The Effectiveness of the Prevention Program 'New Perspectives'
(NP) on Juvenile Delinquency: The Investigation of Mediational
Processes
Master's Thesis Forensic Child and Youth Care Sciences Graduate School of Child Development and Education University of Amsterdam C. J. M. Wibbelink 10259856 S. L. A. de Vries M. Hoeve Amsterdam, September 2014
Table of Contents
Abstract ... 3
Introduction ... 4
New Perspectives ... 6
Mediator: Motivation ... 6
Mediator: Cognitive Distortions ... 7
Mediator: The Social Network ... 9
Present Study ... 10 Methods ... 11 Participants ... 11 Procedure ... 11 Measures ... 13 Analytic Strategy ... 16 Results ... 18 Descriptive Statistics ... 18
Investigation of Change during the Intervention and Intervention Effects ... 19
Assessment of Mediation ... 23
Discussion ... 24
References ... 29
Appendix ... 36
Abstract
Objective: The present study examined the effectiveness of New Perspectives (NP) compared with treatment as usual (TAU) on juveniles' delinquent behavior. There was also investigated whether the hypothesized mediators were related to the observed effects in juveniles'
delinquent behavior. Method: Participants in this randomized controlled trial were 101 juveniles who received NP (N = 47) or treatment as usual (N = 54). Data were collected at three time points: at baseline (preintervention), after 3 months, and after 6 months
(postintervention). Latent growth modeling was used to investigate the change during the intervention, program effectiveness and mediational processes. Results: NP was more effective compared with TAU in reducing the number of committed offenses and the affiliation with deviant peers. There were no effects found for the other hypothesized mediators. Furthermore, there was no evidence for mediation. Conclusion: The results revealed that NP was only more effective in reducing the number of committed offenses and the affiliation with deviant peers. Future studies on the effectiveness of NP are needed. Keywords: New Perspectives, mediators of intervention effects, effectiveness, delinquency, prevention, randomized controlled trial.
The Effectiveness of the Prevention Program 'New Perspectives' (NP) on Juvenile Delinquency: The Investigation of Mediational Processes
Introduction
Juvenile delinquency is a serious problem in our society. For example, in the Dutch community approximately one third of the juveniles between 12 to 17 years admitted that he or she has ever committed a crime (Van der Laan & Blom, 2011). Wartna and colleagues (2011) concluded that in 2008 about 36% of the Dutch juveniles, who have ever committed a crime, reoffends.
The theory of Moffitt (1993) suggested that there are two different groups of antisocial juveniles: juveniles with adolescent-limited antisocial behavior and juveniles with life-course-persistent antisocial behavior. The first group, adolescent-limited antisocial juveniles, begins with antisocial behavior during adolescence. However, their antisocial behavior declines in young adulthood. The second group, life-course-persistent antisocial juveniles, begins with antisocial behavior at an early age, continues throughout the life course with this behavior, and shows a pattern of more serious delinquent behaviors with age. Moffitt (1993) suggested that the second group is probably responsible for the majority of crimes committed by young, and later, adult offenders. Intervening at an early stage of delinquency, in which juveniles are still at the beginning of their antisocial 'career', is therefore important. For this reason, there are a number of interventions aiming to prevent juveniles from starting with antisocial behavior or more serious delinquent behaviors.
There are several studies and reviews that focus on the effectiveness of the
interventions that seek to prevent delinquency (e.g., Petrosino, Turpin-Petrosino, & Buehler, 2003; Piquero, Farrington, Welsh, Tremblay, & Jennings, 2009; Mulvey, Arthur, & Reppucci, 1993; Lӧsel & Beelmann, 2003; Deković et al., 2011; De Vries, Hoeve, Assink, Stams, & Asscher, in press). However, there is still no consensus on the effectiveness of prevention programs. Mulvey and colleagues (1993) discussed program outcomes in terms of primary and secondary prevention. Primary prevention programs can be described as programs that attempt to identify individuals at risk for delinquent behavior. Secondary prevention programs refer to programs aimed at restoring juveniles who had shown some indications of behavior problems, but had not been classified as delinquent. Mulvey and colleagues (1993) concluded that some programs with well-implemented and clearly defined interventions, that included behavioral and family-based change strategies, were found to produce reductions in official
arrest rates. However, there were no reductions found in self-reported delinquency. De Vries and colleagues (in press) studied the effectiveness of prevention programs in preventing juvenile delinquency and examined which program, sample and study characteristics
contributed to the effects. They concluded that prevention programs have a positive effect on the prevention of persistent juvenile delinquency. However, it is important that programs address multiple risk factors of (persistent) delinquency (i.e., multimodal programs) and include program elements, such as behavioral contracting, modeling, and parenting skills training. Programs should also adjust their intensity on the juvenile's risk for reoffending (De Vries et al., in press). In contrast to the findings of De Vries et al. (in press), Deković and colleagues (2011) concluded in their meta-analytic study that prevention programs during early and middle childhood show no convincing evidence in reducing delinquent behavior in adulthood. To conclude, there is still no consensus on the effectiveness of prevention
programs.
Deković and colleagues (2011) pointed out several important unresolved issues. One of these issues was that it is still unknown which program elements are the most important for the prevention and reduction of delinquency. There is a lot of research about the effectiveness of interventions, but research about the mechanisms of change is lacking (Kazdin, 2007). Only a few studies have examined hypothesized mediational processes through which interventions affect juveniles’ problem behavior (DeGarmo, Eddy, Redi, & Fetrow, 2009). This is remarkable, given that information about mediational processes is very valuable for the development and improvement of effective interventions (Cheong, MacKinnon, & Khoo, 2003). Therefore, research on mediational processes is needed.
The present study will be an addition to the existing literature on prevention programs. First, the effectiveness of the Dutch prevention program 'New Perspectives' (Elling &
Melissen, 2007) will be examined. This secondary prevention program is community-based and targets juveniles at risk of the development of (persistent) delinquent behavior. The program has a multimodal format, which has been proven to be effective for the prevention of (persistent) delinquent behavior (De Vries et al., in press). NP is one of the 12 identified secondary prevention programs in the Netherlands for children under 12 years (Van der Put, Assink, Bindels, Stams, & De Vries, 2013). According to Van der Put et al. (2013), there is no research available on the effectiveness of these secondary prevention programs. A number of interventions have been proven effective in other countries, but information about the
effectiveness in the Netherlands is unknown. Therefore, research on the effectiveness of Dutch secondary prevention programs is necessary.
Second, there will be investigated whether the hypothesized mediators are related to the observed effects in juveniles' delinquent behavior. The mediators of interest are:
motivation, cognitive distortions, and the social network of juveniles. New Perspectives
The prevention program 'New Perspectives' (NP) is a short-term, intensive, and community-based program aiming to prevent juveniles to become (more) delinquent. The program consists of an intensive phase of three months and an aftercare phase of three months. The NP program has an outreaching (i.e., the juveniles are visited in their own environment) and multisystematic approach. The program focuses on multiple factors related to delinquency and recidivism, for example, motivation, cognitive distortions, moral
reasoning, and the social network. NP is based on the Transtheoretical Model of Behavior Change of Prochaska & DiClemente (1984). According to this model, behavior change involves a progression through an invariant sequence of stages (Casey, Day, & Howells, 2005). NP is also based on the principles of the Risk-Need-Responsivity model (RNR model) (Andrews, Bonta, & Hoge, 1990). The risk principle states that the intensity of the
intervention should be matched to the offender's risk of recidivism. The need principle states that the intervention should target criminogenic needs, which are aspects of the offender's social, emotional and psychological functioning that increase the likelihood of delinquency. The responsivity principle states that the intervention must be adjusted to the personality, learning style, cognitive abilities, motivation, and cultural background of the offender (Andrews et al., 1990; Andrews & Bonta, 2010). Andrews and Bonta (2010) concluded that interventions based on the RNR model are effective in reducing recidivism.
