J Exp Criminol (2010) 6:131–157 DOI 10.1007/s11292-010-9091-7
Fast Track intervention effects on youth arrests and delinquency Conduct Problems Prevention Research Group
Published online: 16 April 2010 # Springer Science+Business Media B.V. 2010
Abstract This paper examines the effects of the Fast Track preventive intervention on youth arrests and self-reported delinquent behavior through age 19. High-risk youth randomly assigned to receive a long-term, comprehensive preventive intervention from 1st grade through 10th grade at four sites were compared to high-risk control youth. Findings indicated that random assignment to Fast Track reduced court-recorded juvenile arrest activity based on a severity weighted sum of juvenile arrests. Supplementary analyses revealed an intervention effect on the reduction in the number of court-recorded moderate-severity juvenile arrests, relative to control children. In addition, among youth with higher initial behavioral risk, the intervention reduced the number of high-severity adult arrests relative to the control youth. Survival analyses examining the onset of arrests and delinquent behavior revealed a similar pattern of findings. Intervention decreased the probability of any
Members of the Conduct Problems Prevention Research Group, in alphabetical order, include Karen L. Bierman, Department of Psychology, Pennsylvania State University; John D. Coie, Department of Psychology and Neuroscience, Duke University; Kenneth A. Dodge, Center for Child and Family Policy, Duke University; Mark T. Greenberg, Department of Human Development and Family Studies, Pennsylvania State University; John E. Lochman, Department of Psychology, The University of Alabama; Robert J. McMahon, Department of Psychology, University of Washington; and Ellen E. Pinderhughes, Department of Child Development, Tufts University. This work was supported by National Institute of Mental Health (NIMH) grants R18 MH48043, R18 MH50951, R18 MH50952, and R18 MH50953. The Center for Substance Abuse Prevention and the National Institute on Drug Abuse also have provided support for Fast Track through a memorandum of agreement with the NIMH. This work was also supported in part by Department of Education grant S184U30002, NIMH grants K05MH00797 and K05MH01027, and NIDA grants DA16903, DA017589, and DA015226. We are grateful for the close collaboration of the Durham Public Schools, the Metropolitan Nashville Public Schools, the Bellefonte Area Schools, the Tyrone Area Schools, the Mifflin County Schools, the Highline Public Schools, and the Seattle Public Schools. We greatly appreciate the hard work and dedication of the many staff members who implemented the project, collected the evaluation data, and assisted with data management and analyses. We particularly express appreciation to Jennifer Godwin for her work on data analyses for this paper.
132
Conduct Problems Prevention Research Group
juvenile arrest among intervention youth not previously arrested. In addition, intervention decreased the probability of a self-reported high-severity offense among youth with no previous self-reported high-severity offense. Intervention effects were also evident on the onset of high-severity court-recorded adult arrests among participants, but these effects varied by site. The current findings suggest that comprehensive preventive intervention can prevent juvenile arrest rates, although the presence and nature of intervention effects differs by outcome. Keywords Prevention . Arrests . Delinquency . Longitudinal . Juveniles
1 Introduction The age-crime curve for delinquency indicates that aggregate crime rates increase to a peak in the teenage years and then decline (Brandt 2006; Farrington et al. 2007), although some offenders continue their involvement in serious and persistent antisocial behavior well beyond adolescence (Cernkovich and Giordano 2001) and exhibit antisocial personality disorder in adulthood (Lahey et al. 2005). The developmental surge in delinquency is a key focus for prevention efforts, and the variations in the surge for boys versus girls (Butts and Snyder 2007) and for different methods for obtaining the data need to be carefully considered, as noted below. This paper will focus on the impact of a long-term, cross-site, multifaceted prevention trial on adolescent courtrecord arrests and self-reported delinquency. After considering methods for assessing delinquency, prior research findings for the Fast Track project will be overviewed. Analyses in this paper will examine the rates and onset of court-recorded arrests and selfreports of delinquent activity through the high school years for adolescents who had been randomly assigned to the Fast Track intervention or to a control condition. 1.1 Delinquency and source of data Different estimates of prevalence and peaks of delinquency can be found depending on the source of information used. Court records and youth self-reports are commonly used measures, and they both have advantages and disadvantages (Farrington et al. 2007). Self-reports Self-reports of delinquency are considered to be more accurate estimates of the true number of offenses that individual adolescents actually commit in the sense that youth are only arrested for a fraction of their offending behavior (Thornberry and Krohn 2001). The validity of self-reports has been noted because of their ability to predict future court referrals (Farrington et al. 1996). Youths also report large numbers of minor offenses that are not reflected in police or court records (Farrington et al. 2007). For example, youth self-reports of antisocial behavior may contain items such as “fighting” that, while fitting the technical definition of assault, do not necessarily indicate criminal behavior. Fighting is behavior that would lead to school discipline if discovered, but rarely to arrest, even if reported (Pellegrini 2003). Self-reports of delinquency may also be inaccurate either because youths are unwilling to admit their offenses (or cannot recall them) or because they have motives for exaggerating their involvement in
Fast track intervention effects on youth arrests and delinquency
133
