and increases knowledge about psychoactive substances. REBOUND has various effects on the examined subgroups age, gender, and school type: Males, younger students, and students in academic high school benefitted more from the course regarding consumption-related criteria. We suggest a program optimization specific to school form and age, inclusion of a tobacco intervention, and the use of more gender-segregated interventions.

Full text

SAGE Open
October-December 2015: 1 –13
© The Author(s) 2015
DOI: 10.1177/2158244015617515
sgo.sagepub.com
Creative Commons CC-BY: This article is distributed under the terms of the Creative Commons Attribution 3.0 License
(http://www.creativecommons.org/licenses/by/3.0/) which permits any use, reproduction and distribution of
the work without further permission provided the original work is attributed as specified on the SAGE and Open Access pages
(https://us.sagepub.com/en-us/nam/open-access-at-sage).
Article
Although most universal school-based prevention programs
target students aged 6 to 13 (Foxcroft & Tsertsvadze, 2011;
Thomas & Perera, 2006), fewer programs have been designed
for the crucial phases of mid- to post-adolescence. Some pro-
grams focus on substance use, whereas others consider a
larger set of psychosocial variables, such as life skills or mul-
tiple problem behaviors (Botvin, Griffin, Paul, & Macaulay,
2003). Existing programs only aiming to reduce older ado-
lescents’ consumption of psychoactive substances have so
far been shown to be less effective than interventions for
younger students (Foxcroft & Tsertsvadze, 2011; Malwald &
Reese, 2000; Thomas & Perera, 2006; Tobler et al., 2000).
Positive effects shown in meta-analyses included a decrease
in frequency of drunkenness and “binge drinking” in the age
group of 13 years and younger (Botvin et al., 2003; Faggiano
et al., 2005; Poduska et al., 2008). Still, the development of
interventions for older adolescents is desirable, because the
readiness for experimentation, risk taking, and related social
costs increase at this age.
Concepts and Challenges in Promoting
Risk Competence
The life skills program REBOUND targets the age group of
14- to 25-year-olds. REBOUND aims at providing primary
prevention for non-consumers as well as secondary preven-
tion for consumers. To achieve this goal, REBOUND focuses
617515SGOXXX10.1177/2158244015617515SAGE OpenJungaberle and Nagy
research-article2015
1Heidelberg University Hospital, Germany
2FINDER Research, Berlin, Germany
Corresponding Author:
Henrik Jungaberle, FINDER Research, Palisadenstr. 67, 10243 Berlin,
Germany.
Email: Henrik.Jungaberle@finder-akademie.de
Pilot Evaluation Study of the Life
Skills Program REBOUND: Effects
on Substance Use, Knowledge About
Substances, and Risk Perception
Henrik Jungaberle1,2 and Ede Nagy1
Abstract
The main aim of the study is pilot evaluation of the life skills program REBOUND in a school context focusing on substance
use, risk perception, and knowledge about psychoactive substances ( nIG CG+ = 723 students in five schools and 46 classes,
Mage = 14.8, range 14-18) for the total sample and in the subgroups gender, age, and school type. Main goal of the study is
collecting evidence for program optimization. A controlled study was carried out with repeated measurement before and
after the intervention (4-6 months). Multilevel analyses, ANCOVA, and logistic regression analyses were applied to measure
the effects. Overall, significantly lower incidence rates of drunkenness (odds ratio [OR] = .55; p = .033), improved knowledge
about psychoactive substances (p = .006), lower personal (p = .013) and general tobacco risk perception among users
(p = .002), and lower general tobacco (p = .018) and cannabis (p = .000) risk perception in non-users were found in the total
intervention group. In subgroups, significantly lower rates for the incidence of drunkenness can be shown for males (p =
.008) and for younger participants (p = .004). Students at academic high school (German Gymnasium) showed a decrease in
30-day prevalence for alcohol (p = .017) and cannabis (p = .014), and they improved in their knowledge about psychoactive
substances (p = .000). In vocational high school classes (German Realschule), there was an increase in the relative alcohol risk
perception of the students (p = .019). REBOUND contributes to a controlled use of alcohol and increases knowledge about
psychoactive substances. REBOUND has various effects on the examined subgroups age, gender, and school type: Males,
younger students, and students in academic high school benefitted more from the course regarding consumption-related
criteria. We suggest a program optimization specific to school form and age, inclusion of a tobacco intervention, and the use
of more gender-segregated interventions.
Keywords
life skills, risk education, addiction prevention, knowledge about psychoactive substances, risk perception, adolescence

2 SAGE Open
on the promotion of life skills and risk competence
(Fahrenkrug, 1998; Franzkowiak, 1996; Löwe, Braun, &
Kisser, 2008) in the consumption of alcohol and other drugs.
Risk competence is a life skill and as such is important in
preventing and reducing maladaptive coping strategies
known to be a major factor in developing addiction (Jessor &
Jessor, 1977; Lazarus & Folkman, 1984; Resch, 1999;
Silbereisen & Reese, 2001). It is a self-regulative strategy
sensu Bandura (2005) and can be seen as a part of health
behavior (Ajzen, 1991; Becker, 1974).
Four dimensions were hypothesized as constituting risk
competence: (a) self-reflection and risk reflection, (b) verifi-
able knowledge about psychoactive substances and their use,
(c) control, that is, the intention to and success with limiting
one’s use, and (d) an orientation toward positive goals in the
future that exclude destructive substance use.
Challenges for the promotion of risk competence are for
example delaying the onset of use, which is an important pre-
dictor for later addiction (Chou & Pickering, 1992; Laucht &
Schmid, 2007; Lynskey et al., 2003; Robins & Przybeck,
1985). Others are the heterogeneous consumption habits at
this age, differences in gender, and school-form consumption
patterns that need to be considered for intervention and call
for the analysis of subgroups in evaluation (Rothwell, 2005).
For example, male gender is associated with higher con-
sumption of psychoactive substances (Kulis, Yabiku,
Marsiglia, Nieri, & Crossman, 2007; Lev-Ran, Le Strat,
Imtiaz, Rehm, & Le Foll, 2013; van Etten & Anthony, 2001).
The participants of the present study were from voca-
tional high school (Realschule) and academic high school
(Gymnasium). Studies have shown different prevalences of
substance use at different types of schools (Bundeszentrale
für gesundheitliche Aufklärung, 2012; O’Malley, Johnston,
Bachman, Schulenberg, & Kumar, 2006).
Hypotheses
We assume that the participation in REBOUND has a posi-
tive effect on risk competence that will lead to no or a
decreased consumption of alcohol, tobacco, and cannabis;
promote adequate risk perception; and increase participants’
knowledge about psychoactive substances. In addition, we
examine whether the reported effects in predefined sub-
groups hold for REBOUND; that is, participants benefit
more from the program, when they are younger, female, and
students at academic high school.