Prior evaluation studies of NP have shown improvements in different areas, like family, school and peers (Noorda & Veenbaas, 1997; Geldorp, Groen, Hilhorst, Bürmann, & Rietveld, 2004; Buysse, Van den Andel, & Van Dijk, 2008; Nauta, 2008). However, the use of a control group in these studies is lacking and, consequently, the positive results cannot be attributed to the intervention. Using a randomized controlled trial is the best way to evaluate treatment effects (Henggeler & Clingempeel, 2002). Therefore, the current study will make use of a randomized controlled trial.
Mediator: Motivation
As already mentioned above, NP addresses multiple risk factors. One of these risk factors is the motivation of the juveniles. By increasing the motivation of the juvenile, NP attempts to decrease the risk of (re)offending. Thus, increasing the motivation can be seen as
one of the mediational processes of the intervention.
NP tries to increase the motivation of the juveniles by using motivational interviewing. Motivational interviewing (Miller & Rollnick, 2002) is an approach in which the therapist evokes the person's intrinsic motivation and his resources for change (Miller & Rollnick, 2002). Four broad principles underlie motivational interviewing: express empathy towards the client, develop discrepancy between the client's present behavior and his broader goals and values, de-escalate resistance of the client, and support the client's self-efficacy (i.e., the person's belief in his ability to succeed) (Miller & Rollnick, 2002). The Transtheoretical Model of Behavior Change (TTM) of Prochaska & DiClemente (1984) has played an important role in the development of motivational interviewing (DiClemente & Velasquez, 2002). The TTM views behavior change as a progression through gradual steps, where each step involves multiple tasks and require different activities. Motivational interviewing provides the counselor the techniques required for each stage.
Lundahl and Burke (2009) reviewed the research support for motivational interviewing and concluded that motivational interviewing was more effective than no treatment and equally effective compared to other treatments for a variety of problems (e.g., substance use, risky behaviors, emotional well-being). In a systematic review of McMurran (2009) on motivational interviewing with offenders, there was concluded that motivational interviewing can lead to enhanced motivation to change, reduced offending and improved retention in treatment. Improved retention in treatment is important, as McMurran and Theodosi (2007) suggested that failure to complete treatment can enhance the risk of recidivism, especially for high-risk offenders, through, for example, emotional distress and increased antiauthority attitudes.
Therefore, by using motivational interviewing as part of the intervention, it is expected that the juveniles are more motivated, their retention in treatment is improved, and,
consequently, the juveniles are less at risk of developing (persistent) delinquent behavior. Mediator: Cognitive Distortions
Another risk factor of (re)offending that NP addresses, is cognitive distortions. Cognitive distortions refer to inaccurate thoughts, beliefs or attitudes concerning the person's own or others' behavior (Liau, Barriga, & Gibbs, 1998). NP attempts to modify cognitive distortions by using a technique based on Ellis' (1962) ABC model of human disturbance. The 'A' represents the activating event, 'B' represents person's belief, and 'C' stands for the
emotional and behavioral consequences. The ABC model could be seen as the basis of all
cognitive therapies (Muran, 1991). The model aims to give clients insight into their irrational beliefs, or cognitive distortions, and their dysfunctional emotional and behavioral
consequences (Ellis & Dryden, 1997).
There is a distinction made between two types of cognitive distortions: self-debasing cognitive distortions and self-serving cognitive distortions. Self-debasing cognitive distortions are related to internalizing behaviors. These cognitive distortions ensure that the person feels humiliated which may lead to self-harm. Self-serving cognitive distortions are related to externalizing behaviors and help to protect the person against blame or a negative self-concept (Barriga, Landau, Stinson, Liau, & Gibbs, 2000). Self-serving cognitive distortions are more important for the development of delinquent behavior, therefore, more attention is paid to these cognitive distortions.
Barriga & Gibbs (1996) discussed a typology of four categories of self-serving cognitive distortions: self-centered, blaming others, minimizing/mislabeling, and assuming the worst. Self-centered refers to the tendency of a person to only take into account his own needs, rights, (immediate) feelings, and desires; the rights, feelings and needs of others are scarcely considered. Blaming others refers to misattributing blame for one's antisocial acts to outside sources (e.g., another person, group, bad mood). Minimizing/mislabeling refers to describing one's antisocial behavior as necessary, acceptable or even admirable. At last, assuming the worst refers to the tendency of expecting the worst-case scenario and assuming that improvement in one's own and others' behavior is impossible (Barriga & Gibbs, 1996). The four categories are divided in primary and secondary self-serving cognitive distortions. Primary serving cognitive distortions are centered beliefs and attitudes (i.e., self-centered). Secondary cognitive distortions refer to rationalizations that contribute to the neutralizing of feelings of guilt and thereby prevent damage to the person's self-image after an antisocial act (i.e., blaming others, minimizing/mislabeling, and assuming the worst) (Barriga & Gibbs, 1996).
Previous research revealed a positive relationship between self-serving cognitive distortions and antisocial and delinquent behavior (e.g., Lardén, Melin, Holst, & Långstrӧm, 2005; Barriga et al., 2000; Barriga, Hawkins, & Camelia, 2008; Liau et al., 1998). Lardén and colleagues (2005) concluded that delinquent juveniles had more antisocial cognitive
distortions compared to non-delinquent juveniles. According to Liau et al. (1998), highly significant relationships were found between self-serving cognitive distortions and antisocial behavior for non-delinquent and delinquent juveniles. Therefore, by addressing cognitive
distortions of the juveniles, NP aims to prevent juveniles to develop (persistent) delinquent behavior.
Mediator: The Social Network
The last hypothesized mechanism of change is related to the social network of the juveniles. NP is a community-based and outreaching intervention. It is designed to intervene in the social network of the juveniles, if necessary. The intervention focuses on the
relationships with parents and peers.
Poor quality of parenting behavior is an important risk factor for delinquent behavior (Hoeve et al., 2009). NP recognizes the importance of parenting behavior by using various techniques for parents, such as psychoeducation, empowerment, setting house rules and improving communication skills. NP focuses on five behaviors: parental involvement, positive reinforcement, problem solving, discipline, and monitoring (Tan, Brussen, Sewray, Rijnveld, & Bontes, 2010). These parenting behaviors can be summarized into two key dimensions, support and control, that have often been used in the literature to describe parenting behavior (Hoeve et al., 2009; Deković, Wissink, & Meijer, 2004). Supportive parenting behavior refers to behaviors towards the child that ensures that the child feels accepted, approved and comfortable (Rollins & Thomas, 1979). Positive aspects of support include acceptance, affection, warmth, responsiveness, and sensitivity. If parents give very little support, this can be described as hostile, neglect or rejection (Rohner, 2004). Juveniles who receive little support from their parents are more at risk of exhibiting delinquent behavior (Hoeve et al., 2009).
Control can be divided into two categories, authoritarian or restrictive control and authoritative control. Authoritarian control (i.e., control based on restrictive and firm discipline techniques) is often associated with negative developmental outcomes, like delinquent and antisocial behavior (Hoeve et al., 2009). Authoritative control (i.e., control based on child-oriented and inductive discipline techniques) is associated with positive developmental outcomes (Deković et al., 2004). Monitoring is another aspect of control (Hoeve et al., 2009; Deković et al., 2004). Monitoring can be defined as "parents' knowledge of their children's whereabouts, companions, and activities" (Pettit, Bates, Dodge, & Meece, 1999, p. 768). Research has shown that there is a relationship between parental monitoring and the child’s delinquent behavior, where high levels of parental monitoring are associated with less delinquent behavior (e.g., Hoeve et al., 2009; Patterson & Stouthamer-Loeber, 1984).
