Ciocanel2017 Article EffectivenessOfPositiveYouthDe

J Youth Adolescence (2017) 46:483–504
DOI 10.1007/s10964-016-0555-6
EMPIRICAL RESEARCH
Effectiveness of Positive Youth Development Interventions:

A Meta-Analysis of Randomized Controlled Trials
Oana Ciocanel1 • Kevin Power2,3 • Ann Eriksen1 • Kirsty Gillings4
Received: 16 May 2016 / Accepted: 26 July 2016 / Published online: 12 August 2016
Ó Springer Science+Business Media New York 2016
Abstract Positive youth development is thought to be development interventions than high-risk youth. The
essential to the prevention of adolescent risk behavior and studies examined had several methodological flaws, which
the promotion of thriving. This meta-analysis examined the weakened the ability to draw conclusions. Substantial
effects of positive youth development interventions in progress has been made in the theoretical understanding of
promoting positive outcomes and reducing risk behavior. youth development in the past two decades. This progress
Ten databases and grey literature were scanned using a needs to be matched in the intervention literature, through
predefined search strategy. We included studies that the use of high-quality evaluation research of positive
focused on young people aged 10–19 years, implemented a youth development programs.
positive youth development intervention, were outside
school hours, and utilized a randomized controlled design. Keywords Positive youth development Sexual health
Twenty-four studies, involving 23,258 participants, met the Substance use Mental health Academic achievement
inclusion criteria and were included in the analysis. The Prosocial behavior Meta-analysis
impact of the interventions on outcomes including behav-
ioral problems, sexual risk behavior, academic achieve-
ment, prosocial behavior and psychological adjustment Introduction
were assessed. Positive youth development interventions
had a small but significant effect on academic achievement During adolescence, young people must negotiate complex
and psychological adjustment. No significant effects were and inter-related biological, cognitive, emotional and
found for sexual risk behaviors, problem behavior or pos- social-cultural changes. Problem behaviors including sub-
itive social behaviors. Intervention effects were indepen- stance misuse, risky sexual behaviors, school dropout,
dent of program characteristics and participant age. Low- antisocial attitudes and violence increase during early
risk young people derived more benefit from positive youth adolescence (e.g., Dryfoos 1990) and can lead to greater
likelihood of negative behaviors into adulthood. Early
adolescence is therefore an important time to intervene and
& Kirsty Gillings influence the trajectory of an individual’s cognitive, social,
[email protected] emotional and cultural development and their risk behav-
1
Sexual Health and Blood Borne Virus Team, Kings Cross
ior. Positive youth development interventions encompass
Hospital, NHS Tayside, Clepington Road, Dundee DD3 8EA, these two overarching aims. Such programs need to be
UK sensitively designed to capitalise upon the plasticity that
2
School of Natural Sciences, University of Stirling, Stirling, characterizes this developmental period and address the
UK unique challenges inherent in adolescence. As Tolan et al.
3
Area Psychological Therapies Service, NHS Tayside, 7 (1995, p 579) note, ‘‘Intervention designs informed by this
Dudhope Terrace, Dundee DD3 6HG, UK model emphasise developmentally appropriate compo-
4
Psychology Department, Stratheden Hospital, NHS Fife, nents, sensitivity to the impact of timing of intervention,
Cupar, Fife KY15 5RR, Scotland and evaluation of the impact on future development as well
123
484 J Youth Adolescence (2017) 46:483–504
as curtailing or preventing the target symptoms’’. It is the economic costs for adolescents, their families, communi-
above features that may contribute to the potential success ties and services (Scott et al. 2001; Hoffman and Maynard
of positive youth development programs. This paper sys- 2008; Parsonage 2009).
tematically evaluates the effectiveness of such programs,
exploring both their impact upon risk factors (e.g., sexual Positive Youth Development
behaviour, substance use, antisocial behaviour, depression)
and positive outcomes (e.g., academic achievement, Public health experts believe that reducing the prevalence of
prosocial behaviour, psychological adjustment) using modifiable behavior patterns could result in reduced public
meta-analyses. costs, improved overall well-being and health throughout
adolescence and adulthood (e.g., Steinberg 2004). Given the
Adolescent Health Risk Behaviors as a Public Health observed clustering of risk behaviors, it has been suggested
Problem that interventions should take a broad approach and address
multiple problems and their common determinants simul-
Adolescence is a critical period during which many health- taneously (e.g., Bonell et al. 2007; Hawkins et al. 1999;
risk behaviors are initiated, including substance use, sexual Kipping et al. 2012). Positive youth development interven-
risk and antisocial behavior (Degenhardt et al. 2008; World tions aim to address the common determinants of adolescent
Health Organisation 2014). Despite the recent decline in multiple health risk behaviors. As the term implies, positive
some risk behaviors (e.g., smoking and unprotected sex), youth development interventions do not focus solely upon a
young people are still more likely to engage in risky pathology or deficit model. While accepting that a holistic
behaviors than adults over 25 (Eaton et al. 2006). Risky understanding of adolescent development must include
sexual behavior in young people under 25 results in unin- adverse aspects, the positive youth development approach
tended pregnancies and sexually transmitted infections takes the perspective that all young people have inherent
(e.g., Department of Health 2011). In addition, between 6 strengths (e.g., Damon 2004; Roth and Brooks-Gunn 2003a)
and 13 % of adolescents smoke regularly, drink alcohol and that development takes place within relational systems
and use illicit drugs (e.g., Connell et al. 2009; Gunning (e.g., Lerner 2006; Overton 2013). The aims of positive
et al. 2010; McVie and Bradshaw 2005). Aggression and youth development interventions are to support adolescents
antisocial behavior in young people are also problematic, to acquire a sense of competence, self-efficacy, belonging
with approximately a quarter of young people found to and empowerment (e.g., Bowers et al. 2010), thus promoting
carry a weapon and 19 % found to have attacked someone positive behavior and reducing the likelihood of risk
with the intent of seriously hurting them (e.g., Beinart et al. behavior. Effective positive youth development interven-
2002). Taken together, these findings highlight the fre- tions should optimize the interaction between the unique
quency of initiation of health risk behaviors in adolescence. strengths of the individual and their contextual resources
In addition to their frequent occurrence in adolescence, (e.g., healthy relationships with adults, access to commu-
behaviors such as substance use, risky sexual behavior, nity-based activities; Spencer and Spencer 2014). The
smoking and antisocial behavior tend to cluster (Hale and potential advantages of positive youth development inter-
Viner 2012; Jackson et al. 2012a, b; Mistry et al. 2009; ventions have led to major investment in many countries.
Wiefferink et al. 2006). Individuals engaging in one risky For example, in the UK, millions of pounds have been
behavior are more likely to engage in others (DuRant et al. invested in youth development interventions (Scottish
1999). Such behaviors are thought to share common bio- Government 2009) as public health officials see these as
logical and environmental determinants (Beyers et al. essential in promoting the health and well-being of young
2004; O’Connell et al. 2009; Resnick et al. 1997), which people (e.g., HM Government 2010). Therefore, it is
likely shape the development of multiple risk-taking. For important to understand the impact of these interventions
example, substance use before the age of 16 has been and their mechanisms of action.
positively associated with early sexual initiation, poor Although the philosophy and aims of positive youth
contraceptive use, violence and delinquency (Bellis et al. development have been well articulated (e.g., Benson et al.
2008; Hawkins et al. 1999; Parkes et al. 2007). Adolescent 2006; Damon 2004; Larson 2000; Lerner 2006), the core
risk-taking often continues into adulthood, with consequent components of an effective positive youth development
negative outcomes such as poor physical, mental and program remain unclear (Brooks-Gunn and Roth 2014).
sexual health; substance abuse and addiction; poor educa- There is considerable diversity in the operational features
tional and occupational achievement; future morbidity and and activities that currently characterize positive youth
premature mortality (Biglan 2004; Fergusson et al. 2007; development programs. In the first literature review,
Flory et al. 2004; Mirza and Mirza 2008). Youth risk- Catalano et al. (2002) defined positive youth development
taking behavior therefore has substantial personal and as developing cognitive, social, emotional, behavioral and
123
J Youth Adolescence (2017) 46:483–504 485
moral bonding competences; self-efficacy; prosocial on sexual health and have also had mixed findings (e.g.,
behavior; a belief in the future; a clear and positive iden- Catalano et al. 2004; Gavin et al. 2010; Shepherd et al.
tity; self-determination and spirituality. Roth and Brooks- 2010). Some positive youth development reviews have also
Gunn (2003b) believed that for a program to be considered reported reductions in violence and drug use (e.g., Catalano
positive youth development it must (a) foster program et al. 2002; Durlak et al. 2010; Roth and Brooks 2003b).
goals, such as confidence, competence, character, connec- These findings demonstrate the lack of clarity and consis-
tions and caring; (b) provide young people with opportu- tency in the existing literature on the impact of positive
nities and experiences at school, at home and in the youth development programs.
community so that they can develop their interests and The observed variance in findings across reviews of
talents and build new skills and competencies; (c) create a positive youth development interventions may be
supportive atmosphere in which young people can develop explained both the variety in program components and
bonds with the adults involved in delivering the program as differences in review methodology. Existing reviews differ
well as with the other program participants. Positive youth in their inclusion criteria, data pooling methods and the
development interventions also need to be stable and long outcomes examined. Reviews have generally focused on
lasting, so that the participants have sufficient time to form either health or social/behavioral outcomes and some were
and benefit and from positive relationships. The mecha- non-systematic or were limited to a narrative approach
nisms by which positive youth development interventions (e.g., Catalano et al. 2002; Gavin et al. 2010; Roth and
are hypothesized to work are equally diverse. The active Brooks 2003b). Other reviews have limited inclusion to
ingredients are thought to include (1) engaging young programs that have evidence of effectiveness (e.g., Cata-
people in structured and productive activities thus diverting lano et al. 2002). All previous reviews have included a mix
them from unhealthy behavior (Roth et al. 1998), (2) pro- of randomized controlled trials, quasi-experiments and
viding adolescents with additional resources and time to even non-experimental studies. This raises the possibility
develop knowledge, skills and social networks (Pettit et al. of systematic bias affecting the reviews, potentially con-
1997) and (3) addressing risk factors such as low self- flating the effectiveness of positive youth development
esteem, poor educational attainment and low aspirations for interventions or reducing their ability to detect genuine
the future by developing protective factors such as social effects.
and emotional competencies (Catalano et al. 2002). These
examples demonstrate that positive youth development Contributions of this Review
interventions vary considerably in structural and process
features. The mixed findings of previous reviews, the range of
positive youth development interventions and the wide-
Prior Reviews of Interventions to Promote Positive spread interest and investment in positive youth develop-
Youth Development ment has motivated this review. This review aims to
address some limitations of previous reviews by adopting
Despite extensive investment, the effectiveness of positive an inclusive approach. It differs from previous reviews in
youth development interventions in reducing risky behav- several ways. First, this review included all possible ran-
ior and promoting positive behavior is uncertain. Positive domized controlled trials, which provide stronger evidence
youth development programs have been examined in meta- of a program’s impact, since randomized controlled trials
analyses (e.g., Durlak et al. 2010; Shepherd et al. 2010) and have the highest possible internal validity. Second, a sys-
narrative reviews (e.g., Catalano et al. 2002; Clarke et al. tematic strategy was used to identify all possible published
2015; Gavin et al. 2010; Roth and Brooks 2003b). Some of and unpublished studies that provided evidence of program
these reviews have shown that positive youth development impact, regardless of their findings (positive, negative or no
interventions are effective, with others yielding mixed or effects). We included both published and unpublished
inconclusive findings. Investigations focusing on social documents to avoid review bias, since studies with signif-
outcomes have shown positive effects for academic icant results are published whereas those with non-signif-
achievement and cognitive variables and social skills (e.g., icant results remain unpublished (i.e., the ‘‘file-drawer
Catalano et al. 2004; Clarke et al. 2015; Durlak and effect’’; Rosenthal 1979). Third, the impact of positive
Weissberg 2007; Durlak et al. 2010). However, Zief et al. youth development interventions was explored across a
(2006) in a review on after-school programs that combined range of health, social and behavioral outcomes. Fourth,
recreation and/or youth development programming with this review systematically assessed study quality according
academic support services, found that there was limited to established guidelines (i.e., the Cochrane Collaboration
impact on academic and behavioral outcomes. Systematic Risk of Bias Tool). Finally, the evidence on particular
reviews examining health outcomes have focused primarily outcomes across the studies was pooled using meta-
123
analytic methods (where appropriate) to maximize power These goals included bonding, resilience, social, emo-
to detect intervention effects. tional, cognitive, behavioral or moral competence, self-
determination, spirituality, self-efficacy, positive identity
development, belief in the future, recognition for positive
Purpose of the Current Study behavior, opportunities for pro-social behavior and proso-
cial norms.
The purpose of this review and meta-analysis was to syn-
thesize evidence on the effectiveness of positive youth Outcome
development interventions in young people aged
10–19 years. As many outcomes cluster because they Any health or non-health outcome with at least two mea-
shared the similar risk and protective factors, the effects of surements points was included. Outcomes were measured
positive youth development interventions on multiple in several categories; social and emotional skills, positive
health, social and behavioral outcomes were explored. social behavior, mental health issues, sexual risk behavior
These included substance use, sexual risky behavior, psy- and academic performance. Programs that resulted in both
chological adjustment, prosocial behavior and academic significant and non-significant changes in the outcomes
performance. We also examined whether the variation in compared to the control conditions were included, and
the effects was moderated by study, intervention and par- programs that only focused on knowledge and attitude
ticipant characteristics. changes were excluded. Self-reports, official records, and
third party (i.e., parents, teachers) measures (both validated
or not validated) were eligible for inclusion. Table 1 details
Method the outcomes used in this meta-analysis.
The PRISMA guidelines for the conduct of systematic Setting