Method
Study Design and Setting
A controlled study with repeated measurements was con-
ducted in southern Germany from September 2011 to July
2012. In this naturalistic field study, the intervention group
(IG) consisted of school classes (ninth and 10th graders)
taking part in the program over a period of 4 to 5 months.
The control group (CG) consisted of classes of the same age
group. Twenty-nine course instructors were trained for the
pilot phase and implemented the program for their first time.
Due to the number of participating schools in the program’s
developmental phase, randomization in this quasi-experi-
mental design was conducted by classrooms, rather than by
the low number of schools in the development phase and did
not succeed in all cases, due to the practical fact that some
schools were not able to adjust their class schedules in such
a way that the REBOUND teachers could be assigned to the
randomized classes. Blind randomization was not possible
because students, parents, and trainers would know which
group participated in REBOUND and which did not.
The survey was conducted online using EFS-Survey 8.1
(Décieux, Heinz, & Jacob, 2011). It took place within 1 week
before and after the intervention (M processing time 29 min;
SD = 6.6). Response to all questions was mandatory with the
exception of irrelevant subordinate questions.
Intervention
The 16 units of REBOUND were integrated into school cur-
ricula and taught in weekly sessions of 90 min. The sessions
were facilitated by trained course instructors teaching at the
participating school. They were optionally assisted by peer
mentors. The intervention is described in detail in Kröninger-
Jungaberle, Nagy, von Heyden, DuBois, and the REBOUND
Participative Development Group (2014).
REBOUND integrates best-practice elements, such as
promotion of life skills, information about psychoactive sub-
stances, the use of peer mentoring, and normative education
(Cujpers, 2002; Hawks & Scott, 2002; Tobler, et al., 2000).
In addition, its developers created innovative methods, most
notably “explorative film work” consisting of joint scenario
analysis of videos depicting risk behaviors in realistic con-
text. The latter is based on Bandura’s (2005) self-regulation
approach and aims to enhance cooperation and risk assess-
ment among students. In the course of the sessions, health-
promoting norms are identified by reflecting healthy and
harmful cognitive–emotive and behavioral patterns. For
example, the classroom would frequently talk about the abil-
ity to take a conscious break from weekly alcohol consump-
tion or the necessity of training general self-control, body
awareness, and peer-pressure resistance skills, if thinking
about trying a psychoactive substance.
Measures
Our large questionnaire contained five different groups of
items relating to socio-demographic variables, substance
use, risk perception, knowledge about psychoactive sub-
stances, and possible confounding covariates.
The following socio-demographic variables were collected:
school type, gender, age, housing, family setting, pocket money,

Jungaberle and Nagy 3
parents’ occupation, recreational activities, and satisfaction
with school performance. Satisfaction with school perfor-
mance was measured using a 7-point Likert-type scale rang-
ing from 1 = not at all satisfied to 7 = very satisfied.
The frequency of alcohol, tobacco, and cannabis use was
measured by lifetime, 6-month-, and 30-day prevalence. The
information concerning the 6-month prevalence was gath-
ered using a 7-point Likert-type scale ranging from 0 = not at
all to 6 = 40 times or more. Thirty-day prevalence was equiv-
alent to the number of consumption days in a month (0 to
31). In addition, frequency of drunkenness and binge drink-
ing were ascertained for the last month. Binge drinking was
defined as the consumption of at least five alcoholic drinks
on one occasion. Finally, the intensity of last alcohol intoxi-
cation was measured on a 7-point Likert-type scale ranging
from 1 = only slightly tipsy to 7 = so drunk that I could hardly
remain standing on my feet. The items were adapted from a
survey among 12- to 25-year-olds in Germany conducted by
the Federal Centre for Health Education (Bundeszentrale für
gesundheitliche Aufklärung, 2012).
Three different kinds of risk perception were measured:
the general—“how risky is it for everybody?” personal—
“how risky is it for me?” and relative—“how risky is it for me
compared with for other people?” These were assessed for
use of alcohol, tobacco, and cannabis. We also asked non-
consumers about their perception of the risk for the general
population. All were done using a 7-point Likert-type scale
from 1 = harmless to 7 = very dangerous.
We developed a 40-item test to assess knowledge about
alcohol, tobacco, cannabis, and psychoactive substances in
general with 10 items for each category (e.g., nicotine improves
concentration and short term memory; answers given as true,”
not true, do not know). The do not know answers were counted
as wrong answers. Cronbach’s alpha for the total scale was .80
before and .85 after the intervention.
The following covariates were measured on a 5-point
Likert-type scale from 1 = not at all true to 5 = absolutely
true: normality of the consumption of alcohol, tobacco, and
cannabis in the peer group (e.g., drinking of alcohol is com-
pletely normal in my peer group), parental acceptance of
one’s own use of alcohol, tobacco, and cannabis (e.g., for my
parents, it’s ok, if I smoke), not being liked by peers (I am not
liked by other kids), being bullied (other students often make
fun of me), behavioral problems in school (I have had fre-
quent problems in the school, e.g., fighting, disturbing the
lesson), perceived bio-psycho-social stress (compared with
other people my family and I have more problems such as
divorce, illness, financial problems), and the perceived rela-
tionship with parents (in general, I have a good relationship
with my parents).
Study Sample
We were able to include 1,148 students in 48 school classes in
five schools in the quasi-experimental design. Twenty-three
students chose not to participate. IG: n = 679 students in 29
school classes, CG: n = 446 students in 19 school classes.
The sample was constituted by students of academic and
vocational high schools; other German school types were
underrepresented (n = 16 students) and therefore excluded.
The quasi-experimental design resulted in a lack of equiva-
lence between IG and CG regarding a number of variables.
Hence, propensity score matching (Rosenbaum & Rubin,
1983) was carried out to achieve equivalence between IG and
CG, using the “nearest” algorithm of the R-package “MatchIt”
within the framework of the statistical software R.2.15.2 (Ho,
Imai, King, & Stuart, 2011). This achieved equivalence on all
test variables except “amount of pocket money a month”
(MIG = 32.2 Euro, MCG = 26.9 Euro; T = −2.49, p = .013; see
Table 1). Due to the matching procedure, the sample decreased
by 126 persons (11.8% of the total sample) and two classes.
The analyzed sample became after Attrition as follows: n =
723, 48.1% male, 51.9% female, M age: 14.8 years (median =
15; SD = .74, range = 14-17), school type: vocational high
school: 48.7%, academic high school: 51.3% (see Figure 1).
Multilevel structure. The data have a hierarchical structure
(Level 1 = students, Level 2 = classes). Because of the similar
context, higher correlations within the classes were expected
than between them (intraclass correlation coefficient [ICC] ρ).
Therefore, at first, the ICC of the study variables (difference
values between the pre- and post-measurements: T2-T1) was
computed. This procedure was necessary to calculate the design
effects (DE), so that the appropriate analytical methods could be
determined (conventional vs. multilevel analyses). DE is a func-
tion of the ICC and the average cluster size (m), which allows
adjusting the sample size at excessive ICC. DE is calculated
according to the formula 1 + (m − 1) × p. In the case of a DE >
2, the application of a multilevel analysis is recommended (B.