Risk factors for delinquency in the social network are not limited to the parents, but also include relationships with peers. Therefore, NP investigates the relationships of the juveniles with prosocial and deviant peers. The intervention teaches the juveniles skills to resist negative influences of deviant peers and tries to improve the leisure activities of the juveniles (Tan et al., 2010). This is important, because numerous studies have shown that relationships with delinquent peers promote the development and maintenance of delinquent behavior (e.g., Keijsers et al., 2012; Deković et al., 2004; Bowman, Prelow, & Weaver, 2007). Kaplan, Johnson and Bailey (1987) gave several explanations for the link between deviant peers and juveniles' delinquent behavior. First, exposure to deviant peers facilitates antisocial behavior if the juvenile is motivated to behave in such a way. The deviant peers teach the juvenile how to engage in antisocial behavior and they provide the means to engage in such behavior. Second, deviant peers provide the juvenile with rewards for engaging in antisocial behavior (i.e., positive reinforcement). Finally, exposure to deviant peers decreases the effects of social and personal control mechanisms that might prevent the juvenile to engage in
antisocial behavior. According to Dishion, Andrews, & Crosby (1995), deviant peers supply the juvenile with the motivation, attitudes, and rationalization to support antisocial behavior and they provide opportunities to engage in such behavior.
Overall, it can be concluded that it is important to intervene in the social network of the juveniles to reduce the risk of developing (persistent) delinquent behavior.
Present Study
In sum, the aim of the present study is to examine the effects of NP compared with treatment as usual (TAU) on juveniles' delinquent behavior. The study will also focus on the hypothesized mediators through which the intervention exerts its effect on delinquency (i.e., mediational processes). The hypothesized mediators are motivation, cognitive distortions, and the social network.
First, the change that took place during the intervention in juveniles' delinquent
behavior, motivation, cognitive distortions, and the social network will be examined. This will be investigated by conducting several latent growth models. The social network is divided in two important components: parenting behavior (i.e., support, authoritative and authoritarian control, and monitoring) and the relationships with peers (i.e., affiliation with deviant and prosocial peers, and the intensity of contact with peers). Research has shown that the delinquent behaviors of peers and the time spending with peers are both of influence on the link between deviant peers and juveniles' delinquent behavior (e.g., Loeber, Farrington,
Stouthamer-Loeber, Moffitt, & Caspi, 1998; Patterson & Dishion, 1985). Therefore, a distinction is made between the affiliation with peers and the intensity of contact with peers.
Second, the effectiveness of NP compared with TAU will be examined using multiple-group analyses. It is hypothesized, based on previous evaluation studies of NP (Noorda & Veenbaas, 1997; Geldorp et al., 2004; Buysse et al., 2008; Nauta, 2008), that NP has more effect on juveniles' delinquency and the hypothesized mediators, compared with TAU.
At last, it is examined whether the hypothesized mediators actually lead to a change in the delinquent behavior of juveniles. The hypothesis is that the effects of NP on the
hypothesized mediators will lead to reductions in juveniles' delinquent behavior. So, for example, it is expected that by enhancing the motivation of the juveniles, the juveniles will show less delinquent behavior.
Methods Participants
The sample consisted of 101 juveniles (68 boys and 33 girls) with an average age of M = 15.58 year (SD = 1.53, range = 12.30 - 19.30). Of the 101 juveniles, 47 received NP and 54 received TAU. Seventeen percent of the juveniles had a Dutch background. Of the juveniles with another ethnic background, the largest group had a Surinamese (30%) or a Moroccan (27%) background. More than half of the juveniles (55%) lived in a single-parent home and almost half of the juveniles (45%) followed lower secondary professional education (VMBO).
To examine differences between the treatment conditions at pre-test (T1), independent samples t tests for continuous variables and chi-square analyses for categorical variables were conducted. No significant differences (p > .05) were found on demographic variables (gender, age, educational level and ethnicity), hypothesized mediators, and the outcome measure. This suggests that the randomization was successful.
Procedure
Juveniles who were at risk of developing (persistent) delinquent behavior, were referred from multiple sources (e.g., self-referred, treatment professional, court) to a youth care institution in an urban area of Amsterdam, the Netherlands. Juveniles were eligible for participation if they met the following three inclusion criteria: age 12 to 23 years, problems in multiple life domains (e.g., school, family, peers), and at risk of developing (persistent) delinquent behavior. Exclusion criteria were: IQ below 70, substance dependency, psychiatric problems, no residence status in the Netherlands, and/or absence of motivation to stop
committing any delinquent acts. If the juveniles were eligible for participation, they were randomized to NP or TAU by the research staff, using computer generated block
randomization. The ratio of the randomization between NP and TAU was 1:1. Juveniles randomized to TAU were referred to a variety of other standard services of youth care in Amsterdam. Immediately after randomization, the juveniles and their parents were contacted (by phone or face to face) to explain the research procedures. If the juvenile agreed with the research, a written informed consent was obtained and the first assessment was conducted.
The assessments took place before the start of the program (T1), after 3 months (after the intensive intervention phase of NP, T2), after 6 months (after the aftercare phase of NP, T3) and 18 months after registration (T4). The data of T4 were not used in this study. Each juvenile received €20 and each parent received €10 for completing each assessment. The design, procedures and the informed consent of the study were approved by the Ethics Committee of the University of Amsterdam (Faculty of Social and Behavioral Science).
Despite the efforts made, some juveniles have not completed all three assessments (see Figure 1). Eleven juveniles have not completed the second assessment (T2) and 6 juveniles have not completed the third assessment (T3). To assess if the participants who have not completed the second and/or third assessment differed from the participants who have completed all three assessments, Little's MCAR test for missing completely at random was conducted. Little's MCAR test indicated that the data were missing completely at random (χ2(206) = 231.51, p = .107). Therefore, all participants were included in the analyses and expectation maximization was used to estimate the missing values on the mediators (Tabachnick & Fidell, 2013). Missing values on the semicontinuous outcome measure delinquency were not estimated.
Figure 1. Flow diagram. NP = New Perspectives; TAU = Treatment as usual.
Measures
Outcome measure
Delinquent behavior. Delinquency was measured using the 'Self-report Delinquency Scale' (SRD) (Van der Laan & Blom, 2006). This questionnaire consists of 17 self-reported offences, varying from minor offences to more serious offences. The 17 offences can be divided in three types of delinquent behavior: vandalism, property crime and violent crime. The participants were asked to indicate on a list of 17 offences, if they have ever been involved in each of these offences. Examples of items are: 'Have you ever deliberately
destroyed something in a bus, tram, metro or train?' and 'Have you ever used violence to steal something from someone, for example, a purse, wallet or mobile phone?'. Subsequently, if the participant answered with 'yes', they are asked how often they exhibited this delinquent
behavior in the previous three months. In the present study, a total score was calculated based on how often the participant indicated that he had exhibited an offence in the last three
months. Cronbach's alpha varied from .62 to .89 in the present study. Mediators
Motivation. Motivation is assessed by using a short Dutch version of the 'Motivation for Treatment Questionnaire' (MTQ) (Van Binsbergen, 2003). The questionnaire is based on
the different stages of behavior change, according to the TTM of Prochaska and DiClemente (1984). The questionnaire consists of 12 items. Examples of items are: 'I think about my behavior, but do not talk about it with others' and 'I know which help I need, and I also talk about it with others'. The juveniles had to answer on a 2-point scale, 1 = not true and 2 = true. A total motivation score was used for the analyses. Cronbach's alpha ranged from .39 to .61.