reviews and meta-analyses (Moher et al. 2009) were fol-
lowed for the planning, conduct and reporting. Out-of-school programs were the focus of the intervention.
These included all activities targeting young people that
Study Inclusion Criteria were delivered regularly either in a community or a school-
based setting outside normal school hours. Interventions
The inclusion criteria were formulated in accordance with that were delivered primarily during school hours were
the PICOS approach and included the following: excluded, as these were the focus of a recent review
(Durlak et al. 2011). For studies that included more than
Population one intervention, only those interventions that focused on
out-of-school programming as the main intervention were
The focus of our intervention was young people. The included. The following criteria were used to determine the
majority of participants (at least 75 %) at the pre-test were main intervention: (a) if the author identified that out-of-
10–19 years of age. As our interest was in preventive school programming was the main intervention or (b) if the
approaches, programs for specific populations such as youth report gave out-of-school interventions a higher impor-
with learning or physical disabilities were excluded. How- tance in relation to other interventions. Our review also
ever, studies which targeted young people on the basis of excluded interventions that focused on family functions
their pre-existing risk behavior or other forms of targeting, and so were targeted at parents/other family members as
such as young people at a high risk of teenage pregnancy, well as young people. Programs were only included if they
students from poor socioeconomic status families and stu- focused primarily on young people and out-of-school
dents with poor grades, were included in this review. programming to minimize the potential moderating effects
of other variables (i.e. the effect of the program on parents,
Intervention effects of in-school components) on intervention impact.
Positive youth development programs were defined as Design

those that involved voluntary education to promote positive
development (National Youth Agency 2007). Specifically, Studies were eligible if they were randomized controlled
programs needed to address at least one of the 12 positive trials and used a control condition to evaluate positive
youth development goals formulated by Catalano et al. youth development interventions. Waiting list or no treat-
(2002) across social domains, including school, community ment, treatment as usual or alternative treatments were all
and family, or more than one goal in a single domain. considered valid control conditions. Other inclusion criteria
123
Table 1 Outcome categories used in this meta-analysis

Outcome category and subcategory Description Examples N (%)*
Behavioural adjustment
Positive social behaviours Ability of a person to get along Social competence, prosocial 7 (29.1 %)
with others behaviours
Problem behaviour Inability to adequately control Conduct problems (e.g.. 16 (66.6 %)
behaviour in social situations aggression) and substance use
(e.g. alcohol, marijuana or
tobacco)
Psychological adjustment 8 (33.3 %)
Emotional distress Internalised mental health issues Depression, negative affect 3 (12.5 %)
Self-perceptions Thoughts about and perceptions of Self-control, self-efficacy, self- 6 (25 %)
self concept, academic and social
self-efficacy
Academic/school outcomes 11 (45.8 %)
Academic achievement Students’ success in meeting short- Grades, graduation, enrolment in 10 (41.6 %)
or long goals in education university, course failure
Academic adjustment School attendance, school liking 5 (20.8 %)
Sexual health outcomes Sexual risk behaviours and Sexual initiation, contraceptive 11 (45.8 %)
pregnancy rates use; Pregnancy and birth rates
* N (%) number of studies that included outcome measures within each outcome category
were sufficient information to calculate effect sizes and Specifically, keyword searches included variations in
publication in English between 1985 and 2015. ‘‘children and young people’’, ‘‘positive youth develop-
ment’’, ‘‘youth work’’, ‘‘after-school’’ and (‘‘intervention’’
Search Strategy OR ‘‘outcome’’ OR ‘‘program’’ OR ‘‘treatment’’). Terms
are available on request from the first author.
A comprehensive literature search was performed to iden-
tify all published and unpublished studies that met the Study Selection and Data Extraction
above inclusion criteria (see Fig. 1). Four main procedures
were used to identify eligible studies: (a) An electronic Study selection was first performed in two main stages using
search of 10 databases (Applied Social Sciences Index and a screening instrument. First, the titles and abstracts were
Abstracts, Medline, PsycINFO, Embase, CINAHL Plus, scrutinized and excluded as appropriate. Relevant papers
ERIC, Social Services Abstracts, Cochrane Central Register were then retrieved in full and assessed against the inclusion
of Controlled Trials in the Cochrane Library, BibioMap and criteria. Documents that were potentially eligible were fur-
Trials Register of promoting health interventions); (b) a ther reviewed to decide upon the final inclusion. Disagree-
search of relevant registers and youth work-related websites ments were resolved through discussion, and where
(e.g. National Youth Agency, National Council for Volun- necessary, studies were reviewed again. A data extraction
tary Youth Services (NCVYS) Publications, 4-H); (c) ref- form with five sections was used in the initial review to
erence list screening from previous reviews (e.g., Dickson extract information from all articles that met the inclusion
et al. 2013; Harden et al. 2006; Morton and Montgomery criteria. These were (a) general study characteristics (author,
2011) and articles identified through electronic databases; year of publication, country of origin), (b) population char-
(d) information from researchers on unpublished or ongoing acteristics (number of participants, age, gender, grade level
articles or to clarify reports identified through other sources. and risk level), (c) intervention characteristics (dosage, set-
The electronic searches were initially conducted between ting, format, components), (d) methodological characteris-
September and December 2014 and re-run prior to the tics (sample sizes, characteristics of the control group,
analyses in July 2015. design, attrition, follow-up period, intention to treat versus
Search terms were developed based on previous reviews treatment on the treated analysis and outcome measures) and
and empirical studies (e.g., Dickson et al. 2013; Harden (e) statistical data needed for effect size calculations. For
et al. 2006; Morton and Montgomery 2011) to reflect the multiple publications from the same cohort, only studies with
agreed population criteria (young people), intervention up-to-date or comprehensive data were included. Where data
criteria (positive youth development) and research methods. on study methods or results were missing, authors were
123
Records identified through database Additional records identified

searching through other sources
(n = 15,452) (n =2,789)
Records after duplicates removed and screened by title

(n =12,500) Full-text articles excluded, with
reasons
(n =328)
• Not age 10-19: (n=56)

Records screened by abstract • Ineligible intervention: (n=64)
(n =669) • Classroom-based program:
(n=54)
• Other setting (e.g. residential,
clinical): (n=21)
Full-text articles assessed for • Ineligible research design:
eligibility (n=72)
(n =352) • No full text:(n=14)
• Not English language: (n=2)
• Other population (i.e. youth
offenders, orphans, foster care,
Studies included in quantitative substance users, ADHD):
synthesis (meta-analysis) (n=28)
(n = 24 ) • Does not include outcomes of
interest (i.e. HIV,
victimisation, BMI, eating
disorders): (n=9)
• Effect size not calculable:
(n=4)
Fig. 1 CONSORT diagram
contacted with a request to supply the information. In studies Social Sciences (SPSS 21.0) was used to analyze the
where the requested information was unavailable due to data descriptive data.
loss or non-response, where possible the data was included in
the meta-analysis. Effect Size Calculations
Risk Bias Assessment For continuous outcomes, we calculated the standardized