O. Muthén & Satorra, 1995). In addition, determining the ICC
was necessary for the post hoc calculation of the discoverable
effect sizes for hierarchical data. For two covariates and five
dependent variables, DE > 2 was found. The rest of the variables
yielded an average ICC after Fisher’s Z transformation of .024
(min: .00, max: .066). The ICC and DE are shown in Tables 2
and 3 in the two columns on the right side.
Control of Covariates
To statistically control the influence of covariates on the rela-
tionship between group membership and intervention effects,
covariates were selected that showed a high correlation with
at least three of the dependent variables, but low correlations
among each other.
These conditions were met by the following 11 variables,
which were subsequently used as covariates: type of school
(due to no variance at the cluster level not used in multilevel
regression analysis [MLA]), gender, traditional family set-
ting (living with both biological parents), high socioeco-
nomic status, satisfaction with school performance, leisure

4 SAGE Open
Table 1. Comparison and Characteristics of the Intervention and Control Group After the Propensity Score Matching.
CG (n = 211) IG (n = 512)
M SD n %M SD n % Statistics (df)pICC DE
School type χ2(1) = 3.40 .065
Junior high school 114 54 238 46.5
Academic high school 97 46 274 53.5
Gender χ2(1) = 1.48 .253
Male 109 51.7 239 46.7
Female 102 48.3 273 53.3
Age in year 14.8 0.8 14.8 0.7 t(721) = .69 .483
14 73 34.6 186 36.3
15 101 47.9 246 48
16 32 15.2 71 13.9
17 5 2.4 9 1.8
Lifetime prevalence
Alcohol use 156 73.9 368 71.9 χ2(1) = .31 .573
Tobacco use 65 30.8 174 34 χ2(1) = .68 .409
Cannabis use 18 8.5 47 9.2 χ2(1) = .08 .782
Drunkenness 59 37.8 155 41.9 χ2(1) = .75 .385
30 days’ prevalence of use (in days)
Alcohol 2.2 3.4 1.8 0.29 t(346) = 1.75 .100 .07 2.05
Binge drinking 0.53 1.1 0.47 0.97 t(349) = .71 .477 .07 2.1
Drunkenness 0.35 1.1 0.29 0.85 t(721) = .75 .454 .02 1.3
Tobacco 2.1 5.9 2.1 6.1 t(721) = .03 .975 .03 1.5
Cannabis 0.6 0.27 0.07 0.52 t(721) = .42 .682 .03 1.51
6 months’ prevalence of use (scale 1-7)
Alcohol 1.9 1.8 1.6 1.6 t(362) = 1.71 .071 .03 1.38
Tobacco 0.8 1.7 0.8 1.7 t(721) = .16 .874 .07 1.98
Cannabis 0.1 0.32 0.1 0.5 t(614) = .16 .164 .02 1.35
Intensity of the last drunkenness
(scale 1-7)
4.3 2 4.1 1.7 t(90.2) = .5 .615 .00 1.04
Alcohol risk perception among users
Personal 2.1 1.4 2 1.3 t(524) = .83 .406 .00 1
Relative 2.3 1.3 2 1.3 t(524) = 1.32 .186 .00 1.1
General 4.6 1.4 4.6 1.4 t(524) = .62 .535 .02 1.3
Tobacco risk perception among users
Personal 2.3 1.7 2.6 2 t(240) = .1 .325 .02 1.35
Relative 2.3 1.7 2.2 1.6 t(240) = .6 .554 .00 1.08
General 4.9 1.7 5.2 1.6 t(240) = .59 .291 .07 2.13
Cannabis risk perception among users
Personal 2.8 2.3 2.1 1.5 t(23.5) = 1.2 .165 .00 1.1
Relative 2.4 1.7 2.3 1.4 t(63) = .46 .648 .23 4.4
General 4.3 2 4.2 2 t(63) = .16 .876 .03 1.5
General risk perception among non-users
Alcohol 5.1 1.2 5 1.2 t(198) = .68 .494 .03 1.44
Tobacco 6.1 1 5.9 1.3 t(344) = 1.8 .07 .02 1.25
Cannabis 5.9 1.5 5.8 1.4 t(657) = .403 .687 .04 1.61
Knowledge about PAS 16.2 5.7 15.5 5.7 t(721) = 1.36 .173 .07 2.07
Satisfaction with own school
achievement
3.4 1.3 3.4 1.4 U = 54,004 .996 .10 1.15
Pocket money in Euro 26.9 20.5 32.2 35.5 t(650.5) = −2.49 .045*
Leisure activities
Alone 7 3.3 24 4.7 χ2(1) = .68 .408
With friends at home 168 79.6 389 76 χ2(1) = 1.12 .290
With friends in the disco, pub,
or café
70 33.2 165 32.2 χ2(1) = .06 .804
(continued)

Jungaberle and Nagy 5
activities, normality of tobacco use among friends (due to
DE > 2 used only in MLA), being bullied, behavioral prob-
lems in school, and perceived bio-psycho-social stress level.
For the subgroup analyses with a reduced number of vari-
ables, the eligibility conditions were met by the following
four variables: higher status parental occupation, normality of
cannabis use among peers, parental acceptance of students’
tobacco use, and leisure activities in disco, bar, or café.
Data Analysis
Total sample analysis. Because of the nested data structure,
MLAs (analysis type = two level) were applied using the
software Mplus 5.21 (L. K. Muthén & Muthén, 1998) for the
analysis of intervention effects on the five dependent vari-
ables (DV) with DE > 2. For each DV, a difference value of
pre- and post-measurement was calculated to reflect the indi-
vidual change in the outcome variables. In the MLA, a fixed-
effect model was used with a dummy-coded predictor
(group) and random intercept specified for each DV. The
influence of covariates was statistically controlled in the
model.
For dependent variables with a DE < 2, we applied univari-
ate analyses of covariance and logistic regression analyses
using SPSS 0.20. Logistic regression analyses including group
as the predictor and the covariates were used for the analysis of
intervention effects on the incidence rates of alcohol, tobacco,
cannabis, and drunkenness. The intervention effects on the
pre–post difference values of continuously scaled variables
were examined using fixed-factor group ANCOVAs.
Subgroup analyses. The sample was divided into subgroups:
gender, age (aggregated samples of 14- to 15-year-olds and
16- to 17-year-olds), and school type (vocational and aca-
demic high school). Alpha error accumulation in multiple
tests was countered by reducing the number of consumption
variables. The 30-day prevalence of alcohol, tobacco, and
cannabis use and the lifetime prevalence of drunkenness
were included in the analyses.