Cognitive distortions. Cognitive distortions were measured by a Dutch validated version of the 'How I Think Questionnaire' (HIT) (Gibbs, Barriga, & Potter, 2001; Brugman et al., 2011). The HIT is composed of 54 items, 39 items represent self-serving cognitive distortions, 8 items are used to screen for anomalous responses, and the last 7 items are positive fillers. The 39 self-serving cognitive distortions items refer to the four-category typology of self-serving cognitive distortions: self-centered, blaming others,
minimizing/mislabeling, and assuming the worst. All items were assessed using a 6-point Likert scale (1 = disagree strongly to 6 = agree strongly). Examples of items are: 'I make mistakes because I am with the wrong people' and 'You have to hurt people, before they hurt you'. In the present study, a total mean score of the self-serving cognitive distortions items was used. The Dutch version of the HIT showed acceptable reliability and validity (Nas, Brugman, & Koops, 2008; Van der Velden, Brugman, Boorm, & Koops, 2010). In the present study, reliability was good (Cronbach's alpha varied from .91 to .94).
Parenting behavior. In the present study, parenting behavior is divided in four dimensions: support, authoritative control, authoritarian control, and monitoring. The 'Parenting Behavior Questionnaire' (PBQ) (Wissink, Dekovic, & Meijer, 2006) assesses the first three dimensions of parenting behavior.
Support. Support refers to parental warmth and responsiveness and was assessed by 10 items. Examples of items are: 'How often do your parents give you a compliment?' and 'How often do your parents notice, if you're feeling sad or down?'.
Authoritative control. Authoritative control refers to parents' efforts to exercise control by stimulating autonomy and by using democratic means. This dimension consists of 10 items (e.g., 'How often do your parents explain why something is forbidden to you?' and 'How often do your parents ask your opinion?').
Authoritarian control. Authoritarian control refers to parents' behavior to exercise control through strictness and discipline. The scale consists of 10 items. Examples of items are: 'How often are your parents strict?' and 'How often do your parents punish you severely?'.
All items were assessed using a 5-point Likert scale (1 = never to 5 = very often). The mean scores of the three scales were calculated for the present study. Reliability was found to be good for all three scales on the three assessments (Cronbach's alpha varied from .81 to .90).
Monitoring. The last dimension, monitoring, was assessed by using the 'Vragenlijst Toezicht Houden' (VTH), the Dutch version of the parental monitoring scale of Brown, Mounts, Lamborn, and Steinberg (1993). The questionnaire consists of 6 items in which the juveniles were asked how much their parents know about what he or she is doing on a 4-point Likert scale (1 = nothing to 4 = everything). Examples of items are: 'How much do your parents know about how you spent your money?' and 'How much do your parents know about who your friends are?'. A mean score was used for the analyses. Cronbach's alpha ranged from .81 to .83 in the present study.
Relationships with peers. In the present study, the relationships with peers is divided into two dimensions: affiliation with prosocial and deviant peers and the intensity of contact with peers.
Affiliation with peers. The perception of affiliation with prosocial and deviant peers is measured by the Dutch version of the ‘Friends’ scale of the 'Family, Friends & Self Scale' (FFS) questionnaire (Deković et al., 2004; Simpson & McBride, 1992). The questionnaire consists of 17 items. Each item represents a deviant behavior on the 'Deviant peers' scale (10 items) or a prosocial activity on the 'Prosocial peers' scale (7 items). Examples of items are: 'How many of your friends like to play sport?' and 'How many of your friends have quit or want to quit school?'. Juveniles were asked to indicate how many of their friends showed the behavior in question. They could respond on a 5-point Likert scale, 1 = none of my friends to 5 = almost all of my friends. The mean scores of the two scales were used for the analyses. Cronbach's alpha for both scales varied from .71 to 93.
Intensity of contact with peers. The intensity of contact with peers is assessed by a subscale of the 'Basic Peer Questionnaire' (BVL) (Weerman & Smeenk, 2005). The
questionnaire consists of 4 items and the juveniles were asked how often (2 items) and how much time (2 items) they spend with their peers during the week and weekends. The two frequency items were answered through a 3-point scale, ranging from 1 = never/almost never to 3 = 3 or more days or on Saturday and Sunday. An example of an item is: 'How often do you see your friends in the weekend (Saturday and Sunday)?'. The two duration items were also answered through a 3-point scale, ranging from 1 = less than one hour to 3 = all day. An
example of an item is: 'How long have you been together?'. For the present study, a total score was calculated. Reliability ranged from α = .72 to α = .77.
Analytic Strategy
All analyses were performed according to the intention-to-treat principle of Montori and Guyatt (2001). This means that all participants who entered the study at pretest (T1) were included in the analysis, regardless whether the participants had completed the intervention or not.
First, the change during the intervention on the outcome measure and the mediators was assessed by using two-factor latent growth modeling; factor one is the intercept and factor two is the slope (Hess, 2000; Duncan & Duncan, 2004; Muthén & Curran, 1997). Each individual's growth across time can be represented by a growth curve and contained an unique initial measurement point (i.e., intercept) and an unique growth over time (i.e., slope) (Hess, 2000). The intercepts and slopes of the individuals can be summarized in one latent intercept and one latent slope; the two factors. So the collection of the intercepts and slopes of each individual's growth curve results in the latent intercept and slope and gives information about the sample by a mean and a variance. The latent intercept had factor loadings of 1 on the three assessments in the present study, because it was a constant for any individual across time (Duncan & Duncan, 2004). The latent slope had factor loadings of 0, 1 and 2, which
corresponded with the three assessments, respectively T1, T2 and T3. This made it possible to test whether there was a linear growth curve. Demographic characteristics were included in the latent growth models as a covariate if there was a significant association, based on the correlation matrix, between the demographic characteristic and the outcome measure or mediator. Demographic characteristics were: gender, ethnicity (Dutch vs. other ethnic
background), and age (participants younger than 16 years vs. participants older than 16 years). If there was a significant association, the covariate was regressed on the intercept and slope of the latent growth model. The regression on the intercept was held equal in both groups, because of the randomization. The regression on the slope was freely estimated to detect possible moderation effects.
For the semicontinuous outcome measure 'delinquency', a two-part latent growth modeling strategy (Brown, Catalano, Fleming, Haggerty, & Abbot, 2005; Muthén, 2001; Olsen & Schafer, 2001) was used. This strategy decomposed the original distribution of the outcome measure into two parts, the likelihood and the amount of delinquency, whereby each part is modeled simultaneously by separated growth processes. In the first part of the model
(the 'u-part', see Figure 2), non-offending was separated from the rest of the distribution by creating a binary variable that distinguishes any positive level of offending within the three time periods (code '1') from non-offending (code '0'). The offending versus non-offending variable was analyzed as a random-effects logistic growth model (Muthén, 2001; Brown et al., 2005). The slope growth factor of this part of the model indicates whether the likelihood of reporting ever been involved in an offense declines or increases over time.
The second part of the model of (the 'y-part', see Figure 2) studies the frequency of committed offenses among those who had offended. This part consisted of continuous variables that represent the frequency of offenses committed, given that some offending behavior had take place. The continuous variables were analyzed by latent growth modeling and non-offending within each time period was treated as missing data.
To assess the program impact, a multiple-group analysis was conducted. First, the normative development was assessed by a separate analysis of the TAU group. The four growth parameters (intercept mean and variance, and slope mean and variance) found in the model of the TAU group were repeated in the same model of the NP group. Then, it was tested whether the parameters of the NP group were equal to the parameters of the TAU group. A significant difference between a parameter value of the NP group and the TAU group suggests a significant effect of the intervention.