mean differences, or the difference between two means
Two authors independently assessed each study’s divided by their pooled standard deviations. To avoid effect
methodological quality using the Cochrane Collaboration size underestimation (Field 2001) we applied Hedge’s
Risk of Bias Tool (Higgins et al. 2011). A third author g correction, which is usually recommended for a sample
assessed more than half the papers. All disagreements were size lower than 20 (Borenstein et al. 2009). For dichoto-
discussed until a consensus was reached. Seven domains mous outcomes, we calculated an odds ratio and then
were scored with high, low or unclear risk of bias: transformed these (using meta-analysis software) to
Sequence generation, allocation concealment, participant g statistics to allow for across study comparisons (Boren-
blinding and personnel, outcome assessment blinding, stein et al. 2009). When the studies failed to report means,
incomplete outcome data, selective outcome reporting and standard deviations or proportions, effect sizes were cal-
other issues (i.e., baseline differences among groups). Each culated using a t test, F-statistic or p value and sample size
domain was scored as -1 for high risk, 0 for unclear risk (Borenstein et al. 2009). All effect sizes were coded in
and 1 for low risk. These scores were then summed to which positive values indicated favorable intervention
provide an overall quality score, which ranged from -6 to effects such as lower pregnancy rates or less substance use,
6. Higher values signified a lower bias risk. with values of 0.20 considered small, 0.50 as medium and
0.80 as large (Cohen 1988). When a study had multiple
Statistical Analyses measures for the same outcome, an overall effect size was
calculated by averaging the individual effect sizes. There-
Comprehensive Meta-Analysis Program (version 3) was fore, a single mean effect size per study was calculated for
used to carry out the meta-analyses. Statistical Package for each outcome category, which ensured statistical
123
independence (Lipsey and Wilson 2001). For each outcome, 2789 citations through website searches and searching
a separate analysis was performed to examine the inter- lists of previous reviews and studies. After the removal of
vention effect at both post-intervention and follow-up. Post- duplicates, 669 abstracts were screened for relevance. In
intervention effect sizes were calculated for the assessment total, the full text of 352 studies were obtained and
nearest in time to the completion of the program. When a screened for eligibility. Of these, 328 did not meet the
study reported multiple follow-up assessments for a partic- inclusion criteria and were excluded. Twenty-four studies
ular outcome, the longest follow-up period was selected to were included in the final meta-analysis. Nine trials were
examine the robustness of the intervention. reported in multiple companion publications (See Table 2
For clustered randomized trials in which the study had for details).
adjusted for a clustering effect, the analysis results were
imputed to calculate the effect sizes. Conversely, for Characteristics of Included Studies and Programs
studies that did not correct for potential clustering prior to
effect size calculation, we corrected for design effect using Design
the guidelines of Higgins and Green (2009). A random-
effects model was used in our statistical analyses due to the Twenty programs were conducted in the USA and the
heterogeneity between studies in target population, inter- remaining four were conducted in Croatia, Ireland, UK and
ventions employed and outcomes assessed (Hedges and New Zealand. Fourteen were published in peer-reviewed
Vevea 1998). All effect sizes were weighted prior to any journals, eight were technical reports and two were dis-
analysis by multiplying the values with the inverse of their sertation projects. Publication dates ranged from 1992 to
error variance (Lipsey and Wilson 2001). This method 2014, with most studies being published after 2000 (75 %).
ensured that larger studies contributed more to the effect All 24 studies employed randomized controlled designs.
sizes and were given more weight in the analyses. Fifteen used students as the randomization unit; seven used
schools, classes or communities and the remaining two
Statistical Heterogeneity used a combination. Seven studies compared the treatment
group with usual care groups such as regular sex education
Statistical heterogeneity between the studies was assessed (e.g., O’Donnell et al. 2002) or standard alcohol and drug
using the Q statistic and the I2 statistic. A significant education programs (e.g., Komro et al. 2008; Perry et al.
Q rejects the homogeneity null hypothesis and indicates 1996), seven used an alternative treatment and the
whether the effect sizes varied more across the studies than remaining nine used no treatment or wait lists as compar-
that expected from the sampling error alone (Borenstein isons. Twelve studies reported high attrition rates. The
et al. 2009). I2 (Higgins and Thompson 2002) shows the sample sizes ranged from 30 to approximately 5812. In
heterogeneity percentage across the studies (0 % = none, most studies, data were gathered through self-reports. Five
25 % = low, 50 % = moderate, 75 % = high; Higgins studies had data from school records, parents or teachers. A
and Thompson 2002). summary of studies included in the meta-analyses can be
found in Table 2.
Publication Bias
Participants
Finally, the presence of publication bias was assessed using
funnel plots (Sterne and Egger 2001) and Begg and Egger The total participant number randomized across the 24
tests (Begg and Mazumdar 1994). Funnel plots measure studies was 23,258. The mean age at baseline ranged from
effect size against study size, and when there is no evi- 10 to 16. Young people included in the programs attended
dence of publication bias these plots display studies sym- elementary schools (12 %), middle schools (37.5 %), high
metrically around the pooled effect size. The Begg and schools (25 %) or a mixture of grade levels (25 %). The
Egger tests measure the extent of the funnel plots asym- predominant race studied was African American (58.3 %),
metry (with p \ 0.05 indicating the presence of statistically followed by Caucasian (37.5 %) and Native American
significant publication bias). (4 %). Most studies included mixed-sex samples, with
three studies focusing exclusively on females. Fifteen
studies focused on at-risk students, six focused low-risk
Results students and three included both. The identifiers for at-risk
populations included students from low-income back-
Figure 1 summarizes the search and selection process. grounds (n = 12 studies), students of racial or ethnic
The literature search identified 15,452 citations from the minority background (n = 5 studies), and students with
electronic bibliographic database searches and a further low academic achievements (n = 6 studies).
123
Table 2 Characteristics of included studies (N = 24)a
490
1st Author Program name Location Baseline sample Risk level/at- Intervention type Duration Setting Follow-up Outcomes
(year) (age, % female) risk identifier (format) period
123
Allen Teen Outreach Multiple US 695 (12–18 years), High % low Community service and 12 months School Pre-post School failure, teen pregnancy and
(1997) Program cities 85 % income classroom-based grounds and (9 months) school suspensions
discussions about community
service experiences centres
Bonell Teens and England, UK 449 (13–14 years), High % low self- Mentoring, skills 18–20 weeks Pre-school Pre-post. Pregnancy, contraceptive use, dislike
(2013) Toddlers 100 % esteem, school training (group) nurseries (22 weeks of school, self-esteem
disengagement, and
etc. 12 months)
Bird (2014) Leadership and South 86 (11–14 years), n/a High % low Social skills training 16 weeks School Pre-post English and Math grades, academic
Young Carolina, income, poor (individual and group) grounds (5 months) self-efficacy, subjective well-being,
Professionals US grades social self-efficacy, negative and
(LYP) positive affect.
Carter Project Venture US 397 (11 years), n/a Low Classroom based and 12 months School Pre-post (6, Alcohol use
(2007) outdoor experiential grounds, 18 months)
learning, adventure community
camps and community centres
oriented activities
Deane Project K Multiple 1.092 (14–15 years), High % low self- Recreation, mentoring, 14 months Community Pre-post Self-efficacy (social, academic, and
(2012) regions, 47 % efficacy community service centres (immediately career decision), academic
New after and achievement
Zealand 12 months)
Dolan Big Brothers Ireland 164 High % low Mentoring, recreation 12 months Community Pre-post Misconduct, substance use (alcohol
(2011) Big Sisters (10–14 years),51 % income, low centres (10 months, and cannabis use), emotional well-
(BBBS) of self-esteem, 15 months being, school liking, scholastic
Ireland poor social and efficacy, prosoical behaviour,
skills 21 months) academic performance
Feinberg Siblings Are Multiple US 174 (10–11 years), Low Social skills training 12 weeks School Pre-post. Academic performance, aggressive
(2013) Special cities 100 % (group) grounds, (4 weeks, behaviour, depression, prosocial
home 12 months) behaviour, social competence, self-
control
Gottfredson All Stars Baltimore, 447 Low and high % Recreation, skills 96 days School Pre-post Drug use, disruptive behaviour,
(2010) US (mean = 12.2 years) minority, low training, academic grounds (96 days) aggression, delinquency
46 % income support (group)
Grossman The Summer Multiple US Cohort 2: 1.635 High % minority, Work experience, life Approx. School Pre-post (42 for Academic performance, school
(1992): Training and cities (14–15 years), low income skills training, school 295 h over grounds and Cohort 3 and dropout rates, contraceptive use,
Cohort 2 Education Cohort 3: 1.591 year support 2 summers community 54 months for high school graduation rates
and 3 Program (14–15 years), centres Cohort 2)
(STEP)
Grossman Big Brothers Multiple US 1.107 (10–16 years), High % minority Mentoring 18 months School Pre- post Alcohol and drug use, antisocial
(1998) Big Sisters cities 37.6 % and low income grounds and (18 months) behaviour, global self-worth, social
(BBBS) community acceptance and self-confidence
Hahn (1994) The Quantum Multiple US 250 (14–17 years), High % low Mentoring, community Up to School Pre-post Academic achievement, substance
Opportunities cities 51 % income service, youth 60 months grounds and (12 months use, delinquency behaviour
Program development activities, community after program
educational activities ended)
Table 2 continued
1st Author Program name Location Baseline sample (age, Risk level/at-risk Intervention type Duration Setting Follow-up Outcomes
(year) % female) identifier (format) period
Karcher Developmental Texas, US 30 (10–11 years), 60 % Mixed Mentoring, recreation, 12 months Community Pre-post Connectedness to family, school and
(2002) mentoring academic centres (12 months) to future, academic achievement
program
Karcher Developmental Texas, US 73 (9–11 years) 44 % Mixed Mentoring, recreation, 6 months Community Pre-post Connectedness to school and parents,
(2005) mentoring social skills training centres (6 months) academic achievement
program
Komro Project Chicago, US 5.812 (12 years), 50 % High % low Curricula, family 36 months School Pre-post (6, 18 Alcohol and drug use
(2008) Northland income intervention and grounds, and
youth-led community community 30 months)
service projects centres,
family
LeCroy The Go Grrls Arizona, US 118 (mean = 13.5), Low Academic assistance, 12 weeks School Pre-post Self-liking and competence; self-
(2004) Program 100 % Social skills training grounds (immediately efficacy, assertiveness
(group) after)
LoSciuto Across Ages Philadelphia, 729 (11–12 years), High % low Mentoring, community 36 months School Pre-post (9, 20, Well-being, substance use
(1996) US 53 % income service, classroom- grounds, and
based life skills community 32 months)
curriculum centres
Maxfield Quantum Multiple US 1.069 teens High % low Mentoring, academic Up to School 48 months Delinquency behaviour, alcohol and
(2003)** Opportunities cities (14–15 years) income support, case 60 months grounds and drug use, high school grades,
Program management (mixed) community achievement test scores.
centres
Monahan Choices New York, 351 (9–15 years) High % at risk of Abstinence education Wave 1 and Community Pre-post Sexual initiation
(2011) Enhanced US teen pregnancy training, youth 2: centres (average
developmental 12 weeks; 10 weeks)
activities Wave 3:
8 weeks
O’Donnell Reach for New York, 255 (mean = 12), High % minority, Community service, 24 months School Pre-post Violent behaviour, sexual activity
(2002) Health US low income classroom health grounds and (45 months) (i.e. recent sex)
curriculum community
centres
Perry (1996) Project Minnesota, 2.351 (11 years), Low Curricula, family 36 months School Pre-post Alcohol use
Northland US 48.7 % intervention and grounds, (2.5 years)
youth-led community community
service projects centres,
family
Philliber Carrera program Multiple US 1.163 (13–15 years), High % low Varied (i.e. job clubs, 3 years Community Pre-post Delinquency behaviour,
(2001)* cities 55 % income academic assistance, centres (36 months) contraceptive use, sexual activity,
family life, recreation) pregnancy and birth rates,
substance use
Reyna Reducing the Arizona and 734 (14–19 years), Mixed Sex education Between 2 School Pre-post Sexual activity (i..e. number of sexual
(2014) Risk New York, 55 % curriculum, and grounds (3 months, partners, number of unprotected sex
US experiential activities 3 weeks 6 months, acts)
to build skills, 12 months)
academic support
491
123
Interventions
and Vilella-Velez 1992a, b); Grossman 1998 (Rhodes et al. 2002, 2005; Tierney et al. 1995), O’Donnell et al. 2002 (O’Donnell et al. 1999; O’Donnell et al. 2003), Maxfield et al. 2003 (Schirm
et al. 2003; Schirm 2004; Stigler 2006, Rodrı́guez-Planas 2010) and Perry 1996 (Perry 2000; Komro et al. 2001; Perry et al. 2002; Stigler et al. 2006; Perry et al. 2007). *For this program, we
reported in the publication by Philliber et al. (2002). **For this program, we calculated effect sizes based on the data in the Maxfield et al. (2003) instead of Stigler et al. (2006). ***For this
calculated effect sizes based on the data for the full study sample that were available in the unpublished report (Philliber et al. 2001) instead of the information on the smaller sample that was
program, we calculated effect sizes based on the more comprehensive data that were available in the unpublished report (Trenholm et al. 2007) instead of the data reported in the published
The following nine trials were reported in multiple companion publications: Deane 2012, (Qiao and McNaught 2007), Gottfredson et al. 2010 (Cross 2009), Grossman and Sipe 1992 (Walker
Sexual activity, pregnancy and birth
rates, substance use (i.e. cigarette,
Programs were conducted in a range of settings. Specifi-
cally, five studies were conducted in the community, four
alcohol marijuana use) were conducted on school grounds and one program was
conducted in a combined school/family domain. The
Alcohol use majority (n = 15) were delivered in mixed settings, with
Outcomes
five being delivered in a combined school, community and

family settings. Fourteen interventions were conducted in
one geographical locality, with the remaining ten con-
ducted nationally. Interventions varied in duration and
Pre-post (42 to
(12 months)
78 months)
number of sessions. The mean intervention duration was

Follow-up
Pre-post
80 weeks, with studies ranging from 3 to 240 weeks.

period
Seventeen interventions involved at least 20 sessions and

12 had two or three follow-ups. The length of the first
grounds and
community
community
follow-up ranged from immediately after the intervention

grounds,
centres,
centres
family
to 5 years after the intervention. The first follow-up was

Setting
School
School
conducted immediately or within 3 months of the inter-

vention in five studies. Eleven studies had the first follow-
48 months
up after 5–18 months, seven studies involved a second

36 months
Duration
follow-up after 6–20 months and five studies reported a

Up to
third follow-up from 12 to 32 months.

The most common outcome measures examined were
youth-led community
behavioral including problem behaviors (66 %), academic

Abstinence education
curriculum, youth
improvement and school adjustment (45 %) and sexual

intervention and
service projects
Curricula, family
Intervention type
developmental
risk behaviors/pregnancy (45 %). A smaller number of

activities
studies examined the effectiveness of these interventions

(format)
on positive social behaviors (29 %) and psychological

adjustment (33 %). Five interventions were delivered in a
group, two were delivered individually, and the remain-
High % minority,
Risk level/at-risk
ing seventeen combined individual and group interven-

single parent
low-income,
tions. Twenty-one programs were multi-modal and

identifier
involved primary and secondary interventions. Three

Low
single-modal programs provided mentoring, skills train-

ing or academic components. Of the 21 multi-modal
Baseline sample (age,
programs, primary after-school activities covered aca-

1981 (11–13 years)
505 (8–13), 62 %
demic and homework help (n = 8), mentoring (n = 7),

community service projects (n = 9), social or cognitive/
% female)
emotional skill development (n = 16), recreational

activities (n = 6) and job clubs (n = 2). The most
common positive youth development goals reported by
approximately half the programs were pro-social bond-
Milwaukee,
Location
ing, social competence, cognitive competence, emotional

Croatia
US
competence, self-efficacy, self-determination and a belief

in the future.
report (Trenholm et al. 2008)
Families United
Program name
Pregnancy
to Prevent
Northland
Bias Risk
Teenage
Project
Table 2 continued
The bias risk in the studies is summarized in Fig. 2.