The following variables did not show equivalence
between IG and CG before the course (T1) and were excluded:
30-day prevalence of alcohol in the subgroup of vocational
high school students ( MIG = 1.93, MCG = 2.99, t = 2.45,
df = 174.5, p = .015), lifetime prevalence of drunkenness
(rateIG = 44.4%, rateCG = 30%, χ2 = 4.42, df = 1, p = .035),
and 30-day prevalence of tobacco use in the subgroup of aca-
demic high school students ( MIG = 1.53, MCG = .64, t =
−2.1, df = 281.6, p = .03); personal alcohol risk perception in
the subgroup of males ( MIG = 2.03, MCG = 2.51, t = 2.43, df
= 130.15, p = .016), relative alcohol risk perception in the age
group of 14 to 15 years old ( MIG = 2.07, MCG = 2.35, t =
2.03, df = 429, p = .043), overall alcohol risk perception in
non-drug users ( MIG = 4.73, MCG = 5.75, t = 3.02, df = 21,
p = .007), and general cannabis risk perception in drug users
(MIG = 3.85, MCG = 6.00, t = 3.19, df = 15.92, p = .006) in
the age group of 16- to 17-year-olds.
CG (n = 211) IG (n = 512)
M SD n %M SD n % Statistics (df)pICC DE
In social club 79 37.4 178 34.8 χ2(1) = .47 .494
Other 38 18 122 23.8 χ2(1) = 2.94 .087
Socioeconomic status 31 14.7 85 16.6 χ2(1) = .40 .525
Traditional home 154 73 377 76.6 χ2(1) = .03 .858
Normality of the PAS—consumption among friends
Alcohol 211 512 U = 50,611.5 .204 .10 2.53
Tobacco 211 512 U = 51,212.5 .299 .09 2.38
Cannabis 211 512 U = 53,188.5 .772 .6 1.93
Acceptance of the PAS consumption by the parents
Alcohol 211 512 U = 49,728.5 .099 .02 1.32
Tobacco 211 512 U = 53,540.5 .894 .01 1.22
Cannabis 211 512 U = 52,904 .427 .00 1.06
Not to be liked by others 211 512 U = 53,401.5 .881 .02 1.23
Being bullied by other 211 512 U = 53,032.5 .749 .02 1.34
Behavioral problems in the school 211 512 U = 53,169.5 .793 .13 1.91
Perceived bio-psycho-social stress 191 464 U = 43,187 .557 .00 1.09
Relationship to the parents 211 511 U = 49,731 .313 .00 1.06
Note. IG = intervention group; CG = control group; ICC = interclass correlation; DE = design effect; PAS = psychoactive substances; U = Mann–Whitney
U test. ICC and DE refer to the difference value between the pre- and post-measurements (T2 − T1).
p ≤ .05
Table 1. (continued)

6 SAGE Open
Only those subgroups are suitable for analysis in which
significant interaction effects can be found (Brookes et al.,
2001). First, interaction effects were examined between gen-
der, age, and school type as well as between membership in
the IG and CG. ANOVAs and logistic regression analyses
were conducted.
Some variables were dichotomous such as “drunkenness”
(yes/no). The incidence rate of these was calculated using
logistic regression analyses. Again, we controlled for the
influence of the four covariates. In each subgroup, seven
tests were carried out, which contributed to the accumulation
of the alpha error. Thus, sequential Bonferroni correction
sensu Holm (1979) was used, which is less conservative than
the usual Bonferroni correction.
Results
Attrition
Of the total sample (N = 2,153 pre- and post data sets in the
IG and CG), 709 (36 %) student data sets were missing
(nIG = 284; 21.7%; nCG = 425; 50.2%). From these, 15.3%
were caused by having no match between pre- and post-data
due to absent or deviant identification codes. In all, 5.9%
Figure 1. Composition of the sample and attrition from recruitment to analysis.

Jungaberle and Nagy 7
were missing due to response bias (e.g., processing time less
than 10 min, response tendencies) and implausible responses
(pocket money, for example, 9,999 Euros). Finally, 11.7%
were excluded by the propensity score matching. There was
evidence of systematic dropout comparing the analysis sam-
ple (N = 854) with the dropout sample (N = 276). Students
who dropped out were older, t(402) = −4.69, p = .000; more
likely to be male, χ2(1) = 26.8, p = .000; more likely to be
from vocational high school, χ2(1) = 4.52, p = .033; or spent
more of their recreational time with friends at home, χ2(1) = 4.92,
p = .017, or in the disco or café, χ2(1) = 5.46, p = .012. They also
had more behavioral problems in school (U = 94,158.5,
Table 2. Results of the Multilevel Regression Analyses: Intergroup Effects on the Dependent Variables (T2-T1) with EN > 2.
IG CG
Intervention effects on . . . M SD M SD B SE p
30 day’s prevalence of use (in days)
Alcohol 1.2 3.6 0.64 3.75 0.609 .484 .208
Binge drinking 0.67 1.49 0.84 1.59 −0.159 .231 .491
General tobacco—risk perception among users −0.22 1.78 0.48 1.44 −0.647 .278 .020*
Relative cannabis—risk perception among users −0.26 1.8 −0.47 2.56 0.242 .588 .680
Knowledge about psychoactive substances 1.65 6.03 .30 4.79 1.336 .483 .006**
Note. IG = intervention group; CG = control group.
*p < .05. **p < .01.
Table 3. Results of the ANCOVA: Intervention Effects on the Change (T2-T1) in Substance Consumption, in Risk Perception, and in
Knowledge About PAS.
CG (n = 211) IG (n = 512)
Difference T2-T1 n M SD n M SD AMD Partial η2df F p
30-day frequency of use (in days)
Drunkenness 73 0.63 2.58 168 0.48 2.09 −.04 .000 1 0.013 .910
Tobacco 47 1.66 8.44 123 2.11 7.38 .69 .000 1 0.003 .959
Cannabis 15 1.13 1.17 53 1.11 2.60 .20 .001 1 0.071 .791
6-month frequency of use (Scale 1-7)
Alcohol 164 0.40 1.23 371 0.41 1.28 .004 .000 1 0.001 .971
Tobacco 66 0.14 1.64 180 0.23 1.51 .009 .001 1 0.002 .968
Cannabis 22 0.64 1.14 67 0.46 1.25 −.09 .001 1 0.093 .761
Intensity of the last drunkenness
(scale 1-7)
43 0.42 2.14 126 0.13 1.83 −.20 .002 1 0.337 .562
Alcohol risk perception among users
Personal 142 0.13 1.60 314 0.08 1.40 −.04 .000 1 0.074 .786
Relative 142 0.03 1.27 314 0.18 1.37 .14 .002 1 1.07 .302
General 142 0.28 0.04 314 0.08 1.45 −.25 .007 1 3.16 .076
Tobacco risk perception among users
Personal 50 0.46 1.58 141 −0.33 2.01 −.85* .034 1 6.28 .013*
Relative 50 −0.32 1.56 141 −0.08 1.56 .17 .003 1 0.491 .485
Cannabis risk perception among users
Personal 15 −0.20 3.71 41 −0.41 1.92 −.66 .014 1 0.603 .442
Relative 15 −0.47 2.56 41 −0.24 1.83 .21 .002 1 0.103 .750
General 15 −0.40 1.99 42 −0.63 1.87 −.46 .013 1 0.561 .458
General risk perception among non-users
Alcohol 26 0.00 0.84 73 −0.30 1.19 −.28 .013 1 1.16 .283
Tobacco 119 0.05 0.96 266 −0.27 1.39 −.33* .018 1 5.60 .018*
Cannabis 166 −0.13 1.57 392 −0.64 1.61 −.54* .024 1 13.18 .000**
Note. PAS = psychoactive substances; M = non-adjusted mean; AMD = adjusted mean difference values (IG-CG); partial η2 = adjusted effect size of the
predictor “group.”