At last, the mediational processes are tested by using the strategy of Cheong and colleagues (2003). The latent growth model for the mediator and the two-part latent growth model for the outcome measure are combined into one parallel process latent growth model (see Figure 2). Group membership was included in the model as a dummy variable ('1' for the NP group and '0' for the TAU group). It was hypothesized that the intervention influences the growth trajectories of the outcome measure ('su' and 'sy', see Figure 2) indirectly by
influencing the growth trajectory of the mediator. Therefore, it was tested whether NP, compared with TAU, significantly changes the trajectory of the mediator, which in turn significantly affects the outcome measure (see Figure 2). To test the mediated effects, the joint significant test of MacKinnon, Lockwood, Hoffman, West, & Sheets (2002) was used. There is evidence for mediation if each of the two paths (b1 and b2u or b1 and b2y in Figure 2)
is significantly nonzero. If there is evidence for mediation, estimates of the mediated effects (the product of the two paths) and estimates of the standard errors are used to construct confidence intervals by using the multivariate delta method (Taylor, MacKinnon, & Tein, 2007). The mediational models were examined for each hypothesized mediator.
Figure 2. Hypothesized mediational model. Only necessary parameters are included in the model. Bin
= binary outcome; Con = continuous outcome; Med = mediator; I = intercept; Iu = intercept of binary part of the model, Iy = intercept of continuous part of the model; S = slope; Su = slope of binary part of the model, Sy = slope of continuous part of the model; b1, b2u, b2y = coefficients making up the mediational model.
Results Descriptive Statistics
Descriptive statistics for the measures on the three measurements, separated by group, and tests for mean differences on the postintervention measurement are presented in Table 1. The tests for mean differences indicate that there were no significant differences between the NP group and TAU group on the postintervention measurement. Prevalence rates of
delinquency were also analyzed. On the preintervention measurement, 43% of the juveniles in the NP group had committed an offense against 41% of the juveniles in the TAU group. After three months, 30% of the juveniles in the NP group and 33% of the juveniles in the TAU had committed an offense. At last, on the postintervention measurement, 21% of the juveniles in the NP group and 23% of the juveniles in the TAU group had committed an offense. So there was a decrease in the delinquent behavior of the juveniles.
The correlations between the demographic characteristics, mediators and the outcome measure can be found in Table 1 in the Appendix. There were significant correlations between gender and deviant and prosocial peers, ethnicity and support and age and monitoring. Also the correlations between age, gender and authoritarian control and delinquency were
significant. The demographic characteristics that had a significant relationship with the outcome variable or mediators, were involved in the corresponding latent growth model.
Table 1
Means and Standard Deviations for the NP Group (N = 47) and TAU Group (N = 54) and Tests for Mean Differences on the Postintervention Measurement
Preintervention After 3 months Postintervention
Variable M SD M SD M SD ta Motivation 0.49 NP 2.21 0.40 2.16 0.40 2.22 0.31 TAU 2.29 0.41 2.28 0.31 2.26 0.40 Cognitive distortions 0.37 NP 2.66 0.69 2.36 0.56 2.39 0.62 TAU 2.58 0.69 2.53 0.66 2.44 0.73 Support 0.51 NP 3.34 0.91 3.57 0.88 3.41 0.96 TAU 3.52 0.91 3.56 0.95 3.52 1.08 Authoritative control 0.74 NP 3.60 0.76 3.68 0.62 3.59 0.72 TAU 3.84 0.65 3.81 0.62 3.69 0.74 Authoritarian control 0.44 NP 2.72 0.76 2.74 0.71 2.68 0.76 TAU 2.81 0.69 2.67 0.72 2.74 0.69 Monitoring -0.07 NP 2.83 0.62 2.86 0.60 3.05 0.62 TAU 2.89 0.61 2.94 0.55 3.04 0.60 Deviant peers 0.75 NP 1.71 0.87 1.64 0.67 1.52 0.64 TAU 1.63 0.67 1.72 0.79 1.64 0.89 Prosocial Peers -0.75 NP 3.40 0.63 3.20 0.96 3.43 0.85 TAU 3.40 0.72 3.36 0.69 3.30 0.98 Intensity of contact 0.08 NP 2.40 0.50 2.45 0.53 2.29 0.50 TAU 2.45 0.47 2.32 0.50 2.30 0.53 Delinquent behaviorb 1.93 NP 0.95 1.65 0.60 1.03 0.30 0.67 TAU 1.31 2.40 0.75 1.29 0.98 2.42
Note. NP = New Perspectives; TAU = treatment as usual.
a Test statistics are based on the mean differences between the NP and TAU group on the postintervention
measurement.
b
Due to missing values on this variable, NP group (N = 43) and TAU group (N = 52).
Investigation of Change during the Intervention and Intervention Effects
Latent growth modeling was conducted to investigate the change during the intervention and possible differences between the growth trajectories of the NP and TAU group.
The results of the latent growth models for the mediators are presented in Table 2. The 19
results revealed that there was a significant decrease in the intensity of contact with peers in both groups. There was a significant decrease in cognitive distortions in the NP group, but not in the TAU group. On the other mediators, there were no significant increases or decreases over time. The latent growth model of the mediator motivation could not be estimated, because the model was misspecified in the NP group. Apparently, there was no linear growth over time in the NP group.
To assess the differences between the growth trajectories of the NP group and TAU group, the initial levels and the growth rate of the mediators and outcome measure were compared by using a chi-square difference test for the comparison between a constrained and an unconstrained multiple-group latent growth model. The results of these comparisons are presented in Tables 2 and 3. Table 2 shows that there was a significant difference between the slope means of the NP group and TAU group on affiliation with deviant peers. The NP group showed a (nonsignificant) decrease over time, whereas the TAU group showed a
(nonsignificant) increase over time. On the other mediators, there were no significant
differences between the slope means of the two groups. There were no significant differences in intercept means between the NP and TAU group on the mediators. This indicated that there were no initial differences between both groups. Furthermore, there were no significant differences in intercept and slope variances between the groups, which means that the within-group variability in the initial scores and growth rates were comparable.
The results for the outcome measure are shown in Table 3. No fit indices (e.g., RMSEA, CFI, SRMR) were available for the two-part latent growth model. For the binary part of the model, chi-square tests for model fit were p > 0.05. The intercept means and variances of the binary and continuous part of the two-part latent growth model were not significantly different between the NP group and TAU group. The slope variances of both groups in the binary and continuous part of the two-part latent growth model could not be estimated, because the parameters were fixed at zero for identification purposes. The TAU group showed a significant negative slope mean on the binary part of the two-part latent growth model, whereas the NP group showed a nonsignificant negative slope mean. This indicates that the likelihood of reporting ever been involved in an offense significantly declined over time in the TAU group. However, this difference in slope means was not
significant. At the same time, in the continuous part of the two-part growth model, there was a significant difference between the two groups in their slope means. The slope mean of the NP group was negative (-0.143, SE = 0.099), although not significant. In the TAU group there was a slight, nonsignificant increase (0.195, SE = 0.135).
The demographic characteristics gender and age were included in the model as
covariates. In the nonrestricted, baseline model, there was an effect of gender on the intercept of the binary part (-1.340, SE = 0.656, p = 0.041) and the continuous part (-0.324, SE = 0.158, p = 0.041) of the model. This means that girls had significant lower start values compared to boys. In the experimental group, there was a significant effect of age on the slope in the binary part of the model (-1.180, SE = 0.479, p = 0.014). This means that the decrease over time in the likelihood of committing an offense was greater for juveniles older than 16 years, compared to juveniles younger than 16 years.
In sum, the results of the latent growth models indicate that there were no significant differences in the growth rate between the NP group and TAU group on the mediators, except for affiliation with deviant peers. Furthermore, there was a significant difference between the two groups in the decrease of the number of committed offenses.