Overall, the studies did not provide sufficient information
West (2008)
1st Author
to judge the randomization procedure quality, with 16

Trenholm
(2007)
studies having an unclear rating. Only seven studies

(year)
***
described the randomization sequence and only two
123
Fig. 2 Risk of bias ratings

across included studies
Table 3 Effect sizes by outcome category

Outcome Mean effect size (Hedge’s g) N 95 % Confidence intervals Heterogeneity
Q statistic df I2
Behavioural adjustment
Positive social behaviours 0.04 7 -0.11–0.21 11.84* 6 49.34*
Problem behaviour 0.05 16 -0.00–0.11 20.75 15 27.73
Psychological adjustment
All combined 0.17* 8 0.04–0.31 21.30* 7 67.14
Emotional distress 0.14* 3 -0.002–0.29 1.30 2 0.00
Self-perceptions 0.19* 6 0.02–0.37 19.13* 5 73.87
Academic/school outcomes
Academic achievement 0.22* 10 0.07–0.38 28.30* 9 68.20
Academic adjustment 0.09 5 -0.02–0.20 4.15 4 3.75
Sexual health outcomes 0.05 11 -0.00–0.12 8.41 10 0.00
Significant effects are expressed in boldface Random effects model. N = number of studies containing a measure for each category, g = mean
Hedges’g, CI = 95 % confidence interval; Q statistic = test of heterogeneity; df = degrees of freedom; I2 = proportion of observed dispersion
* Denotes mean effect is significantly different from zero at the .05 level
a
Indicated where heterogeneity is high
indicated how the allocation concealment was conducted. Intervention Effects

As seen in Fig. 2, participants and personnel blinding rates
were found to be at a high risk across all studies. Never- Table 3 shows the mean effect sizes, the 95 % confidence
theless, given the nature of the interventions, the blinding intervals and the corresponding statistics for each outcome
of participants or personnel was often not possible. Only category. Forest plots were created for each of the five
three studies reported a blinding of the outcome assessors. outcome categories. Effect sizes ranged from 0.04 to 0.22,
Attrition bias was high in 13 of the included studies, and despite all being positive (i.e., favoring the interven-
uncertain in two and low risk in nine. The findings in these tion condition), only three were significantly different from
studies may be biased and may not reflect the true effects of zero. Specifically, the analyses indicated significant effects
the intervention as the results may have been influenced by in two areas; academic/school outcomes and psychological
the characteristics of the participants who dropped out of adjustment. The largest positive effect size was found in
the studies. Reporting bias was assessed as low risk in 18 academic achievement (g = 0.22), with the lowest effect
studies, as these papers appeared to have provided results size found in positive social behaviors (g = 0.04). Inter-
on the expected outcomes. Three studies were assessed as vention effects were based only on post-intervention data.
high risk, as they had incomplete information on the Follow-up data were only combined for two outcome cat-
expected outcomes. egories—psychological adjustment and academic
123
Study name Outcome Statistics for each study Hedges's g and 95% CI
Hedges's Lower Upper
g limit limit p-Value
Feinberg 2013 Combined 0.040 -0.249 0.329 0.787
Philliber 2001 Combined 0.031 -0.165 0.227 0.757
Trenholm 2007 SUBSTANCE USE -0.121 -0.316 0.075 0.226
Gottfredson 2010 Combined -0.025 -0.221 0.171 0.803
West 2008 SUBSTANCE USE 0.023 -0.101 0.147 0.716
Dolan 2011 Combined 0.132 -0.284 0.548 0.534
Allen 1997 CONDUCT PROBLEMS 0.518 0.142 0.895 0.007
Grossman_1992_Cohort 2 SUBSTANCE USE 0.112 -0.165 0.389 0.428
Grossman_1992_Cohort 3 SUBSTANCE USE -0.010 -0.320 0.300 0.950
Hahn 1994 CONDUCT PROBLEMS 0.148 -0.525 0.821 0.666
Maxfield 2003 Combined -0.075 -0.206 0.056 0.264
Grossman 1998 Combined 0.160 -0.748 1.068 0.730
LoSciuto 1996 SUBSTANCE USE 0.198 0.002 0.393 0.048
Perry 1996 SUBSTANCE USE 0.156 0.017 0.294 0.027
Komro 2008 SUBSTANCE USE 0.006 -0.056 0.068 0.850
Carter 2006 SUBSTANCE USE 0.190 -0.041 0.422 0.106
0.047 -0.010 0.104 0.107
-1.00 -0.50 0.00 0.50 1.00
Favours Control Favours intervention
Fig. 3 Effect sizes for problem behaviors
achievement—when the category included at least two problems and substance use measures into the same anal-
studies. For the remaining outcomes, the follow-up effects ysis. Eight studies investigated the impact of interventions
were reported for each study. on substance use, two measured conduct problems and six
measured both conduct and substance use problems.
Heterogeneity The results of the meta-analysis indicated no statistically
significant differences between groups (g = 0.04; 95 %
Heterogeneity was found in studies that reported the fol- CI = -0.01, 0.10; ns; see Fig. 3). The statistical hetero-
lowing outcomes; self-perception, academic achievement geneity across the studies was neither important nor sig-
and sexual risk behaviors—which indicated the likelihood nificant (Qtotal = 20.75; I2 = 27.73 %; ns). Only one study
of moderating variables. However, there was no evidence (Dolan et al. 2011) measured the problem behavior after
of heterogeneity in studies that reported on problem 21 months but showed no significant effects (g = 0.04;
behavior, academic adjustment or sexual health outcomes. 95 % CI = -0.30, 0.39; ns).
Behavioral Adjustments Positive Social Behavior
Problem Behavior This category was reported in seven studies and included
outcomes such as getting along with others, social com-
Sixteen studies, which included 52 effect sizes, were petence and prosocial behavior. Measures included teach-
averaged to show the intervention effects on problem ers, parents and self-reports, the latter being the most
behavior. To increase the statistical power and because frequently used. The results showed no significant statis-
moderator analyses showed no significant differences in tical differences between groups at post-treatment
intervention effects between conduct problems and sub- (g = 0.04; 95 % CI = -0.11, 0.21; ns; see Fig. 4) and the
stance use (t = 0.87; ns), we decided to pool all conduct effects’ heterogeneity was moderate but significant
pp
LeCroy 2004 POSITIVE SOCIAL BEHAVIOURS 0.249 -0.115 0.614 0.180
Feinberg 2013 POSITIVE SOCIAL BEHAVIOURS 0.158 -0.103 0.420 0.236
Gottfredson 2010 POSITIVE SOCIAL BEHAVIOURS -0.133 -0.328 0.061 0.180
Karcher 2002 POSITIVE SOCIAL BEHAVIOURS 0.792 0.017 1.568 0.045
Karcher 2005 POSITIVE SOCIAL BEHAVIOURS 0.264 -0.268 0.797 0.331
Dolan 2011 POSITIVE SOCIAL BEHAVIOURS -0.239 -0.576 0.099 0.166
Grossman 1998 POSITIVE SOCIAL BEHAVIOURS 0.002 -0.184 0.188 0.983
0.048 -0.116 0.211 0.568
-1.00 -0.50 0.00 0.50 1.00
Fig. 4 Effect sizes for positive social behaviors
123
pp
LeCroy 2004 SELF-PERCEPTIONS 0.059 -0.237 0.354 0.697
Feinberg 2013 EMOTIONAL DISTRESS 0.020 -0.242 0.282 0.881
Bonell 2013 SELF-PERCEPTIONS 0.248 -0.022 0.517 0.072
Bird 2014 Combined 0.246 -0.175 0.667 0.253
Dolan 2011 SELF-PERCEPTIONS 0.149 -0.055 0.354 0.153
Deane 2012 SELF-PERCEPTIONS 0.452 0.313 0.590 0.000
Grossman 1998 SELF-PERCEPTIONS 0.003 -0.161 0.167 0.971
LoSciuto 1996 EMOTIONAL DISTRESS 0.197 0.001 0.392 0.049
0.177 0.040 0.313 0.011
-1.00 -0.50 0.00 0.50 1.00
Fig. 5 Effect sizes for psychological adjustment
pp
Feinberg 2013 ACADEMIC ACHIEVEMENT 0.159 -0.146 0.464 0.307
Gottfredson 2010 ACADEMIC ACHIEVEMENT -0.047 -0.239 0.145 0.630
Bird 2014 ACADEMIC ACHIEVEMENT 0.078 -0.343 0.499 0.716
Karcher 2002 ACADEMIC ACHIEVEMENT 0.781 0.007 1.555 0.048
Karcher 2005 ACADEMIC ACHIEVEMENT 0.497 -0.040 1.034 0.070
Dolan 2011 ACADEMIC ACHIEVEMENT 0.169 -0.026 0.364 0.089
Allen 1997 ACADEMIC ACHIEVEMENT 0.477 0.272 0.683 0.000
Hahn 1994 ACADEMIC ACHIEVEMENT 0.672 0.312 1.032 0.000
Maxfield 2003 ACADEMIC ACHIEVEMENT 0.050 -0.089 0.188 0.480
Grossman 1998 ACADEMIC ACHIEVEMENT 0.028 -0.410 0.466 0.900
0.229 0.071 0.387 0.004
-1.00 -0.50 0.00 0.50 1.00
Fig. 6 Effects sizes for academic achievement
(Qtotal = 11.84; I2 = 49.34 %; p \ 0.05). One study (2013) conducted a follow-up at 10 months post-intervention
(Karcher et al. 2002) had an impact on heterogeneity and and reported no lasting significant effects on depression levels
overall effect size, so after excluding this study, hetero- (g = -0.04; 95 % CI = -0.38, 0.30; ns).
geneity was lower (Qtotal = 7.86; I2 = 36.40 %; ns), but
the effect size was smaller (g = 0.01; ns). Only one study Academic/School Outcomes
(Dolan et al. 2011) measured positive social behaviors after
21 months and showed a positive significant effect Academic/school outcomes were reported in 11 studies and
(g = 0.27; 95 % CI = 0.00, 0.55; p \ 0.05). included measures in relation to academic achievement and
academic adjustment. Academic achievement outcomes
Psychological Adjustment/Internalizing Behavior were measured in 10 studies using school records, teacher,
parent and self-report. Academic adjustment was reported
Eight studies, including 17 effect sizes, were averaged to in five studies based on self-report. A significant difference
show the intervention effects on psychological adjustment. in academic achievement outcomes was found between the
The meta-analysis indicated a small but significant treatment groups after the intervention (g = 0.22; 95 % CI = 0.07,
effect (g = 0.17; 95 %CI = 0.04, 0.31; p \ 0.05; see Fig. 5). 0.38; p \ 0.05; see Fig. 6). Low but significant hetero-
The homogeneity analysis indicated a moderate degree of geneity was found across the studies (Qtotal = 28.30; -
statistical heterogeneity (Qtotal = 27.16; I2 = 67.15 %; I2 = 68.20 %; p \ 0.01). On average, the positive youth
p \ 0.01). The results suggested that exposure to positive development interventions had a 0.22 standard deviation
youth development interventions improved psychological improvement in scholastic performance relative to the
adjustment compared to the control condition, which was control groups. Generally, however, the analysis found no
equivalent to a 0.17 standard deviation in magnitude. Three significant overall intervention effect on academic adjust-
studies examined long-term intervention effects on self-per- ment (g = 0.09; 95 % CI = -0.02, 0.20; ns) and the effect
ception and found that this did not change over time heterogeneity was low and non-significant (Qtotal = 4.15;
(g = 0.23; 95 % CI = 0.05, 0.42; p \ 0.01). Feinberg et al. I2 = 3.75 %; ns). Three studies measured academic/school
123

Philliber 2001 Combined 0.144 -0.046 0.335 0.138
Bonell 2013 Combined -0.221 -0.778 0.336 0.436
Trenholm 2007 Combined -0.015 -0.351 0.321 0.930
Monahan 2011 SEXUAL RISK BEHAVIOUR 0.009 -0.267 0.285 0.949
Reyna 2014 SEXUAL RISK BEHAVIOUR 0.070 -0.116 0.256 0.461
Allen 1997 PREGNANCY 0.464 -0.354 1.283 0.266
O'Donnell 2002 Combined 0.355 -0.113 0.823 0.137
Grossman_1992_Cohort 2 Combined 0.002 -0.140 0.144 0.978
Grossman_1992_Cohort 3 Combined 0.002 -0.143 0.147 0.978
Hahn 1994 PREGNANCY 0.363 -0.002 0.728 0.051
Maxfield 2003 Combined 0.061 -0.090 0.213 0.427
0.056 -0.009 0.121 0.093
-1.00 -0.50 0.00 0.50 1.00
Fig. 7 Effects sizes for sexual risk behaviors
outcomes that also indicated no lasting significant effects Table 4 shows that the only moderation effect found
(g = 0.07; 95 % CI = -0.05, 0.20; ns). was for the youth risk level moderator in relation to posi-
tive social behavior. Interventions delivered to low or
Sexual Health Outcomes mixed risk youth (k = 4; g = 0.23; p \ 0.05) were more
likely to produce a significant positive effect than inter-
As the moderator analyses showed no significant differ- ventions applied to high-risk young people. One study
ences in the intervention effects between sexual risk (Karcher et al. 2002) significantly contributed to this result
behaviors and pregnancies (t = 0.32, ns), all measures (g = 0.79; p \ 0.05). No other significant moderation
were pooled. A meta-analysis of the 11 studies on sexual effects were found. However, several trends emerged.
health outcomes showed no statistically significant differ- Intervention characteristics that showed small significant
ences between the positive youth development intervention trends were delivered in community-based settings (k = 3;
and the control group (g = 0.05; 95 % CI = -0.00, 0.12; g = 0.30; p \ 0.05, psychological adjustment outcome),
ns), with non-significant heterogeneity found across the were in mixed settings (k = 6; g = 0.29; p \ 0.05; aca-
studies (Qtotal = 8.41; I2 = 0 %; ns). All but two studies demic achievement outcome), or lasted for more than
(Bonell et al. 2013; Trenholm et al. 2007) showed a posi- 1 year (k = 4; g = 0.20; p \ 0.05; psychological adjust-
tive effect size though none were statistically significant. ment; and k = 4; g = 0.25; p \ 0.05; academic achieve-
Only one study measured effect over a longer follow-up ment outcomes). Sample characteristics with significant
(Bonell et al. 2013) and a small, non-significant effect was trends included interventions designed for middle-school
reported (g = 0.03; 95 % CI = -0.39, 0.45; ns) (Fig. 7). youth (k = 4; g = 0.16; p \ 0.05, psychological adjust-
ment outcome) and high-school youth ((k = 3; g = 0.36;
Subgroup Analyses p \ 0.05; academic achievement outcome).
The 24 studies were grouped by one of the five primary
Homogeneity analyses were performed for each set of intervention types; academic and skills training (n = 6),
effect sizes (see Table 4). Significant heterogeneity was recreation (n = 2), community service projects (n = 2) and
present in three of the outcome variables (positive social mixed (i.e. life skills/recreation, life skills/community or
behaviors, psychological adjustment and academic education/work experience) (n = 4). The effect sizes did not
achievement) so potential moderators of these effects were differ significantly between the groups, suggesting that the
examined. Effects for problem behavior and sexual health primary intervention type did not have a strong influence on
outcomes were not heterogeneous, so moderators were not the overall positive youth development intervention effect
examined. Twenty-one analyses were performed with on any outcome. However, mentoring interventions showed
seven potential moderators across the three outcome vari- a significant impact in relation to psychological adjustment
ables. These moderators had three intervention character- outcome (k = 5; g = 0.21; p \ 0.01).
istics (setting, duration and type), two sample
characteristics (youth risk level and age) and one study Publication Bias
characteristic (publication). Emotional distress and self-
perceptions were pooled into psychological adjustments Publication bias was not detected, as the funnel plot shapes
because the groups were too small for separate analyses. were symmetrical for all analyses (data not shown). The
123
Table 4 Moderators of effect sizes