*p < .05. **p < .01.

8 SAGE Open
p = .000), a higher level of perceived biopsychosocial stress
(U = 94,158.5, p = .000), and used more alcohol, t(262) =
−5.32, p = .000, tobacco, t(228) = −3.16, p = .002, and can-
nabis, t(65) = −3.38, p = .001, in the last 30 days. At the same
time, they had a higher personal risk perception regarding
alcohol, t(365) = −2.03, p = .043, and tobacco, t(452) =
−2.13, p = .033, and a higher general risk perception regard-
ing cannabis, t(150) = −1.99, p = .049. Finally, they had bet-
ter knowledge about cannabis.
Results in the Total Sample
Sensitivity analyses. The calculation of effect sizes for hierar-
chical data with the five dependent variables with a DE > 2
was carried out using Optimal Design V.2.0 (Raudenbush,
2011). For the 46 classrooms with an average cluster size of
15.72, 1 − β = .80, α = .05 and
ρ
= .11, a detectable effects
size of d > .35 was calculated.
The sensitivity for non-hierarchical data was computed
using the software G-Power (Faul, Erdfelder, Lang, &
Buchner, 2007). Detectable effect sizes of d > .15 were found
for ANCOVA with n = 723, two groups (df = 1), 10 covari-
ates, 1 − β = .80, and α = .05. For the logistic regression
analysis with odds ratio (OR) = .55 (drunkenness) and OR =
.55 (traditional family setting), a power of 1 − β > .94 was
calculated (two tailed), Pr(Y = 1/X = 0), H0 = 0.5 1, α = .05,
n = 723, RotherX
2 = .08, Binomial × parm π = .68.
Substance use, risk perception, and knowledge about psychoac-
tive substances. Results of the multilevel regression analysis
are shown in Table 2. Significant regression weights were
found while controlling for covariates in the dependent vari-
ables general risk perception of tobacco use for users (B =
−.647, p = .02) and knowledge about psychoactive sub-
stances (B = 1.336, p = .006). The participation in REBOUND
was associated with a significant decrease in general risk
perception of tobacco use for smokers but remained on a
high level (4.9 on a 7-point scale) and resulted in a highly
significant increase in knowledge about psychoactive sub-
stances (see Table 3 for the results of the ANCOVAs).
The adjusted mean difference (AMD) provides informa-
tion about the difference between the mean values in the IG
and CG (comparisons of main effects were Bonferroni-
adjusted). Based on the adjusted effect sizes (partial η2) of
the predictor group, effect sizes f were calculated. A signifi-
cant influence of the predictor group was found regarding the
T2 − T1 difference values of the DV for personal risk per-
ception of tobacco use for users (AMD = −.85, F = 6.28,
df = 1, p = .018, f = .14) and general risk perception of
tobacco use (AMD = −.33, F = 5.6, df = 1, p = .018, f = .14),
and cannabis use for non-users (AMD = −.54, F = 13.18,
df = 1, p = .000, f = .16).
These results indicate that after participating in
REBOUND, users of tobacco rated their tobacco consump-
tion as less dangerous than tobacco users in the CG. In addi-
tion, non-user in the IG compared with the non-user in the
CG showed a decrease in general risk perception of tobacco
and cannabis.
Incidence rates. The results of the logistic regression analyses
are shown in Table 4. The best model fit was found by the
Hosmer–Lemeshow global model fit tests on the DV inci-
dence of cannabis use and incidence of drunkenness.
Although the best fit (Nagelkerke pseudo-R2) was found for
the DV incidence of cannabis use, the contribution of the
predictor group was not significant (Wald test). However, the
Wald test was significant for the predictor group for the DV
incidence of drunkenness. Students in the IG had a decreased
risk of initiating first drunkenness compared with the CG
(OR = .55; 95% confidence interval [CI] = [.32, .95]).
Using logistic regression, traditional family setting sig-
nificantly predicted the incidence of drunkenness (Wald =
3.9, df = 1, OR = .55, 95% CI = [.3, .99], p = .048). The prob-
ability for initiating first drunkenness in students who live
with their two biological parents is .55 times larger. All sig-
nificant results in the total sample are shown in Figure 2.
Table 4. Logistic Regression Analyses of the Intervention Effects on the Incidence Rates of Alcohol, Tobacco, and Cannabis Use and
Drunkenness Experiences.
Effects of the intervention on the incidence of . . .
Hosmer–Lemeshow
test
Pseudo-R2B SE Wald pOR (CI = 95%) χ2 statistic (df)p
Alcohol use 8.5 (8) .387 .038 .13 .27 0.22 .640 1.13 [.67, 1.93]
Tobacco use 11.3 (8) .189 .054 −.02 .32 0.00 .954 1.02 [.55, 1.9]
Cannabis use 4.6 (8) .801 .125 .35 .432 0.67 .411 1.43 [.61, 3.32]
Drunkenness 10.27 (8) .247 .048 −.60 .28 4.53 .033* 0.55 [.32, .95]
Note. The Hosmer–Lemeshow goodness-of-fit test for testing the model validity with χ2 statistic. Pseudo-R2 = model fit index by Nagelkerke; Wald =
Wald test (with one degree of freedom) analyzed the predictive power of the predictor “group” for significance (*p < .05). SE = standard error; OR =
odds ratio; CI = confidence interval.
*p < .05.

Jungaberle and Nagy 9
Results in the Subgroups
Sensitivity analyses. The results of the ANCOVA and logistic
regression analyses are presented in Table 5 and Figure 3 for
all subgroups. Due to the formation of subgroups with
smaller sample sizes, a decrease of statistical power is to be
expected (Kleist, 2007). To determine the discoverable effect
sizes in the subgroup analyses, sensitivity analyses (Faul,
et al., 2007) were conducted. Detectable effect sizes of d >
.146 were calculated for the subgroup of students at academic
high school (ANCOVA with two groups (df = 1), four covari-
ates, 1 – β = .80, and α = .05. The detectable effect sizes for
students at vocational high schools were d > .51. For the sub-
groups of boys and the age group of 14- to 15-year-olds, a
statistical power of 1 − β > .88 was calculated (two tailed),
Pr(Y = 1 / X = 0), H0 = 0.5 1, α = .05, RotherX
2= .08, Binomial
× parm π = .46 and .49, for the logistic regression analyses
with OR = .37 (incidence of drunkenness in boys) and OR =
.44 (incidence of drunkenness among 14- to 15-year-olds).