Table 2
Parameter Estimates with Standard Errors, and the Difference between these Estimates for both Groups (NP and TAU), for the Latent Growth Models of the Mediators
Mediator NP1 TAU1 Δχ2(1) CFI
Cognitive distortions 0.968 Intercept mean 2.531 (0.086)** 2.603 (0.083)** 0.41 Intercept variance 0.172 (0.091) 0.193 (0.085)* 0.03 Slope mean - 0.128 (0.044)** - 0.071 (0.040) 1.06 Slope variance 0 0 0.19 Support 0.980 Intercept mean 3.498 (0.146)** 3.539 (0.119)** 0.05 Intercept variance 0.407 (0.163)* 0.744 (0.189)** 2.14 Slope mean 0.041 (0.063) - 0.004 (0.048) 0.33 Slope variance 0.018 (0.082) 0 0.41 Authoritative control 1.000 Intercept mean 3.680 (0.086)** 3.843 (0.074)** 2.36 Intercept variance 0.035 (0.103) 0.068 (0.089) 0.08 Slope mean - 0.053 (0.048) - 0.049 (0.047) 0.00 Slope variance 0 0 0.44 Authoritarian control 0.930 Intercept mean 2.826 (0.143)** 2.935 (0.113)** 0.70 Intercept variance 0.491 (0.170)** 0.302 (0.090)** 1.08 Slope mean - 0.044 (0.064) 0.084 (0.064) 2.56 Slope variance 0 0.066 (0.038) 1.57 Monitoring 1.000 Intercept mean 2.915 (0.097)** 2.940 (0.089)** 0.06 Intercept variance 0.197 (0.080) 0.214 (0.071)** 0.92 Slope mean 0.100 (0.056) 0.055 (0.042) 0.48 Slope variance 0 0 0.00 Deviant peers 1.000 Intercept mean 1.707 (0.114)** 1.711 (0.101)** 0.00 Intercept variance 0.323 (0.148)* 0.305 (0.119)* 0.05 Slope mean - 0.103 (0.057) 0.051 (0.046) 4.67* Slope variance 0.082 (0.048) 0.003 (0.047) 1.63 Prosocial Peers 0.929 Intercept mean 3.474 (0.105)** 3.455 (0.094)** 0.02 Intercept variance 0.285 (0.143)* 0.326 (0.123)** 0.01 Slope mean - 0.025 (0.069) - 0.063 (0.073) 0.15 Slope variance 0.048 (0.070) 0.066 (0.071) 0.05 Intensity of contact 0.980 Intercept mean 2.458 (0.060)** 2.440 (0.061)** 0.04 Intercept variance 0.044 (0.908) 0.165 (0.060)** 2.56 Slope mean - 0.077 (0.030)** - 0.085 (0.031)** 0.04 Slope variance 0.015 (0.035) 0 0.47
Note. CFI = comparative fit index; NP = New Perspectives; TAU = treatment as usual.
1 A value of zero means that this parameter estimate was negative, very small and nonsignificant. *
p < 0.05. ** p < 0.01.
Table 3
Parameter Estimates with Standard Errors, and the Difference between these Estimates for both Groups (NP and TAU), for the Two-Part Latent Growth Model of the Outcome Measure
Outcome measure NP TAU Δχ2(1)
Delinquent behavior: binary
Intercept mean 0.311 (0.618) na1 0.25
Intercept variance 0.963 (0.953) 11.483 (6.490) 2.70
Slope mean - 0.450 (0.393) - 0.628 (0.287)* 0.15
Slope variance na1 na1 na
Delinquent behavior: continuous
Intercept mean 0.219 (0.170) 0.452 (0.239) 0.87 Intercept variance 0.069 (0.079) 0.254 (0.143) 1.40
Slope mean - 0.143 (0.099) 0.195 (0.135) 4.09*
Slope variance na1 na1 na
Note. na = not applicable; NP = New Perspectives; TAU = treatment as usual.
1
Parameter was fixed at zero for identification purposes.
* p < 0.05.
Assessment of Mediation
There is evidence for mediation if each of the two paths of the mediational model (b1
and b2u or b1 and b2y in Figure 2) is significantly nonzero. This means that the path between
the group and the mediators and the path between the mediators and the outcome measure must be significantly different from zero. However, the latent growth models revealed that there was only an intervention effect on the mediator affiliation with deviant peers. This means that participation in NP was not related to a decrease or increase in the growth rate over time of the other mediators. Therefore, it was not necessary to test these mediational models. So only the mediational model for the affiliation with deviant peers was examined.
The latent growth model of the growth trajectory of the mediator affiliation with deviant peers and the two-part latent growth model of the growth trajectory of the outcome measure delinquency were combined into one parallel process latent growth model. Group membership (NP or TAU group) was introduced as a covariate to explain possible differences in growth trajectories. The relations between the growth factor (slope) of the mediator and the growth factors (intercept and slope) of the outcome measure were assessed. The intercepts of the mediator and the outcome measure were regressed on group membership, to control for initial differences between the two groups. Also age and gender were introduced into the model as covariates to control for differences between boys and girls and juveniles younger than 16 years and juveniles older than 16 years. The joint significant test of MacKinnon et al. (2002) was used to assess if there is evidence for mediation. No fit indices (e.g., RMSEA,
CFI, SRMR) were available due to the fact that a two-part latent growth model was part of the one parallel process latent growth model.
The results of the one parallel process latent growth model revealed that the path between group membership and the growth factor of the mediator (b1) was nonsignificant
(-0.098, SE = 0.070, p = 0.160). This means that participation in NP was not related to the growth rate in the affiliation with deviant peers. This result is not in agreement with the previous results of multiple-group analysis. The paths between the growth factor of the mediator and the growth factors of the binary (b2u) and continuous (b2y) part of the outcome
measure were also nonsignificant (3.641, SE = 2.988, p = 0.223, respectively 0.507, SE = 1.025, p = 0.621). Changes over time in the affiliation with deviant peers did not predict changes in the likelihood of committing an offense and the number of committed offenses. The tests of mediation effects with the joint significant test showed that the first mediated effect (b1b2u) and the second mediated effect (b1b2y) were both nonsignificant (0.357, 95% CI
[-1.610, 2.324], respectively 0.050, 95% CI [-0.771, 0.870]). To conclude, there was no evidence for mediation.
Discussion
The present study examined the effects of NP on the outcome measure delinquency and on the hypothesized mediators motivation, cognitive distortions and the social network. The study showed that NP was more effective compared with TAU in reducing the number of committed offenses. There were no effects found for the hypothesized mediators, except for affiliation with deviant peers. NP was more effective in reducing the affiliation with deviant peers. In both groups, there was a decrease over time in the intensity of contact with peers. In the NP group, there was a decrease in juveniles' cognitive distortions, whereas there was no decrease in the TAU group. In the TAU group there was a decrease in the likelihood of committing an offense, but this was not found in the NP group. However, these differences between the NP and TAU group were not significant. Furthermore, NP did not succeed in improving the motivation of the juveniles, their relationships with prosocial peers and the likelihood of committing an offense. Also there were no improvements found in parenting behavior (i.e., support, authoritative and authoritarian control, and monitoring).
We also tested whether the change during the intervention in the hypothesized mediators, mediates the effect of NP on juveniles' delinquent behavior. There was no evidence found for mediation. This means that the change during the intervention in the
mediators had no effect on the change over time in the likelihood of committing an offense and in the number of committed offenses.
NP was only more effective than TAU in decreasing the number of committed offenses and the affiliation with deviant peers. An explanation for not finding any difference between the two groups on the other measures could be the use of an active control condition. The TAU group was a heterogeneous group and even two juveniles received an evidence-based treatment. This could have resulted in an underestimation of the mean effect size. In addition, there is assumed that treatment as usual in the Netherlands is of better quality compared to other countries (e.g., the VS) (Mulder, Daalder, & Leeuw, 2013).