Study feature Positive social behaviours Psychological adjustment Academic achievement
k g (95 % CI) k g (95 % CI) k g (95 % CI)
Setting
Community 2 .03 (-.41, .42) 3 .30* (.12, .48) 2 .30 (-.11, .72)
School 2 .02 (-.32, .37) 2 .12 (-.16, .42) 2 .00 (-.37, .37)
Mixed 3 .11 (-.18, .40) 2 .09 (-.11, .30) 6 .29* (.07, .51)
Duration
C1 year 5 -.02 (-.20, .15) 4 .20* (.01, .39) 8 .25* (.07, .42)
\1 year 2 .19 (-.08, .47) 3 .17 (-.09, .44) 2 .12 (-.26, .50)
Type
Academic/skills traininga 2 .19 (-.11, .50) 2 .13 (-.20, .47) 2 .12 (-.28, .53)
Community projectsb
Mentoringc 4 .04 (-.20, .29) 5 .21* (.04, .38) 4 .27 (-.04, .59)
Recreationd
Mixed 2 .28 (-.08, .64)
Youth risk level
Low/Mixed 4 .23* (.04, .42) 3 .36* (.03, .63)
High 3 -.08 (-.21, .03) 7 .19* (.01, .37)
Age
Elementary (5) 3 .27 (-.03, .58) 3 .37 (-.01, .77)
Middle (6–8) 2 .00 (-.28, .29) 3 .16* (.01, .31) 2 .00 (-.39, .39)
High-school (9–12) 3 .36* (-.06, .66)
Mixed 2 -.09 (-.37, .19) 3 .09 (-.02, .21) 2 .11 (-.28, .51)
Publication
Published 5 .14 (-.08,.36) 3 .17 (-.07, .42) 5 .28* (.03, .53)
Unpublished 2 -.09 (-.39, .21) 4 .21 (.00, .42) 5 .19 (-.04, .42)
Random effects model used; SE = standard error. Youth risk level moderator eliminated from the analysis for psychological adjustment due to
missingness; Bold-significant results total between Q; Only one of the moderator analyses revealed significant differences (p \ .05);* Significant
effects; Each model evaluates each moderator individually, without controlling for the other listed moderators a (Bird 2014; Feinberg 2013;
Monahan et al. 2011; Komro et al. 2008; LeCroy 2004; Perry et al. 1996; Reyna and Mills 2014; Trenholm et al. 2007; West et al. 2008);
b
(Carter et al. 2007; Gottfredson et al. 2010); c(Bonell et al. 2013; Deane 2012; Dolan et al. 2011; Grossman and Tierney 1998; Karcher et al.
2002; Karcher 2005; LoSciuto et al. 1996); d (Allen et al. 1997; O’Donnell et al. 2002); e (Grossman and Sipe 1992; Hahn 1994; Philliber et al.
2001; Maxfield et al. 2003)
results of the Egger and Begg tests were non-significant for of opportunity in which to potentially alter the course of
asymmetries on all outcomes (psychological adjustment: high-risk behavior. Positive youth development interven-
Egger p = 0.36, Begg p = 1.00; school outcomes: Egger tions aim to prevent the escalation of risk behavior and
p = 0.22, Begg p = 0.18; positive social behaviors: Egger enhance personal growth by drawing upon the strengths of
p = 0.11, Begg p = 0.13; Problem behavior: Egger young people and their contextual assets. Such programs
p = 0.12, Begg p = 0.26; sexual health outcomes: Egger have received significant investment in the past decade.
p = 0.19, Begg p = 0.21). The impact of these programs, as assessed by a number of
previous reviews, is unclear. This lack of clarity may be
accounted for variance in program components, method-
Discussion ological shortcomings in trial evaluations (e.g., measure-
ment of relevant constructs) and differences in review
High-risk behaviors occur frequently in adolescence and methodology. Consequently, there is a lack of specificity in
tend to cluster together. They are associated with a range of the outcomes that positive youth development interven-
adverse physical, psychological and occupational out- tions impact upon and their effective components. The aim
comes, which can persist into adulthood and carry signifi- of this systematic review and meta-analysis was to examine
cant personal, societal and economic costs. Adolescence is the effectiveness of positive youth development interven-
a developmentally sensitive period that presents a window tions on a wide range of outcomes and to offer an updated
123
review of the evidence base. To our knowledge, this is the groups. Cohort studies (e.g., the 4-H Study of Positive Youth
first meta-analysis to examine both published and unpub- Development; Lerner 2006) have already begun to measure
lished randomized controlled trials and to report on the these constructs and the relationships between them. Incor-
effects of such interventions on multiple outcomes. poration of such measures into randomized control
Twenty-four studies were included in the meta-analysis, methodology would provide a robust test of relational
nine of which have not been included in prior reviews. development systems theories upon which many positive
Many of the included studies were affected by systematic youth development programs are based.
bias and attrition rates were high. Positive youth develop- The findings of this review are consistent with prior
ment interventions did not lead to a significant reduction in quantitative (Durlak et al. 2010) and narrative reviews
antisocial/violent behavior, substance abuse or risky sexual (Roth and Brooks 2003b; Catalano et al. 2004; Gavin et al.
behaviors or improve positive social behaviors in compar- 2010) for some outcomes but not others. For example, the
ison to control interventions. However, significant effects for behavioral adjustment evidence contradicts the positive
two outcome variables—psychological adjustment and effects found in Durlak et al. (2010) in relation to problem
academic achievement—were found. Positive youth devel- behavior and positive social behaviors. However, they are
opment interventions were associated with modest but sig- consistent with Zief et al.’s (2006) and Durlak et al.’s
nificant improvements in three areas; self-perception, (2010) findings regarding the non-significant effects on
emotional distress and academic achievement. In particular, drug use. The academic adjustment results were in agree-
significant improvements in self-perception were found to ment with the findings from prior reviews, in that these
reduce emotional distress and improve academic achieve- interventions did not report any beneficial effects (Zief et al.
ment. These effects were small, with many individual studies 2006; Durlak et al. 2010). With regard to sexual health
failing to detect significant effects. This provides partial outcomes, the present results contradict the conclusions
support for the notion that enhancing the assets of adoles- offered by prior reviews, which reported significant effects
cents can support them to thrive academically and manage on sexual risk behavior and pregnancy rates (Shepherd et al.
emotional difficulties. Given that poor academic achieve- 2010; Gavin et al. 2010; Catalano et al. 2002). The con-
ment and low self-perception are associated with lower flicting findings between this review and previous reviews
earnings, poor health (Bonell et al. 2005; Wellings et al. may be accounted for by a number of factors. These include
2001; Emler 2001; Hallfors et al. 2006; Wheeler 2010) and differing inclusion criteria, study design and sample size. A
delinquency (Maguin and Loeber 1996; Donnellan et al. conservative approach was adopted through selection of
2005), improving youth academic achievement and self- randomized controlled designs, utilization of an inclusive
perception could result in improved economic well-being approach to study inclusion (e.g., published and unpub-
and possibly positive health outcomes (Maynard 1996). lished data) to maximize the statistical power and a robust
In addition to exploring which behaviors are modifiable measure of quality assessment. The focus upon both posi-
through positive youth development intervention, it is tive and negative behavioral outcomes reflects an increasing
important to gain an understanding for whom they work and consensus over the last two decades to the benefits for
how programs may be tailored to particular individual and research, policy and practice of integrating promotion and
contextual characteristics. This review addresses this ques- prevention approaches to youth development (e.g., Brooks-
tion in a limited way through the examination of program and Gunn and Roth 2014).
individual moderators. In this review, program characteris- Whilst there has been considerable progress in research
tics were not associated with the strength of program out- on positive youth development, this review highlights the
comes, suggesting that positive youth development limited extent to which conclusions can be drawn about
intervention effects were independent of setting. Program positive youth development intervention effects on ado-
effects were similar regardless of age. However, young lescent health and well-being. These limitations include the
people deemed low-risk were more likely to benefit from lack of high-quality studies and studies delivered outside
positive youth development interventions than high-risk the US. Nevertheless, this lack of high-quality evidence is
youth. It will be important for future research to examine not evidence of a lack of effectiveness. Future investment
both individual and contextual moderators of program and research is therefore required to replicate USA-based
effectiveness. Factors such as socioeconomic status, eth- programs on adolescent populations in different coun-
nicity, access to and engagement with ecological assets (e.g., tries/communities, and the quality of the evaluation studies
community resources) are likely to be important. Such needs to be improved through more rigorous designs and
research would highlight the characteristics of individuals minimization of potential biases. Evaluation of positive
most likely to benefit from existing positive youth develop- youth development programs needs to go beyond measur-
ment interventions and which modifications will be required ing the effects on behavior alone to encompass measures of
to extend the reach of programs to marginalized or high-risk individual positive youth development (e.g., the ‘‘5Cs’’,
123
hope, future expectations; Bowers et al. 2010) and con- inclusion of urban, high-risk youth, predominantly African
textual resources (e.g., parents and other adults, community American samples. Therefore, these results cannot be
activities and institutions). Measuring these constructs is generalized to all adolescents or to all programs outside the
complex and consensus is yet to be reached (Roth and US. Given these limitations, it is not currently possible to
Brooks-Gunn 2016; Tolan 2016). Future studies should give strong recommendations for the use of positive youth
also include adequate sample sizes and repeated follow- development programs. In summary, the findings of this
ups, on the basis that positive youth development is likely a review are encouraging but further research using robust
dynamic process rather than an endpoint (Scales et al. methodology is required.
2016). There is a need for clearer descriptions of the pro-
gram goals, components that contribute to program out-
comes and their implementation features. When more Conclusion
programs have been rigorously evaluated and adequately
reported, it may be possible to pool the evidence to Positive youth development interventions are the focus of
determine program effectiveness. Further research is nee- significant investment, especially in the US and UK. This
ded to understand the critical intervention components that review found that the effects of positive youth development
contribute to the prevention of risk-taking behaviors and to on various outcomes were either non-significant or modest
robustly test novel interventions. in magnitude. Additionally, the evidence base for their
The findings of this review and the conclusions drawn effectiveness was dominated by a high number of USA-
about the positive youth development intervention effects based studies, many of which were poor quality. Never-
need to be interpreted in the context of their limitations and theless, the results of this review support the effectiveness
the current state of evaluation research in this field. A of positive youth development interventions on academic
paucity of rigorous research evidence exists on the impact achievement and psychological adjustment. Low-risk
of positive youth development interventions in adolescents; young people appear to benefit particularly from these
hence, this meta-analysis had only a few studies, some of programs. Substantial progress has been made in theoreti-
which had small sample sizes and validity problems. cal development of positive youth development. Improve-
Although we identified a sample of 24 studies, the number ments are now needed in the way studies are designed,
of studies included in any single analysis was much lower, evaluated and reported so that we can draw more concrete
particularly in the analyses that examined the impact of conclusions as to their real potential in reducing risk
these interventions on certain outcomes. As a result, our behavior and encouraging adolescents to thrive.
power to detect significant differences was reduced. All the
included studies suffered from some internal validity Author Contributions OA, KG and AE conceived of the study. OA
conducted the literature searches, data analysis and drafted the
problems due to methodological flaws. Most studies had manuscript. KG participated in study design, statistical analysis,
high performance bias, selection bias and detection bias quality assessment and helped to draft the manuscript. KP contributed
risk, and therefore, may have overestimated the positive to study design, quality assessment and manuscript revision. AE
youth development intervention effect. However, evidence participated in study design and co-ordination. All authors read and
approved the final manuscript.
suggests that although adequate procedures to ensure ran-
dom sequence allocation and allocation concealment are Conflicts of interest The authors report no conflicts of interest.
often followed, these are frequently underreported (Hill
et al. 2002). In addition, participant blinding is often Ethical Standards As a review of previous research, this study did
not directly involve the use of human participants or animals.
impossible in these types of interventions. There was a
significant study heterogeneity associated with several Ethical Approval No specific ethical approval was sought in the
outcomes, and moderator analyses were unable to explain production of this review.
this variability. Subgroup analyses only included a small
Informed Consent Where relevant, authors of studies included in
number of studies, which led to a low statistical power, so this review were contacted for permission to use their data and to seek
the results may have failed to detect some important sub- unpublished data.
group differences. The non-significant p value might have
been due to low power and not necessarily related to effect
size consistency (Borenstein et al. 2009). As with any References
systematic review, we cannot be sure that our search
identified all possible studies, especially all unpublished References marked with an asterisk indicate studies included in the
meta-analysis.
studies, which have a tendency to report more null effects. *Allen, J. P., Philliber, S., Herrling, S., & Kuperminc, G. P. (1997).