Interactions. Regarding the IG/CG and the subgroups, the
following interactions were found with regard to the study
variables: gender by group for the dependent variable drunk-
enness incidence (Wald = 3.99, df = 1, OR = .74, 95% CI =
[.54, .99], p = .044), school type by group for the dependent
variable drunkenness incidence (Wald = 7.97, df = 1, OR =
.76, 95% CI = [.63, .92], p = .05), 30-day alcohol use preva-
lence (F = 16.31, df = 1, p = .000), 30-day tobacco use preva-
lence (F = 6.29, df = 1, p = .013), 30 days’ cannabis use
prevalence (F = 9.3, df = 1, p = .03), relative alcohol risk
perception (F = 5.76, df = 1, p = .17), and age group by group
for the dependent variable drunkenness incidence (Wald =
3.99, df = 1, OR = .34, 95% CI = [.54, .99], p = .46).
Gender. A significant result in the subgroup gender was found
for the incidence of drunkenness for boys (Wald = 7.1, df = 1,
OR = .37, 95% CI = [.18, .77], p = .008). The probability for
an initiation of the first drunkenness for boys in the interven-
tion group is 0.37 times higher than in the control group.
Figure 2. Comparison of the IG and CG with respect to significant changes before and after the intervention.
Note. IG = intervention group; CG = control group.
*p ≤ .05. **p ≤ .01

10 SAGE Open
Table 5. Results of ANOVA and Logistic Regression Analyses: Intervention effects in the Subgroups of Boys, the Age Group of 14- and
15-Year-Old, and School Types.
FAMD Partial η2W B SE
OR (CI =
95%) df p
Sex: Male
Incidence drunkenness 7.1 −.99 .37 .37 [.18, .77] 1 .008**
Age group: 14 to 15 years old
Incidence drunkenness 8.1 −.81 .29 .44 [.25, .77] 1 .004**
School type: Academic high school
Incidence drunkenness 7.1 −.99 .37 .37 [.18, .77] 1 .008**
30 days’ prevalence
alcohol 1.44 −.99 .022 1 .017*
cannabis 6.83 −2.51 .176 1 .014*
Substance knowledge 12.95 2.37 .036 1 .000**
School type: Vocational high school
Relative alcohol-risk perception 5.57 .46 .024 1 .024*
Note. AMD = adjusted mean difference of the groups (IG-CG), partial η2 = adjusted effect size of the predictor group, W = Wald test: tests the predictive
power of the predictor “group” for significance (*p < .05); SE = standard error; OR = odds ratio; CI = 95% confidence interval; IG = intervention group;
CG = control group.
*p < .05. **p < .01.
Figure 3. Comparison of IG and CG for males, age 14 to 15 and academic high school concerning significant changes before and after
the intervention.
Note. IG = intervention group; CG = control group; LTP = lifetime prevalence.
*p < .05. **p ≤ .01.

Jungaberle and Nagy 11
School type. The highest number of significant intervention
effects was found in the subgroup of academic high school
students. After the course, students in the intervention group
consumed significantly less alcohol (F = 1.44, df = 1, p =
.017) and less cannabis (F = 6.83, df = 1, p = .014) in the past
30 days, and knew more about psychoactive substances (F =
12.95, df = 1, p = .000). Participants of REBOUND in voca-
tional high schools reported a higher relative risk perception
of alcohol than the control group (F = 5.571, df = 1, p = .019).
Age groups. The younger REBOUND participants (age group
of 14 to 15 years) showed a significantly lower incidence of
drunkenness (Wald = 8.1, df = 1, OR = .44, 95% CI = [.25,
.77], p = .004). The probability for initiating first drunken-
ness was 0.44 times higher than in the control group.
Discussion
Summary of Results
The present study examined effects of the life skills and risk
education program REBOUND on substance use, risk per-
ception, and knowledge about psychoactive substances. The
total sample and the subgroups gender, age, and school types
were examined.
In the total sample, fewer students initiated first drunken-
ness compared with controls. The decreased initiation of
drunkenness directly after the course can be interpreted as a
first indication of risk competent dealing with alcohol and
not just a change of cognitions. Similar effects were found
previously in life-skills programs (Botvin, et al., 2003;
Faggiano, et al., 2005; Poduska, et al., 2008).
After the course, the personal risk perception in smokers
and the general tobacco and cannabis risk perception in non-
users decreased, but remained on a high level. However, no
increase in the incidence of tobacco or cannabis consumption
was found, although the average age of onset in tobacco con-
sumption in Germany currently is at 14.3 years (Bundeszentrale
für gesundheitliche Aufklärung, 2012); this is when the
course starts. This decrease in risk perception could be inter-
preted as iatrogenic effect, which would need to be cared for
in a program revision. The alternative view is seeing it as a
desirable reduction in unrealistic high-risk perception (pre-
sumably acquired through scare tactics). Such a dispropor-
tionate risk perception may trigger a trivialization of risks
when actually being confronted with the reality of drug use
(see switching-risk effect; Zwick, 2005). This is particularly
true for cannabis with its polarized public image. Given the
political changes in cannabis politics, further research is
needed to determine such effects. In providing science-based
trustworthy and balanced information, REBOUND supports
an autonomous and more adequate assessment of risks in this
age group.
Subgroup analyses showed that male students in the inter-
vention group were less likely to initiate first drunkenness.
This indicates increased risk competence with regard to the
control of alcohol. Because the risk for alcohol-related prob-
lems in males is generally increased (Lev-Ran, et al., 2013),
this finding represents an important component of the effec-
tiveness of REBOUND. The program also had a stronger
influence on the consumption behavior of academic high
school students (Gymnasium) compared with vocational
high school students: The 30-day prevalence was reduced for
both alcohol and cannabis use in this school type.
Students in academic high school after the course show
increased substance knowledge, indicating better handling of
objectively verifiable information. This education bias con-
firms findings in previous studies showing that students with
stronger cognitive and linguistic capacities benefit more
from prevention courses (e.g., O’Malley et al., 2006).
In vocational high school, REBOUND contributed to
increasing the relative alcohol risk perception, suggesting a
reduction of the self-serving bias (underestimation of one’s
own vulnerability compared with others; Weinstein, 1980)
and indicating an increase in critical reflection and success-
ful risk pedagogy.
After the course, the subgroup of younger participants
(14- to 15-year-olds) had a lower incidence of initiating
drunkenness. This is a favorable effect considering the
increased risk of a pathological development connected with
early onset of alcohol consumption (Chou & Pickering,
1992; Laucht & Schmid, 2007; Lynskey, et al., 2003; Robins
& Przybeck, 1985).
The results of the subgroup analyses allow for a more pre-
cise interpretation of the findings in the total sample. The
increase of knowledge in the total sample predominantly
originates from the participants in academic high school, and
the decline in incidence of drunkenness is due to the younger
and male course participants.