There were no improvements found in motivation, the likelihood of committing an offense, relationships with prosocial peers, support, authoritative and authoritarian control and monitoring. These results are partially comparable with the results of the evaluation study of Buysse et al. (2008) and Geldorp et al. (2004). They found an improvement in delinquent behavior, but less improvement in other domains, like relationships with friends. In the study of Nauta (2008), there was concluded that younger participants reoffend more often compared to older participants. This is in agreement with the results of the present study.
An explanation for the non-improvements in some domains can be found in the intensity of the program. In a meta-analytic study of De Vries and colleagues (in press), there was concluded that the intensity of prevention programs is related to the effectiveness. Highly intensive programs could have a negative effect on low-risk juveniles. This is related to the risk principle of the RNR model, where NP is based on. This principle states that the intensity of the program should be adjusted to the juvenile's risk for reoffending (Andrews et al., 1990; Andrews & Bonta, 2010). When evaluating juveniles in the NP program, several studies concluded that there is a subgroup with no delinquent behavior and/or a low risk for
reoffending (Geldorp et al., 2004; De Vries, Hoeve, & Asscher, 2014a). For these juveniles, the intervention may be too intensive and perhaps even counterproductive. However, there is also a subgroup of juveniles with persistent delinquent behavior and a high risk of reoffending (Buysse et al., 2008; Geldorp et al., 2004; De Vries et al., 2014a). For these juveniles a more intensive intervention is needed. In addition to this, results of the study of De Vries and colleagues (2014a) about the program integrity of NP, revealed that in 22% of the evaluated juveniles, there was a suspicion or diagnosis of mild mental handicap. In the present study, nearly a quarter of the juveniles (24%) experienced cognitive distortions. So it could be that the problems of the juveniles were too severe for the intervention. In fact, the referral of juveniles to more intensive and specific interventions after the NP program is one of the main
purposes of NP (Elling & Melissen, 2007). The creators of NP recognize that many juveniles who receive NP will need more intensive and specific interventions after the program (Elling & Melissen, 2007). To conclude, the including of juveniles with low or high risk of
reoffending or juveniles with severe problems in the NP program, may explain the non-improvements in several domains.
Another explanation for the non-improvements could be that NP intervenes in life domains, if necessary (Elling & Melissen, 2007). Some juveniles needed help with
relationships with peers, but perhaps others needed more help with relationships with their parents. Van der Braak and Konijn (2006) have investigated the implementation of NP in practice and concluded that the intervention is applied according to the method. However, they also concluded that the method was constantly adapted to the juvenile. This finding is also found in the study of Spapens, Van Wensch and Von der Furh (2000). They concluded that the rules of the intervention are not always applied uniformly. NP has a multimodal format, but a tailored approach is preferred, so not all modules are carried out as standard. This is also shown in the study of De Vries and colleagues (2014a). Originally, NP is an individually targeted intervention. But although the importance of a family-oriented approach is now acknowledged, family-oriented techniques are used less frequently compared to other more individual-oriented techniques, such as motivational interviewing (De Vries et al., 2014a). For example, in only 34% of the cases, parenting skills were addressed (De Vries et al., 2014a). To conclude, not carrying out all methods as standard, could explain why, in some domains, no improvements were found.
A final explanation may be found in the Transtheoretical Model of Behavior Change (TTM) of Prochaska & DiClemente (1984), where NP is based on. This model states that changing behavior is a long-term and intensive process, which involves progress through six stages of change: precontemplation, contemplation, preparation, action, maintenance, and termination (Prochaska & Velicer, 1997). The precontemplation stage is the stage in which someone has no intention to take action. In the contemplation stage, someone has the
intention to change. In the preparation stage, someone has the intention to take action and has already taken some steps. In the action stage, someone has made specific, overt modification in their behavior. The maintenance stage is the stage in which someone has made specific, overt modifications in their lifestyle and is working to prevent a relapse. At last, in the termination stage, someone has no longer the temptation to relapse. Prochaska, Redding, and Evers (2008) concluded that it takes at least six months before someone enters the
maintenance stage. Prochaska and Velicer (1997) studied interventions based on the TTM and 26
concluded that it takes time for participants to progress to the action stage. In fact, some treatments will produce effects months or years after the treatment ended. This is also known as 'sleeper effects', which can be defined as delayed effects of therapy (Bell, Lynes, & Kolvin, 1989). So it could be that the intervention has some delayed effects, which are not been detected yet.
This study has several limitations. A first limitation of the present study is that it targets only the short-term effectiveness of NP. Follow-up assessments will give more insight in the change over a long time. Follow-up assessments are also needed to prevent 'sleeper effects'.
A second limitation is that the present study only includes self-report measures. Self-report measures could result in social desirability bias, which is the tendency of some people to respond in a culturally acceptable manner (Podsakoff, MacKenzi, Lee, & Podsakoff, 2003). This has been partly overcome by using the cognitive distortions questionnaire. This
questionnaire contained a scale to measure social desirability bias. All juveniles scored low to normal on this scale. However, future studies should include measures from different sources, like parent reports and official judicial records, to enhance reliability.
A third limitation of the present study is that the power is too low. This may have resulted in an underestimation of the effects. Power calculations prior to the start of the study indicated that 90 juveniles per condition were required to detect differences in behaviors at post-test, assuming a medium effect size, an alpha of 0.05 and power of 0.95 (De Vries, Hoeve, Asscher, & Stams, 2014b). However, the NP group consisted of 47 juveniles and the TAU group of 54 juveniles. Shapiro (1995) concluded that at least 60 participants per group are needed to detect differences between psychosocial interventions. Both groups, the NP and TAU group, contained less participants.
The power of the present study is also too low to detect mediation effects. Cheong (2011) investigated the accuracy of estimates and statistical power for testing mediation with the parallel process latent growth modeling approach. He concluded that with three
measurements and a statistical power of 0.80, very large samples (e.g., 1000 participants) are needed to detect mediation effects. Future studies that will examine mediational processes should include more participants and more measurements.
A fourth limitation of the present study is directed to the measurement of motivation. Cronbach's alpha of the motivation questionnaire ranged from 0.39 to 0.61, which means that this questionnaire had unacceptable to poor internal consistency (Kline, 2000). Inferences based on this questionnaire are therefore not reliable.
A fifth limitation concerns the measurement of delinquency. Besides the comment that delinquency is based on self-reports, only information about participation in and the number of committed offenses is included. Seriousness and versatility of the committed offenses is unknown. De Vries and colleagues (in press) concluded that studies that measured
participation in and frequency of delinquent behaviors showed larger effect sizes of interventions, compared to studies that measured seriousness and versatility. Therefore, follow-up studies should also include information about the seriousness and versatility of the committed offenses.
Finally, the last limitation is that juveniles in the NP group received other
interventions, at the same time they received NP. More than half of the juveniles received at least one intervention and even nine juveniles received an evidence-based intervention. Therefore, it could not be concluded that the change over time in the mediators and outcome measure was due to the intervention.
In summary, the present study is the first RCT study that examined the effectiveness of a Dutch secondary prevention program. Besides the addition to the existing literature on prevention programs, the present study also examined hypothesized mediational models. It was found that NP was only more effective than TAU in decreasing the number of committed offenses and reducing the affiliation with deviant peers. On the other measures, NP was not more effective compared with TAU. There was no evidence found for mediation effects. Future studies on the effectiveness of NP should include more participants, different sources (e.g., parents, judicial records) and a follow-up period.
References
Allen, J. P., Moore, C., Kuperminc, G., & Bell, K. (1998). Attachment and adolescent psychosocial functioning. Child Development, 69, 1406-1419.