Other limitations include an overreliance on studies con- Preventing teen pregnancy and academic failure: Experimental
ducted in the US, the use of self-report measures and the evaluation of a developmentally based approach. Child
123
Development, 68(4), 729–742. doi:10.1111/j.1467-8624.1997. States: Research findings on evaluations of positive youth
tb04233.x. development programs. Prevention and Treatment, 5(1), 15.
Begg, C. B., & Mazumdar, M. (1994). Operating characteristics of a doi:10.1037/1522-3736.5.1.515a.
rank correlation test for publication bias. Biometrics, 50(4), Catalano, R. F., Berglund, M. L., Ryan, J. A., Lonczak, H. S., &
1088–1101. Hawkins, J. D. (2004). Positive youth development in the United
Beinart, S., Anderson, B., Lee, S., & Utting, D. (2002). Youth at risk? States: Research findings on evaluations of positive youth
A national survey of risk factors, protective factors and problem development programs. The Annals of the American Academy
behaviour among young people in England, Scotland and Wales. of Political and Social Science, 591(1), 98–124. doi:10.1177/
London: Communities that Care. 0002716203260102.
Bellis, M. A., Hughes, K., Calafat, A., Juan, M., Ramon, A., Clarke, A. M., Hussein, Y., Morreale, S., Field, C. A., & Barry, M. M.
Rodriguez, J. A., et al. (2008). Sexual uses of alcohol and drugs (2015). What works in enhancing social and emotional skills
and the associated health risks: A cross sectional study of young development during childhood and adolescence? A review of the
people in nine European cities. BMC Public Health, 8, 155. evidence on the effectiveness of school-based and out-of-school
doi:10.1186/1471-2458-8-155. programmes in the UK. Retrieved from National University of
Benson, P. L., Scales, P. C., Hamilton, S. F., & Semsa, A, Jr. (2006). Ireland Galway. https://aran.library.nuigalway.ie/handle/10379/
Positive youth development: Theory, research, and applications. 4981.
In W. Damon, R. M. Lerner, & R. M. Lerner (Eds.), Handbook Cohen, J. (1988). Statistical power analysis for the behavioural
of child psychology, Vol. 1: Theoretical models of human sciences. Hillside. NJ: Lawrence Earlbaum Associates.
development (6th ed., pp. 894–941). Hoboken, NJ: Wiley. Connell, C. M., Gilreath, T. D., & Hansen, N. B. (2009). A
Beyers, J. M., Toumbourou, J. W., Catalano, R. F., Arthur, M. W., & multiprocess latent class analysis of the co-occurrence of
Hawkins, J. D. (2004). A cross-national comparison of risk and substance use and sexual risk behavior among adolescents.
protective factors for adolescent substance use: The United Journal of Studies on Alcohol and Drugs, 70(6), 943–951.
States and Australia. Journal of Adolescent Health, 35(1), 3–16. doi:10.15288/jsad.2009.70.943.
doi:10.1016/j.jadohealth.2003.08.015. Cross, A. B., Gottfredson, D. C., Wilson, D. M., Rorie, M., &
Biglan, A. (Ed.). (2004). Helping adolescents at risk: Prevention of Connell, N. (2009). The impact of after-school programs on the
multiple problem behaviors. New York: Guilford Press. routine activities of middle-school students: Results from a
*Bird, J. M. (2014). Evaluating the effects of a strengths-based, randomized, controlled trial. Criminology & Public Policy, 8(2),
professional development intervention on adolescents’ aca- 391–412. doi:10.1111/j.1745-9133.2009.00555.x.
demic, social, and emotional outcomes. (Unpublished doctoral Damon, W. (2004). What is positive youth development? The Annals
thesis), University, of South Caroline, United States. Retrieved of the American Academy of Political and Social Science,
from University of South Caroline website http://scholarcom 591(1), 13–24. doi:10.1177/0002716203260042.
mons.sc.edu/etd/2790/?utm_source=scholarcommons.sc.edu%2 *Deane, K. L. (2012). Project K in black & white: A theory-driven &
Fetd%2F2790&utm_medium=PDF&utm_campaign=PDFCover randomised controlled trial evaluation of a youth development
Pages. programme (Unpublished doctoral thesis) University of Auckland,
Bonell, C., Allen, E., Strange, V., Copas, A., Oakley, A., Stephenson, J., New Zealand. Retrieved from University of Auckland website
et al. (2005). The effect of dislike of school on risk of teenage https://researchspace.auckland.ac.nz/handle/2292/19753.
pregnancy: Testing of hypotheses using longitudinal data from a Degenhardt, L., Chiu, W. T., Sampson, N., Kessler, R. C., Anthony, J.
randomised trial of sex education. Journal of Epidemiology and C., Angermeyer, M., et al. (2008). Toward a global view of
Community Health, 59(3), 223–230. doi:10.1136/jech.2004.023374. alcohol, tobacco, cannabis, and cocaine use: Findings from the
Bonell, C., Fletcher, A., & McCambridge, J. (2007). Improving WHO World Mental Health Surveys. PLoS Med, 5(7), e141.
school ethos may reduce substance misuse and teenage preg- doi:10.1371/journal.pmed.0050141.
nancy. BMJ, 334(7594), 614–616. doi:10.1136/bmj.39139. Department of Health (2011). Abortion statistics, England and Wales:
414005.AD. 2010. Retrieved from https://www.gov.uk/government/statistics/
*Bonell, C., Maisey, R., Speight, S., Purdon, S., Keogh, P., Wollny, abortion-statistics-england-and-wales-2010.
I., et al. (2013). Randomized controlled trial of ‘‘teens and Dickson, K., Vigurs, C. A., & Newman, M. (2013). Youth work: A
toddlers’’: A teenage pregnancy prevention intervention com- systematic map of the research literature. Dublin, Department of
bining youth development and voluntary service in a nursery. Children and Youth Affairs. Retrieved from http://www.drugs.ie/
Journal of Adolescence, 36(5), 859–870. doi:10.1016/j.adoles resourcesfiles/ResearchDocs/Ireland/2013/SystematicMapping.
cence.2013.07.005. pdf.
Borenstein, M., Hedges, V., Higgins, J. P., & Rothstein, H. R. (2009). *Dolan, P., Brady, B., O’Regan, C., Russell, D., Canavan, J., &
Introduction to meta-analysis. Chichester, UK: Wiley. Forkan, C. (2011). Big brothers big sisters of Ireland: Evaluation
Bowers, E. P., Li, Y., Kiely, M. K., Brittian, A., Lerner, J. V., & study: Report one: Randomised controlled trial and implemen-
Lerner, R. M. (2010). The five Cs model of positive youth tation report. Child and Family Research Centre, University of
development: A longitudinal analysis of confirmatory factor Galway, Foroige. Retrieved from http://www.foroige.ie/sites/
structure and measurement invariance. Journal of Youth and default/files/big_brother_big_sister_report_3.pdf.
Adolescence, 39(7), 720–735. doi:10.1007/s10964-010-9530-9. Donnellan, M. B., Trzesniewski, K. H., Robins, R. W., Moffitt, T. E.,
Brooks-Gunn, J., & Roth, J. (2014). Promotion and prevention in & Caspi, A. (2005). Low self-esteem is related to aggression,
youth development: Two sides of the same coin? Journal of antisocial behavior, and delinquency. Psychological Science,
Youth and Adolescence, 43(6), 1004–1007. doi:10.1007/s10964- 16(4), 328–335. doi:10.1111/j.0956-7976.2005.01535.x.
014-0122-y. Dryfoos, J. (1990). Adolescents at risk: Prevalence and prevention.
*Carter, S., Straits, K. J., & Hall, M. (2007). Project venture: New York: Oxford University Press.
Evaluation of an experiential, culturally based approach to DuRant, R. H., Smith, J. A., Kreiter, S. R., & Krowchuk, D. P. (1999).
substance abuse prevention with American Indian youth. The The relationship between early age of onset of initial substance
Journal of Experiential Education, 29(3), 397–400. use and engaging in multiple health risk behaviors among young
Catalano, R. F., Berglund, M. L., Ryan, J. A., Lonczak, H. S., & adolescents. Archives of Pediatrics and Adolescent Medicine,
Hawkins, J. D. (2002). Positive youth development in the United 153(3), 286–291. doi:10.1001/archpedi.153.3.286.
123
Durlak, J. A., & Weissberg, R. P. (2007). The impact of after-school *Hahn, A., Leavitt T., & Aaron P. (1994). Evaluation of the Quantum
programs that promote personal and social skills. Chicago, Opportunities Program (QOP). Did the program work? A report
Illinois: Collaborative for Academic, Social, and Emotional on the post secondary outcomes and cost-effectiveness of the
Learning. Retrieved from http://www.uwex.edu/ces/4h/after QOP program (1989–1993). Retrieved from http://files.eric.ed.
school/partnerships/documents/ASP-Full.pdf. gov/fulltext/ED385621.pdf.
Durlak, J. A., Weissberg, R. P., Dymnicki, A. B., Taylor, R. D., & Hale, D. R., & Viner, R. M. (2012). Policy responses to multiple risk
Schellinger, K. B. (2011). The impact of enhancing students’ behaviours in adolescents. Journal of Public Health, 34(suppl 1),
social and emotional learning: A meta-analysis of school-based i11–i19. doi:10.1093/pubmed/fdr112.
universal interventions. Child Development, 82(1), 405–432. Hallfors, D., Cho, H., Brodish, P. H., Flewelling, R., & Khatapoush,
doi:10.1111/j.1467-8624.2010.01564.x. S. (2006). Identifying high school students ‘‘at risk’’ for
Durlak, J. A., Weissberg, R. P., & Pachan, M. (2010). A meta- substance use and other behavioral problems: Implications for
analysis of after-school programs that seek to promote personal prevention. Substance Use and Misuse, 41(1), 1–15. doi:10.
and social skills in children and adolescents. American Journal 1080/10826080500318509.
of Community Psychology, 45(3–4), 294–309. doi:10.1007/ Harden, A., Brunton, G., Fletcher, A., Oakley, A., Burchett, H., &
s10464-010-9300-6. Backhans, M. (2006). Young people, pregnancy and social
Eaton, D., Kann, L., Kinchen, S., Ross, J., Hawkins, J., Harris, W.A., exclusion: A systematic synthesis of research evidence to identify
Lowry, R., McManus, T., Chyen, D., Shanklin, S., Lim, C., effective, appropriate and promising approaches for prevention
Grunbaum, J.A., & Weschler, H. (2006). Youth risk behavior and support. London: Institute of Education.
surveillance—United States 2005. Morbidity & Mortality Hawkins, J. D., Catalano, R. F., Kosterman, R., Abbott, R., & Hill, K.
Weekly Report 55(SS05), 1–108. Retrieved from Centers for G. (1999). Preventing adolescent health-risk behaviors by
Disease Prevention. https://www.cdc.gov/mmwr/preview/ strengthening protection during childhood. Archives of Pedi-
mmwrhtml/ss5505a1.htm. atrics and Adolescent Medicine, 153(3), 226–234. doi:10.1001/
Emler, N. (2001). Self esteem: The costs and causes of low self worth. archpedi.153.3.226.
York, York Publishing Services. Retrieved from https://www.jrf. Hedges, L. V., & Vevea, J. L. (1998). Fixed-and random-effects
org.uk/report/self-esteem-costs-and-causes-low-self-worth. models in meta-analysis. Psychological methods, 3(4), 486–504.
*Feinberg, M. E., Solmeyer, A. R., Hostetler, M. L., Sakuma, K. L., doi:10.1037/1082-989X.3.4.486.
Jones, D., & McHale, S. M. (2013). Siblings are special: Initial Higgins, J. P., Altman, D. G., Gøtzsche, P. C., Jüni, P., Moher, D.,
test of a new approach for preventing youth behavior problems. Oxman, A. D., et al. (2011). The Cochrane Collaboration’s tool
Journal of Adolescent Health, 53(2), 166–173. doi:10.1016/j. for assessing risk of bias in randomised trials. BMJ, 343, d5928.
jadohealth.2012.10.004. doi:10.1136/bmj.d5928.
Fergusson, D. M., Horwood, L. J., & Ridder, E. M. (2007). Conduct Higgins, J. P., & Green, S. (Eds.). (2009). Cochrane handbook for
and attentional problems in childhood and adolescence and later systematic reviews of interventions (Vol. 5.0.2). Chichester:
substance use, abuse and dependence: Results of a 25-year Wiley-Blackwell.
longitudinal study. Drug and Alcohol Dependence, 88, S14–S26. Higgins, J., & Thompson, S. G. (2002). Quantifying heterogeneity in
doi:10.1016/j.drugalcdep.2006.12.011. a meta-analysis. Statistics in Medicine, 21(11), 1539–1558.
Field, A. P. (2001). Meta-analysis of correlation coefficients: A doi:10.1002/sim.1186.
Monte Carlo comparison of fixed-and random-effects methods. Hill, C. L., LaValley, M. P., & Felson, D. T. (2002). Discrepancy
Psychological Methods, 6(2), 161–180. doi:10.1037/1082-989X. between published report and actual conduct of randomized
6.2.161. clinical trials. Journal of Clinical Epidemiology, 55(8), 783–786.
Flory, K., Lynam, D., Milich, R., Leukefeld, C., & Clayton, R. doi:10.1016/S0895-4356(02)00440-7.
(2004). Early adolescent through young adult alcohol and HM Government (2010). Positive for youth: A new approach to
marijuana use trajectories: Early predictors, young adult cross-government policy for young people aged 13–19.
outcomes, and predictive utility. Development and Psy- Retrieved from https://www.gov.uk/government/uploads/sys
chopathology, 16(01), 193–213. doi:10.1017/S095457940404 tem/uploads/attachment_data/file/175496/DFE-00133-2011.pdf.
4475. Hoffman, S. D., & Maynard, R. A. (2008). Kids having kids:
Gavin, L. E., Catalano, R. F., David-Ferdon, C., Gloppen, K. M., & Economic costs and social consequences of teen pregnancy.
Markham, C. M. (2010). A review of positive youth develop- Washington DC: The Urban Institute.
ment programs that promote adolescent sexual and reproductive Jackson, C., Geddes, R., Haw, S., & Frank, J. (2012a). Interventions
health. Journal of Adolescent Health, 46(3), S75–S91. doi:10. to prevent substance use and risky sexual behaviour in young
1016/j.jadohealth.2009.11.215. people: A systematic review. Addiction, 107(4), 733–747.
*Gottfredson, D. C., Cross, A., Wilson, D., Rorie, M., & Connell, N. doi:10.1111/j.1360-0443.2011.03751.x.
(2010). An experimental evaluation of the All Stars prevention Jackson, C., Sweeting, H., & Haw, S. (2012b). Clustering of
curriculum in a community after school setting. Prevention substance use and sexual risk behaviour in adolescence: Analysis
Science, 11(2), 142–154. doi:10.1007/s11121-009-0156-7. of two cohort studies. BMJ Open, 2(1), e000661. doi:10.1136/
*Grossman, J. B., & Sipe, C. L. (1992). Summer Training and bmjopen-2011-000661.
Education Program (STEP): Report on long-term impacts. *Karcher, M. J. (2005). The effects of developmental mentoring and
Retrieved from http://ppv.issuelab.org/resource/report_on_ high school mentors’ attendance on their younger mentees’ self-
long_term_impacts_step_program. esteem, social skills, and connectedness. Psychology in the
*Grossman, J. B., & Tierney, J. P. (1998). Does mentoring work? An Schools, 42(1), 65–77. doi:10.1002/pits.20025.
impact study of the Big Brothers Big Sisters program. Evalu- *Karcher, M. J., Davis, C, I. I. I., & Powell, B. (2002). The effects of
ation Review, 22(3), 403–426. doi:10.1177/ developmental mentoring on connectedness and academic
0193841X9802200304. achievement. School Community Journal, 12(2), 35–50.
Gunning, N., Jotangia, D., Nicholson, S., Ogunbadejo, T., Reilly, N., Kipping, R. R., Campbell, R. M., MacArthur, G. J., Gunnell, D. J., &
Simmonds, N., et al. (2010). Smoking, drinking and drug use Hickman, M. (2012). Multiple risk behaviour in adolescence.
among young people in England in 2009. London: Information Journal of Public Health, 34(suppl 1), i1–i2. doi:10.1093/
Centre for Health and Social Care. pubmed/fdr122.
123
*Komro, K. A., Perry, C. L., Veblen-Mortenson, S., Farbakhsh, K., initiation and sexual activity among urban middle schoolers in
Toomey, T. L., Stigler, M. H., et al. (2008). Outcomes from a the Reach for Health service learning program. Journal of