Methodological Strengths and Weaknesses of the
Study
The methodological strengths of the study are the multilevel
analyses, the propensity score matching, and the control of
covariates. These strengths can counteract the disadvantages
of the study design. In this first evaluation of REBOUND,
only a small number of schools participated. This weakness,
as well as school internal organizational issues, made only a
partial randomization at the class level possible. Due to this
pilot design, a spillover effect could have happened
(Angelucci & Di Maro, 2015), that is, a possible exchange of
the course content between IG and CG, and was not con-
trolled for. Randomization problems are well known regard-
ing educational settings. Due to the small sample size and the
strong dropout, only small effects (0.10 < f < 0.24) could be
found. More students with consumption experiences dropped
out from the CG or were excluded in the matching proce-
dure. Parts of the systematic dropout can be explained by low

12 SAGE Open
compliance of those students who did not participate in the
course.
Strengths of the subgroup analyses are the examination of
the interactions between subgroups and intervention in rela-
tion to the dependent variables, the statistical controlling of
the influence of covariates, and the Bonferroni–Holm adjust-
ment. The sample equivalence was checked for all study
variables, and intervention effects were examined only for
those variables for which no differences between IG and CG
were found before the course. This meant only a few crite-
rion variables could be examined.
Conclusion and Practical Implications
There is a need for life skills programs that include substance
use addressing young people in late adolescence and early
adulthood (Foxcroft & Tsertsvadze, 2011; Thomas & Perera,
2006; Tobler, et al., 2000). In this phase, universal school-
based programs typically are working with inhomogeneous
groups with significant differences regarding resilience and
risk behaviors.
Conceptually, the program is geared toward medium-
(after 2 years) and long-term (over 5 years) self-regulation
(Bandura, 2005). First improvements of risk competence were
found concerning knowledge and initiation of drunkenness in
the total sample; in subgroups, frequency of cannabis use was
reduced. The negative effects on risk perception occurred with
no negative effects on consumption measured after the course.
REBOUND intends to stimulate a qualitative change of func-
tions and social contexts of consumption (Franzkowiak, 2001;
Koller, 2004), which has yet to be proved. Long-term effects
of REBOUND as well as the long-term impact of the mea-
sured risk perception changes should be analyzed in follow-up
studies. Future evaluation with a larger school sample should
include program effects on life skills and resilience variables.
Teacher and parent outcomes should be added.
Risk education components of the program should more
intensely consider the issue of tobacco use (Faggiano, et al.,
2005), which is also strongly related to the consumption of
cannabis (Fahrenkrug, 1998).
The first REBOUND program evaluation shows encour-
aging results for prevention programs that are risk compe-
tence–based and not grounded on scare tactics or abstinence
orientation. The results as well as an analysis of a large quali-
tative data set also provided information for the adaptation
and improvement of the program to version 1.0 in 2014.
REBOUND is now ready for a new randomized controlled
trial (RCT) study and a multimethod evaluation of long-term
effects. It is a promising self-regulation and risk competence
program for age groups starting at 14 years which are
engaged in multiple risk behaviors.
Declaration of Conflicting Interests
The author(s) declared no potential conflicts of interest with respect
to the research, authorship, and/or publication of this article.
Funding
The author(s) received no financial support for the research and/or
authorship of this article.
References
Ajzen, I. (1991). The theory of planned behavior. Organizational
Behavior and Human Decision Processes, 50, 179-211.
Angelucci, M., & Di Maro, V. (2015). Program evaluation
and spillover effects (Policy Research Working Paper, No.
WPS7243). Washington, DC: World Bank Group.
Bandura, A. (2005). The primacy of self-regulation in health pro-
motion. Applied Psychology: An International Review, 54,
245-254.
Becker, M. H. (1974). The health belief model and personal health
behavior. Health Education Monographs, 2, 234-508.
Botvin, G. J., Griffin, K. W., Paul, E., & Macaulay, A. P. (2003).
Preventing tobacco and alcohol use among elementary school
students through life skills training. Journal of Child &
Adolescent Substance Abuse, 12(4), 1-17.
Brookes, S. T., Whitley, E., Peters, T. J., Mulheran, P. A., Egger, M.,
& Davey Smith, G. (2001). Subgroup analyses in randomised
controlled trials: Quantifying the risks of false-positives and
false-negatives. Health Technology Assessment (Winchester,
England), 5(33), 1-56.
Bundeszentrale für gesundheitliche Aufklärung. (2012).
Drogenaffinitätsstudie Jugendlicher in der Bundesrepublik
Deutschland 2011. Der Konsum von Alkohol, Tabak und ille-
galen Drogen: aktuelle Verbreitung und Trends [Study on drug
affinity of youth in the Federal Republic of Germany 2011. The
consumption of alcohol, tobacco and illicit drugs: Current dis-
semination and trends]. Köln, Germany: Author.
Chou, S. P., & Pickering, R. P. (1992). Early onset of drinking
as a risk factor for lifetime alcohol-related problems. British
Journal of Addiction, 87, 1199-1204.
Cujpers, P. (2002). Effective ingredients of school-based drug pre-
vention programs: A systematic review. Addictive Behaviors,
27, 1009-1023.
Décieux, P. J., Heinz, A., & Jacob, R. (2011). Online-Erhebungen
mit EFS Survey: Ein Überblick über grundlegende Funktionen
des Erhebungsprogrammes [Online surveys with EFS sur-
vey: An overview on basic functions of the survey platform].
Munich, Germany: GRIN Verlag.
Faggiano, F., Vigna-Taglianti, F., Versino, E., Zambon, A., Borraccino,
A., & Lemma, P. (2005). Universal school-based prevention for
illicit drug use. Cochrane Database of Systematic Reviews, 12,
Article CD003020. Retrieved from http://onlinelibrary.wiley.
com/doi/10.1002/14651858.CD003020.pub3/abstract
Fahrenkrug, H. (1998). Risikokompetenz - eine neue Leitlinie für
den Umgang mit “riskanten Räuschen?” [Risk Competence—A
new guideline for dealing with “risky intoxication”?].
Suchtmagazin, 24(3), 23-27.
Faul, F., Erdfelder, E., Lang, A.-G., & Buchner, A. (2007).
G*Power 3: A flexible statistical power analysis program
for the social, behavioral, and biomedical sciences. Behavior
Research Methods, 39, 175-191.
Foxcroft, D., & Tsertsvadze, A. (2011). Universal school-based
prevention programs for alcohol misuse in young people.
Documentation Centre on Drug Use, 5. Retrieved from http://
www.drugsandalcohol.ie/15098

Jungaberle and Nagy 13
Franzkowiak, P. (1996). Risikokompetenz—Eine neue Leitorientierung
für die primäre Suchtprävention? [Risk competence – A new
guideline for primary prevention?] Neue Praxis, 26, 409-425.
Franzkowiak, P. (2001). Risikokompetenz in der Suchtprävention.
Potentiale und Probleme [Risk competence in substance misuse
prevention. Potentials and problems]. Prävention, 24, 102-104.