Andrews, D. A., & Bonta, J. (2010). Rehabilitating criminal justice policy and practice. Psychology, Public Policy, and Law, 16, 39-55.
Andrews, D. A., Bonta, J., & Hoge, R. D. (1990). Classification for effective rehabilitation: rediscovering psychology. Criminal Justice and Behaviour, 17, 19-52.
Armsden, G. C., & Greenberg, M. T. (1987). The inventory of parent and peer attachment: Individual differences and their relationship to psychological well-being in
adolescence. Journal of Youth and Adolescence, 16, 427-454.
Barriga, A. Q., & Gibbs, J. C. (1996). Measuring cognitive distortion in antisocial youth: Development and preliminary validation of the “How I Think” questionnaire. Aggressive Behavior, 22, 333-343.
Barriga, A. Q., Hawkins, M. A., & Camelia, C. R. (2008). Specificity of cognitive distortions to antisocial behaviours. Criminal Behaviour and Mental Health, 18, 104-116.
Barriga, A. Q., Landau, J. R., Stinson, B. L., Liau, A. K., & Gibbs, J. C. (2000). Cognitive distortion and problem behaviors in adolescents. Criminal Justice and Behavior, 27, 36-56.
Bell, V., Lyne, S. and Kolvin, I. (1989) Play group therapy: Processes and patterns and delayed effects. In Schmidt, M. H. and Remschmidt, H. (eds), Needs and Prospects of Child and Adolescent Psychiatry (pp. 149–163). Bern: Hogrefe and Huber Publishers. Bowman, M. A., Prelow, H. M., & Weaver, S. R. (2007). Parenting behaviors, association with deviant peers, and delinquency in African American adolescents: A mediated moderation model. Journal of Youth and Adolescence, 36, 517-527.
Brown, E. C., Catalano, C. B., Fleming, C. B., Haggerty, K. P. & Abbot, R. D. (2005). Adolescent substance use outcomes in the Raising Healthy Children Project: A two-part latent growth curve analysis. Journal of Consulting and Clinical Psychology, 73, 699-710.
Brown, B. B., Mounts, N., Lamborn, S. D., & Steinberg, L. (1993). Parenting practices and peer group affiliation in adolescence. Child Development, 64, 467-482.
Brugman, D., Nas, C. N., Van der Velden, F., Barriga, A. Q., Gibbs, J. C., Potter, G. B., & Liau, A. K. (2011). Hoe Ik Denk vragenlijst. Handleiding. Amsterdam: Boom test uitgevers.
Buist, M. D. K. L., Dekovic, M., Meeus, W., & Van Aken, M. A. (2004). Gehechtheid en internaliserend en externaliserend probleemgedrag bij adolescenten. Kind en Adolescent, 25, 80-91.
Buysse, W., Van den Andel, A., & Van Dijk, B. (2008). Evaluatie Nieuwe Perspectieven Amersfoort 2005-2007. Amsterdam: DSP-Groep BV.
Casey, S., Day, A., & Howells, K. (2005). The application of the transtheoretical model to offender populations: Some critical issues. Legal and Criminological Psychology, 10, 157-171.
Cheong, J. (2011). Accuracy of Estimates and Statistical Power for Testing Meditation in Latent Growth Curve Modeling. Structural Equation Modeling, 18, 195-211.
Cheong, J., MacKinnon, D. P., & Khoo, S. T. (2003). Investigation of mediational processes using parallel process latent growth curve modeling. Structural Equation Modeling, 10, 238-262.
Clingempeel, W. G., & Henggeler, S. W. (2002). Randomized clinical trials, developmental theory, and antisocial youth: Guidelines for research. Development and
Psychopathology, 14, 695-711.
DeGarmo, D. S., Eddy, J. M., Reid, J. B., & Fetrow, R. A. (2009). Evaluating mediators of the impact of the Linking the Interests of Families and Teachers (LIFT) multimodal preventive intervention on substance use initiation and growth across adolescence. Prevention Science, 10, 208-220.
Deković, M., & Meeus, W. (1997). Peer relations in adolescence: Effects of parenting and adolescents' self-concept. Journal of Adolescence, 20, 163-176.
Deković, M., Slagt, M. I., Asscher, J. J., Boendermaker, L., Eichelsheim, V. I., & Prinzie, P. (2011). Effects of early prevention programs on adult criminal offending: a
meta-analysis. Clinical Psychology Review, 31, 532-544.
Deković, M., Wissink, I. B., & Meijer, A. M. (2004). The role of family and peer relations in adolescent antisocial behaviour: comparison of four ethnic groups. Journal of
Adolescence, 27, 497-514.
De Vries, S. L. A., Hoeve, M., & Asscher, J. J. (2014a). Onderzoek naar de programma- integriteit van Nieuwe Perspectieven in Amsterdam. Amsterdam: Universiteit van
Amsterdam.
De Vries, S. L. A, Hoeve, M., Asscher, J. J., & Stams, G. J. J. M. (2014b). The effects of the prevention program 'New Perspectives' (NP) on juvenile delinquency and other life domains: study protocol for a randomized controlled trial. BMC Psychology, 2, 1-10.
De Vries, S. L. A, Hoeve, M., Assink, M., Stams, G. J. J. M., & Asscher, J. J. (in press). Practitioner Review: Effective ingredients of prevention programs for youth at risk of persistent juvenile delinquency–recommendations for clinical practice. Journal of Child Psychology and Psychiatry.
DiClemente, C. C., & Velasquez, M. M. (2002). Motivational interviewing and the stages of change. In W.R. Miller & S. Rollnick (Eds.), Motivational interviewing: Preparing people for change (2nd Ed. (pp. 201-216)). New York: Guilford Press.
Dishion, T. J., Andrews, D. W., & Crosby, L. (1995). Antisocial boys and their friends in early adolescence: Relationship characteristics, quality, and interactional process. Child Development, 66, 139-151.
Duncan, T. E., & Duncan, S. C. (2004). An introduction to latent growth curve modeling. Behavior Therapy, 35, 333-363.
Elling, M. W., & Melissen, M. (2007). Handboek Nieuwe Perspectieven. Woerden: Adviesbureau Van Montfoort.
Ellis, A. (1962). Reason and emotion in psychotherapy. Secaucus, NJ: Lyle Stuart. Ellis, A., & Dryden, W. (1997). The practice of rational emotive behavior therapy. New
York: Springer publishing company.
Geldorp, M., Groen, H., Hilhorst, N., Bürmann, A., & Rietveld, M. (2004). Evaluatie Nieuwe Perspectieven 1998-2003. Amsterdam: DSP-groep.
Gibbs, J. C., Barriga, A. Q., & Potter, G. (2001). The How I Think Questionnaire. Champaign, IL: Research Press.
Hess, B. (2000). Assessing program impact using latent growth modeling: A primer for the evaluator. Evaluation and Program Planning, 23, 419-428.
Hoeve, M., Dubas, J. S., Eichelsheim, V. I., Van der Laan, P. H., Smeenk, W., & Gerris, J. R. (2009). The relationship between parenting and delinquency: A meta-analysis.
Journal of Abnormal Child Psychology, 37, 749-775.
Kaplan, H. B., Johnson, R. J., & Bailey, C. A. (1987). Deviant peers and deviant behavior: Further elaboration of a model. Social Psychology Quarterly, 50, 277-284.
Kazdin, A. E. (2007). Mediators and mechanisms of change in psychotherapy research. Annual Review of Clinical Psychology, 3, 1-27.
Keijsers, L., Branje, S., Hawk, S. T., Schwartz, S. J., Frijns, T., Koot, H. M., ... & Meeus, W. (2012). Forbidden friends as forbidden fruit: Parental supervision of friendships, contact with deviant peers, and adolescent delinquency. Child Development, 83, 651-666.