randomized controlled trial of a multi-component alcohol use Adolescent Health, 31(1), 93–100. doi:10.1016/S1054-
preventive intervention for urban youth: Project Northland 139X(01)00389-5.
Chicago. Addiction, 103(4), 606–618. doi:10.1111/j.1360-0443. O’Connell, M. E., Boat, T., & Warner, K. E. (Eds.). (2009).
2007.02110.x. Preventing mental, emotional, and behavioral disorders among
Komro, K. A., Perry, C. L., Williams, C. L., Stigler, M. H., young people: Progress and possibilities. Washington: National
Farbakhsh, K., & Veblen-Mortenson, S. (2001). How did Project Academies Press.
Northland reduce alcohol use among young adolescents? Anal- O’Donnell, L., Stueve, A., San Doval, A., Duran, R., Haber, D.,
ysis of mediating variables. Health Education Research, 16(1), Atnafou, R., et al. (1999). The effectiveness of the Reach for
59–70. doi:10.1093/her/16.1.59. Health Community Youth Service learning program in reducing
Larson, R. W. (2000). Towards a psychology of positive youth early and unprotected sex among urban middle school students.
development. American Psychologist, 55, 170–183. American Journal of Public Health, 89(2), 176–181. doi:10.
*LeCroy, C. W. (2004). Experimental evaluation of ‘‘Go Grrrl’’ 2105/AJPH.89.2.176.
preventive intervention for early adolescent girls. Journal of Overton, W. F. (2013). A new paradigm for developmental science:
Primary Prevention, 25(4), 457–473. doi:10.1023/B:JOPP. Relationism and relational-developmental systems. Applied
0000048112.10700.89. Developmental Science, 17(2), 94–107.
Lerner, R. M. (2006). Developmental science, developmental sys- Parkes, A., Wight, D., Henderson, M., & Hart, G. (2007). Explaining
tems, and contemporary theories. In W. Damon, R. M. Lerner, & associations between adolescent substance use and condom use.
R. M. Lerner (Eds.), Handbook of child psychology, Vol. 1: Journal of Adolescent Health, 40(2), 180-e1. doi:10.1016/j.
Theoretical models of human development (6th ed., pp. 1–17). jadohealth.2006.09.012.
Hoboken, NJ: Wiley. Parsonage, M. (2009). Diversion: A better way for criminal justice
Lipsey, m W, & Wilson, D. B. (2001). Practical meta-analysis (Vol. and mental health. London: Sainsbury Centre for Mental Health.
49). Thousand Oaks, CA: Sage publications. Perry, C. L., Lee, S., Stigler, M. H., Farbakhsh, K., Komro, K. A.,
*LoSciuto, L., Rajala, A. K., Townsend, T. N., & Taylor, A. S. Gewirtz, A. H., et al. (2007). The impact of Project Northland on
(1996). An outcome evaluation of across ages: An intergener- selected MMPI-A problem behavior scales. Journal of Primary
ational mentoring approach to drug prevention. Journal of Prevention, 28(5), 449–465.
Adolescent Research, 11(1), 116–129. doi:10.1177/074355489 Perry, C. L., Williams, C. L., Komro, K. A., Veblen-Mortenson, S.,
6111007. Forster, J. L., Bernstein-Lachter, R., et al. (2000). Project
Maguin, E., & Loeber, R. (1996). Academic performance and Northland high school interventions: Community action to
delinquency. Crime and Justice, 20, 145–264. reduce adolescent alcohol use. Health Education & Behavior,
*Maxfield, M., Castner, L., Maralani, V., & Vencill, M. (2003). The 27(1), 29–49. doi:10.1177/109019810002700105.
Quantum Opportunity Program Demonstration: Implementation Perry, C. L., Williams, C. L., Komro, K. A., Veblen-Mortenson, S.,
findings. Washington, DC: Mathematica Policy Research Inc. Stigler, M. H., Munson, K. A., et al. (2002). Project Northland:
Maynard, R. A. (1996). Kids having kids: A special report on the cost Long-term outcomes of community action to reduce adolescent
of adolescent childbearing. New York: Robin Wood Foundation. alcohol use. Health Education Research, 17(1), 117–132. doi:10.
McVie, S., & Bradshaw, P. (2005). Adolescent smoking, drinking and 1093/her/17.1.117.
drug use. Centre for Law and Society: University of Edinburgh. *Perry, C. L., Williams, C. L., Veblen-Mortenson, S., Toomey, T. L.,
Mirza, K. A. H., & Mirza, S. (2008). Adolescent substance misuse. Komro, K. A., Anstine, P. S., et al. (1996). Project Northland:
Psychiatry, 7(8), 357–362. doi:10.1016/j.mppsy.2008.05.011. Outcomes of a communitywide alcohol use prevention program
Mistry, R., McCarthy, W. J., Yancey, A. K., Lu, Y., & Patel, M. during early adolescence. American Journal of Public Health,
(2009). Resilience and patterns of health risk behaviors in 86(7), 956–965. doi:10.2105/AJPH.86.7.956.
California adolescents. Preventive Medicine, 48(3), 291–297. Pettit, G. S., Laird, R. D., Bates, J. E., & Dodge, K. A. (1997).
doi:10.1016/j.ypmed.2008.12.013. Patterns of after-school care in middle childhood: Risk factors
Moher, D., Liberati, A., Tetzlaff, J., & Altman, D. G. (2009). and developmental outcomes. Merrill-Palmer Quarterly, 1982,
Preferred reporting items for systematic reviews and meta- 515–538.
analyses: The PRISMA statement. Annals of Internal Medicine, *Philliber, S., Kaye, J., & Herrling, S. (2001). The national
151(4), 264–269. doi:10.7326/0003-4819-151-4-200908180- evaluation of the Children’s Aid Society Carrera-Model
00135. Program to prevent teen pregnancy. Accord, NY: Philliber
*Monahan, D., Greene, V. L., Ditmar, M., & Roloson, T. (2011). Research Associates.
Effectiveness of an abstinence-only intervention sited in a Philliber, S., Kaye, J. W., Herrling, S., & West, E. (2002). Preventing
neighborhood community center. Journal of Children and pregnancy and improving health care access among teenagers:
Poverty,. doi:10.1080/10796126.2011.532481. An evaluation of the Children’s Aid Society-Carrera Program.
Morton, M., & Montgomery, P. (2011). Youth empowerment Perspectives on Sexual and Reproductive Health, 34(5),
programs for improving self-efficacy and self-esteem of adoles- 244–251. doi:10.2307/3097823.
cents. Campbell Systematic Reviews, 7, 5. Qiao, C., & McNaught, H. (2007). Evaluation of project K. Retrieved
National Youth Agency (2007). The NYA Guide to Youth Work in from Ministry of Social Development website http://thehub.
England. Leicester: National Youth Agency. Retrieved from superu.govt.nz/sites/default/files/40181_project_k_0.pdf.
http://www.nya.org.uk/wp-content/uploads/2014/06/The_NYA_ Resnick, M. D., Bearman, P. S., Blum, R. W., Bauman, K. E., Harris,
Guide_to_Youth_Work_and_Youth_Services.pdf. K. M., Jones, J., et al. (1997). Protecting adolescents from harm:
O’Donnell, L., San Doval, A., Duran, R., Haber, D., Atnafou, R., Findings from the National Longitudinal Study on Adolescent
Piessens, P., et al. (2003). Reach for health: A school sponsored Health. JAMA, 278(10), 823–832. doi:10.1001/jama.1997.
community youth service intervention for middle school students. 03550100049038.
Los Altos, CA: Sociometrics. *Reyna, V. F., & Mills, B. A. (2014). Theoretically motivated
*O’Donnell, L., Stueve, A., O’Donnell, C., Duran, R., San Doval, A., interventions for reducing sexual risk taking in adolescence: A
Wilson, R., et al. (2002). Long-term reductions in sexual randomized controlled experiment applying fuzzy-trace theory.
123
Journal of Experimental Psychology: General, 143(4), adolescent alcohol prevention: What worked in Project North-
1627–1648. doi:10.1037/a0036717. land? Prevention Science, 7(3), 269–280.
Rhodes, J. E., Reddy, R., & Grossman, J. B. (2005). The protective Tierney, J.P., Grossman, J., & Resch, N. (1995). Making a difference:
influence of mentoring on adolescents’ substance use: Direct and An impact study of big brother big sisters. Philadephia, CA,
indirect pathways. Applied Developmental Science, 9(1), 31–47. Public/Private Ventures. Retrieved from http://www.ppv.org/
doi:10.1207/s1532480xads0901_4. ppv/publications/assets/111_publication.pdf.
Rhodes, J. E., Reddy, R., Grossman, J. B., & Maxine Lee, J. (2002). Tolan, P. (2016). Positive youth development interventions: Advanc-
Volunteer mentoring relationships with minority youth: An ing evaluation theory and practice. Applied Developmental
analysis of same-versus cross-race matches. Journal of Applied Science, 20(3), 147–149. doi:10.1080/10888691.2015.1014485.
Social Psychology, 32(10), 2114–2133. doi:10.1111/j.1559- Tolan, P. H., Guerra, N. G., & Kendall, P. C. (1995). A
1816.2002.tb02066.x. developmental-ecological perspective on anti-social behaviour
Rodrı́guez-Planas, N. (2010). Longer-term impacts of mentoring, in children and adolescents: Toward a unified risk and interven-
educational services, and incentives to learn: Evidence from a tion framework. Journal of Consulting and Clinical Psychology,
randomized trial in the United States. Bonn, Institute for the Stud 63(4), 579–584. doi:10.1037/0022-006X.63.4.579.
of Labor. Retrieved from ftp.iza.org/dp4754.pdf. Trenholm, C., Devaney, B., Fortson, K., Clark, M., Quay, L., &
Rosenthal, R. (1979). The file drawer problem and tolerance for null Wheeler, J. (2008). Impacts of abstinence education on teen
results. Psychological Bulletin, 86(3), 638–641. doi:10.1037/ sexual activity, risk of pregnancy, and risk of sexually transmit-
0033-2909.86.3.638. ted diseases. Journal of Policy Analysis and Management, 27(2),
Roth, J. L., & Brooks-Gunn, J. (2003a). What exactly is a youth 255–276. doi:10.1002/pam.20324.
development program? Answers from research and practice. *Trenholm, C., Devaney, B., Fortson, K., Quay, K., Wheeler, J., &
Applied Developmental Science, 7, 94–111. Clark, M. (2007). Impacts of Four Title V, Section 510
Roth, J. L., & Brooks-Gunn, J. (2003b). Youth development Abstinence Education Programs. Final Report. Department of
programs: Risk, prevention and policy. Journal of Adolescent Health and Human Services, US, Mathematica Policy Research,
Health, 32(3), 170–182. Inc.
Roth, J. L., & Brooks-Gunn, J. (2016). Evaluating youth development Walker G., & Vilella-Velez, F (1992). Testing the model. In: Anatomy
programs: Progress and promise. Applied Developmental of a demonstration: The Summer Training and Education
Science, 20(3), 188–202. doi:10.1080/10888691.2015.1113879. Program (STEP) from pilot through replication and post
Roth, J., Brooks-Gunn, J., Murray, L., & Foster, W. (1998). program impacts. Philadelphia, PA: Public/Private Ventures.
Promoting healthy adolescents: Synthesis of youth development Walker G., & Vilella-Velez, F. (1992). Long-term impacts. In:
program evaluations. Journal of Research on Adolescence, 8(4), Anatomy of a demonstration: The Summer Training and
423–459. doi:10.1207/s15327795jra0804_2. Education Program (STEP) from pilot through replication and
Scales, P. C., Benson, P. L., Oesterle, S., Hill, K. G., Hawkins, J. D., post program impacts. Philadelphia, PA: Public/Private
& Pashak, T. J. (2016). The dimensions of successful young Ventures.
adult development: A conceptual and measurement framework. Wellings, K., Nanchahal, K., Macdowall, W., McManus, S., Erens,
Applied Developmental Science, 20(3), 150–174. B., Mercer, C. H., et al. (2001). Sexual behaviour in Britain:
Schirm, A., & Rodriguez-Planas, N. (2004). The Quantum Opportu- Early heterosexual experience. The Lancet, 358(9296),
nity Program demonstration: Initial post-intervention impacts. 1843–1850. doi:10.1016/S0140-6736(01)06885-4.
Washington, DC: Mathematica Policy Research Inc. *West, B., Abatemarco, D., Ohman-Strickland, P. A., Zec, V., Russo,
Schirm, A., Rodriguez-Planas, N., Maxfield, M., & Tuttle, C. (2003). A., & Milic, R. (2008). Project Northland in Croatia: Results and
The Quantum Opportunity Program demonstration: Short-term lessons learned. Journal of Drug Education, 38(1), 55–70.
impacts. Washington, DC: Mathematica Policy Research Inc. doi:10.2190/DE.38.1.e.
Scott, S., Knapp, M., Henderson, J., & Maughan, B. (2001). Financial Wheeler, S. B. (2010). Effects of self-esteem and academic perfor-
cost of social exclusion: Follow up study of antisocial children mance on adolescent decision-making: An examination of early
into adulthood. BMJ, 323(7306), 191. doi:10.1136/bmj.323. sexual intercourse and illegal substance use. Journal of Adoles-
7306.191. cent Health, 47(6), 582–590. doi:10.1016/j.jadohealth.2010.04.
Scottish Government. (2009). Valuing Young people: Principles and 009.
connections to support young people achieve their potential. Wiefferink, C. H., Peters, L., Hoekstra, F., Ten Dam, G., Buijs, G. J.,
Edinburgh: Scottish Government. & Paulussen, T. G. (2006). Clustering of health-related behaviors
Shepherd, J., Kavanagh, J., Picot, J., Cooper, K., Harden, A., Barnett- and their determinants: Possible consequences for school health
Page, E., et al. (2010). The effectiveness and cost-effectiveness interventions. Prevention Science, 7(2), 127–149.
of behavioural interventions for the prevention of sexually World Health Organisation. (2014). Global status report on health
transmitted infections in young people aged 13–19: A systematic and alcohol. Geneva: Switzerland, World Health Organization.
review and economic evaluation. Health Technology Assess- Zief, S. G., Lauver, S., & Maynard, R. A. (2006). Impacts of after-
ment, 14(7), 1–230. doi:10.3310/hta14070. school programs on student outcomes: A systematic review.
Spencer, M. B., & Spencer, T. R. (2014). Exploring the promises, Campbell Systematic Reviews,. doi:10.4073/csr.2006.3.
intricacies and challenges to positive youth development.
Journal of Youth and Adolescence, 43, 1027–1035.
Steinberg, L. (2004). Risk taking in adolescence: What changes, and Oana Ciocanel is a Health Psychologist employed by NHS Tayside.
why? Annals of the New York Academy of Sciences, 1021(1), She received her Masters degree in Health Psychology from the
51–58. doi:10.1196/annals.1308.005. University of Stirling. Her major research interests are sexual health
Sterne, J. A., & Egger, M. (2001). Funnel plots for detecting bias in and behavior change.
meta-analysis: Guidelines on choice of axis. Journal of Clinical
Epidemiology, 54(10), 1046–1055. doi:10.1016/S0895-4356(01) Kevin Power is Area Head of NHS Tayside Psychological Therapies
00377-8. Service and an Honorary Professor of Psychology at the University of
Stigler, M. H., Perry, C. L., Komro, K. A., Cudeck, R., & Williams, Stirling. He received his MA (Hons) degree from University of
C. L. (2006). Teasing apart a multiple component approach to Edinburgh, Masters in Applied Science from University of Glasgow
123
and Ph.D. in Psychology from University of Stirling. His major research Kirsty Gillings is a Consultant Clinical Psychologist employed by
interests include eating disorders, anxiety disorders, including posttrau- NHS Fife. She received her Ph.D. in Psychology from University of
matic stress disorder, suicidal behaviour and service improvement. Warwick and her doctorate in Clinical Psychology from the
University of Birmingham. Her major research interests include the
Ann Eriksen is the Executive Lead for Sexual Health and Blood psychological correlates of addictive behaviors, risk behavior and
Borne Virus at NHS Tayside. She received her MA (Hons) in History trauma.
from the University of Edinburgh. Her major research interests are
teenage pregnancy, sexual health and young people and viral
hepatitis.
123