Hawks, K., & Scott, K. (2002). Prevention of psychoactive sub-
stance abuse: A selected review of what works in the area of
prevention. Geneva, Switzerland: World Health Organization.
Ho, D., Imai, K., King, G., & Stuart, E. A. (2011). MatchIt:
Nonparametric preprocessing for parametric causal inference.
Journal of Statistical Software, 42(8), 1-28. doi:10.18637/jss.
v042.i08
Holm, S. (1979). A simple sequentially rejective multiple test pro-
cedure. Scandinavian Journal of Statistics, 6, 65-70.
Jessor, R., & Jessor, S. L. (1977). Problem behavior and psycho-
social development: A longitudinal study of youth. San Diego,
CA: Academic Press.
Kleist, P. (2007). Vorsicht bei Subgruppenanalysen! [Caution with
subgroup analyses!] Schweiz Med Forum, 7, 794-799.
Koller, G. (2004). Risikokompetente Suchtprävention:
Voraussetzung und Wegbegleitung für Jugendliche, Rausch- und
Risikobalance zu entwickeln [Risk competent prevention: Premise
and companion for young people to develop balance concerning
intoxication and risk]. AJS-Informationen, 47(2), 14-19.
Kröninger-Jungaberle, H., Nagy, E., von Heyden, M., DuBois,
F., & the REBOUND Participative Development Group.
(2014). REBOUND: A media-based life skills and risk edu-
cation programme. Health Education Journal, 74, 705-719.
doi:10.1177/0017896914557097
Kulis, S., Yabiku, S. T., Marsiglia, F. F., Nieri, T., & Crossman, A.
(2007). Differences by gender, ethnicity, and acculturation in
the efficacy of the keepin’ it real model prevention program.
Journal of Drug Education, 37, 123-144.
Laucht, M., & Schmid, B. (2007). Früher Einstieg in den Alkohol-
und Tabakkonsum–Indikator für eine erhöhte Suchtgefährdung?
[Early onset in alcohol and tobacco consumption—An indica-
tor of increased risk for addiction?]. Zeitschrift für Kinder- und
Jugendpsychiatrie und Psychotherapie, 35(2), 137-143.
Lazarus, R. S., & Folkman, S. (1984). Stress, appraisal, and cop-
ing. New York, NY: Springer.
Lev-Ran, S., Le Strat, Y., Imtiaz, S., Rehm, J., & Le Foll, B.
(2013). Gender differences in prevalence of substance use dis-
orders among individuals with lifetime exposure to substances:
Results from a large representative sample. The American
Journal on Addictions, 22(1), 7-13. doi:10.1111/j.1521-
0391.2013.00321.x
Löwe, U., Braun, E., & Kisser, R. K. (2008). Tackling injuries
among adolescents and young adults: Strategy and framework
for action. Vienna: KfV (Austrian Road Safety Board).
Lynskey, M. T., Heath, A. C., Bucholz, K. K., Slutske, W. S.,
Madden, P. A., Nelson, E. C., . . . Martin, N. G. (2003).
Escalation of drug use in early-onset cannabis users vs co-twin
controls. Journal of the American Medical Association, 289,
427-433.
Malwald, E., & Reese, A. (2000). Effektivität suchtpräventi-
ver Lebenskompetenzprogramme - Ergebnisse deutscher
Evaluationsstudien [Effectiveness of substance abuse-preven-
tive life-skill programs. Results from German evaluation stud-
ies]. Sucht aktuell, 7, 8-12.
Muthén, B. O., & Satorra, A. (1995). Complex sample data in struc-
tural equation modeling. Sociological Methodology, 25, 267-316.
Muthén, L. K., & Muthén, B. O. (1998). M Plus statistical analysis
with latent variables: User’s guide. Los Angeles, CA: Author.
O’Malley, P. M., Johnston, L. D., Bachman, J. G., Schulenberg,
J. E., & Kumar, R. (2006). How substance use differs among
American secondary schools. Prevention Science, 7, 409-420.
doi:10.1007/s11121-006-0050-5
Poduska, J. M., Kellam, S. G., Wang, W., Brown, C. H., Ialongo, N.
S., & Toyinbo, P. (2008). Impact of the Good Behavior Game,
a universal classroom-based behavior intervention, on young
adult service use for problems with emotions, behavior, or
drugs or alcohol. Drug and Alcohol Dependence, 95(Suppl. 1),
S29-S44.
Raudenbush, S. W. (2011). Optimal design software for multi-
level and longitudinal research (Version 3.01). Retrieved from
http://hlmsoft.net/od/
Resch, F. (1999). Entwicklungspsychopathologie des Kindes- und
Jugendalters: Ein Lehrbuch [Developmental psychology
in childhood and adolescence]. Beltz, Psychologie-Verlag-
Union.
Robins, L. N., & Przybeck, T. R. (1985). Age of onset of drug
use as a factor in drug and other disorders. NIDA Research
Monograph, 56, 178-192.
Rosenbaum, P. R., & Rubin, D. B. (1983). The central role of the
propensity score in observational studies for causal effects.
Biometrika, 70(1), 41-55. doi:10.1093/biomet/70.1.41
Rothwell, P. (2005). Subgroup analysis in randomised controlled
trials: Importance, indications, and interpretation. Lancet, 365,
176-186.
Silbereisen, R. K., & Reese, A. (2001). Substanzgebrauch: Illegale
Drogen und Alkohol. Opladen, Germany: Leske + Budrich.
Thomas, R. E., & Perera, R. (2006). School-based programmes for
preventing smoking. Documentation Centre on Drug Use, 3.
Retrieved from http://www.drugsandalcohol.ie/16635/
Tobler, N. S., Roona, M. R., Ochshorn, P., Marshall, D. G., Streke,
A. V., & Stackpole, K. M. (2000). School-based adolescent
drug prevention programs: 1998 meta-analysis. The Journal of
Primary Prevention, 20, 275-336.
van Etten, M. L., & Anthony, J. C. (2001). Male–female differences
in transitions from first drug opportunity to first use: Searching
for subgroup variation by age, race, region, and urban status.
Journal of Women’s Health & Gender-Based Medicine, 10,
797-804.
Weinstein, N. D. (1980). Unrealistic optimism about future life events.
Journal of Personality and Social Psychology, 39, 806-820.
Zwick, M. M. (2005). Risk as perceived by the German pub-
lic: Pervasive risks and “switching” risks. Journal of Risk
Research, 8, 481-498. doi:10.1080/13669870500064150
Author Biographies
Henrik Jungaberle is a prevention and drug researcher. His focus
is interdisciplinary research in the fields of health education and
psychoactive drugs, including life-time drug use pathways and non-
dependent use of psychedelics and entactogenes.
Ede Nagy is a psychologist and research associate at the University
of Heidelberg, Institute of Medical Psychology. His recent research
activities include the evaluation of interventions in health promo-
tion, prevention, and systemic family therapy training.