Ciocanel2017 Article EffectivenessOfPositiveYouthDe

Uploaded by

Copyright:

Available Formats

Ciocanel2017 Article EffectivenessOfPositiveYouthDe

Uploaded by

Document Information

Original Title

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

Ciocanel2017 Article EffectivenessOfPositiveYouthDe

Uploaded by

Copyright:

Available Formats

J Youth Adolescence (2017) 46:483–504

Effectiveness of Positive Youth Development Interventions:

The PRISMA guidelines for the conduct of systematic Setting

Positive youth development programs were defined as Design

Table 1 Outcome categories used in this meta-analysis

Records identified through database Additional records identified

Records after duplicates removed and screened by title

• Not age 10-19: (n=56)

Fig. 1 CONSORT diagram

Risk Bias Assessment For continuous outcomes, we calculated the standardized

five being delivered in a combined school, community and

number of sessions. The mean intervention duration was

80 weeks, with studies ranging from 3 to 240 weeks.

Seventeen interventions involved at least 20 sessions and

follow-up ranged from immediately after the intervention

to 5 years after the intervention. The first follow-up was

conducted immediately or within 3 months of the inter-

up after 5–18 months, seven studies involved a second

follow-up after 6–20 months and five studies reported a

third follow-up from 12 to 32 months.

behavioral including problem behaviors (66 %), academic

improvement and school adjustment (45 %) and sexual

risk behaviors/pregnancy (45 %). A smaller number of

studies examined the effectiveness of these interventions

on positive social behaviors (29 %) and psychological

ing seventeen combined individual and group interven-

tions. Twenty-one programs were multi-modal and

involved primary and secondary interventions. Three

single-modal programs provided mentoring, skills train-

programs, primary after-school activities covered aca-

demic and homework help (n = 8), mentoring (n = 7),

emotional skill development (n = 16), recreational

ing, social competence, cognitive competence, emotional

competence, self-efficacy, self-determination and a belief

The bias risk in the studies is summarized in Fig. 2.

to judge the randomization procedure quality, with 16

studies having an unclear rating. Only seven studies

described the randomization sequence and only two

Fig. 2 Risk of bias ratings

Table 3 Effect sizes by outcome category

indicated how the allocation concealment was conducted. Intervention Effects

-1.00 -0.50 0.00 0.50 1.00

Favours Control Favours intervention

Fig. 3 Effect sizes for problem behaviors

Behavioral Adjustments Positive Social Behavior

-1.00 -0.50 0.00 0.50 1.00

Favours Control Favours intervention

Fig. 4 Effect sizes for positive social behaviors

-1.00 -0.50 0.00 0.50 1.00

Favours Control Favours intervention

Fig. 5 Effect sizes for psychological adjustment

-1.00 -0.50 0.00 0.50 1.00

Favours Control Favours intervention

Fig. 6 Effects sizes for academic achievement

Hedges's Lower Upper

-1.00 -0.50 0.00 0.50 1.00

Favours Control Favours intervention

Fig. 7 Effects sizes for sexual risk behaviors

Table 4 Moderators of effect sizes