certain types of crime. The latter is often evidenced in data reflecting improbably high self-reported frequencies for some youth, which then necessitates scaling procedures for analyzing these data. Youth who have been involved in an extended prevention trial, such as Fast Track, are aware of their connection to the research process, as are members of their family, and this can have a biasing effect on self-report data in several ways. It could lead participants to want to please the intervention staff by looking good and thus tend to underreport delinquent activity more than control participants. Or it could work the other way because intervention youth will have had greater extended contact with adult employees of the project and be more likely to trust them and to disclose more about themselves than would control participants. Either or both of these biases could be magnified by repeated interviewing over time. Police are blind to intervention status of an apprehended youth; therefore, arrest data are not subject to this bias. Court records of arrests Arrest data reflect an important subset of adolescent antisocial activities that are of particular importance for prevention efforts, due to their seriousness and cost to society. Prevention programs that target reductions in adolescent offending are particularly interested in reducing crimes as defined legally, given the seriousness of the costs and consequences associated with these crimes. Arrests reflect public priorities for defining and controlling crime, and hence provide an index of more serious criminal activities, as they are defined legally. However, official police and court records may underrepresent the true extent of juvenile delinquency because most crimes are never reported, and of those that are, many juvenile offenders are not arrested or adjudicated to court. Local jurisdictions also can vary in the criteria they set for arrests. A particular problem with arrest data for incarcerated youth is that once incarcerated, youths’ opportunities to be re-arrested are restricted, although this problem is also evident for youths’ self-reports of their delinquent behavior as well. Farrington et al. (2007) have examined the rates of court petitions and selfreported offending for boys in the Pittsburgh Youth Study. Clear differences in rates of offenses were evident (80 self-reports of offenses for every offense petitioned to court). As the youths moved into middle adolescence, more of their self-reported offenses became apparent in court records, but conversely, because of the increasing delinquency rates, they received court petitions for fewer of their offenses. When chronic offenders were determined either by their court records or by their selfreports, there was little overlap. Because their research originally focused on court and arrest records alone, early delinquency researchers concluded that delinquency was predominantly a lower class problem, a conclusion that has not been found to be supported in self-reports of delinquency (Farrington et al. 2007). Use of both self-reports and court-records of arrests Despite these differences in rates of offenses, most researchers have found that the correlated risk factors for court and self-reported delinquency are generally similar (Kirk 2006; West and Farrington 1973). Farrington et al. (2007) concluded that researchers should always measure both self-reports and official court records when studying offending because of the merits and disadvantages of each type of measure.
134
Conduct Problems Prevention Research Group
One consideration is the extent to which actual delinquent activity will be captured by the measure. Police do not apprehend all those who commit criminal acts, so police records will capture fewer of these acts. Numbers of youth arrested will be larger than those who are charged and still larger than those who go to trial and are convicted. Ideally, conviction would be the gold standard if one is concerned with validity because the arrestee would have been given the opportunity to prove innocence. Unfortunately, many other factors contribute to police and judicial decision-making in the path from apprehension to trial and one must be reconciled to the idea that convictions capture just a small amount of adolescent criminal activity. Prevention trials involve far fewer participants than surveys of crime, and numbers become of great importance in deciding on an outcome measure for prevention trial research. For this reason, arrests are to be preferred to convictions, and self-reports preferred to arrest data, even though there are biases operating at all levels of these various measures. 1.2 Prevention of delinquency In response to the evolving risk factors for adolescent delinquency, prevention programs must target both the promotion of age-related individual competencies and the promotion of protective contextual supports (Conduct Problems Prevention Research Group [CPPRG] 1992, in press). Elementary school prevention programs can improve children’s ability to tackle the new social challenges of adolescence. However, for high-risk children living in unstable and risky contexts without effective protective supports, the challenges of adolescence may also undermine the gains produced by early preventive efforts. For example, in adolescence, peer effects on youths’ problem behaviors become more pronounced. Alienation from conventional sources of social support from parents, teachers, and nondeviant peers can lead high-risk youth to join with other adolescents like themselves (e.g., Cairns et al. 1988). Adolescents who associate with deviant peers have a substantially increased risk for adolescent problem behaviors, as adolescents reinforce each others’ antisocial beliefs within deviant peer groups (e.g., Dishion et al. 1994; Dodge et al. 2006). Thus, a successful program for preventing serious antisocial problems among high-risk children may require a long-term intervention commitment that addresses appropriate developmental risks from childhood through adolescence. Some existing evidence suggests that comprehensive preventive interventions can reduce the rates of delinquency. The most effective preventive interventions for children with conduct problem behaviors typically have multiple components that address parent and youth processes to reduce the likelihood of later problems. Multiple-component interventions have had documented reductions in aggressive behavior and self-reported delinquency (e.g., Lochman and Wells 2003, 2004; Vitaro et al. 1999) and have had significant effects (Eddy et al. 2003; Hawkins et al. 2005) or nonsignificant trend effects (Boisjoli et al. 2007) in reducing rates of criminal records. Despite these encouraging findings, none of these multi-component programs focused their implementation specifically on children who showed the highest levels of aggressive and conduct problem behavior at home and school at entry into elementary school, and thus were at greatest risk for life-course persistent antisocial behavior.
Fast track intervention effects on youth arrests and delinquency
135
The Fast Track prevention program is a multisite, multicomponent intervention program targeting those children at the highest risk for life-course persistent conduct problems (Conduct Problems Prevention Research Group 1992). The multiple-gate screening procedure that was employed in kindergarten, which used teacher and parent ratings, has been effective at identifying children at risk for sustained trajectories of conduct problems (Hill et al. 2004; Lochman & Conduct Problems Prevention Research Group 1995). The Fast Track preventive intervention design targeted the primary risk factors for antisocial behavior, including deficient social problem-solving and emotional coping skills, poor peer relations, weak academic skills, disruptive classroom environments, poor parenting practices, and poor home-school relations, with components that addressed social and classroom risk factors (Bierman et al. 1996) as well as family risk factors, including problematic communication between parents and school (McMahon et al. 1996). This unusually comprehensive and longterm intervention began in 1st grade and continued through 10th grade. The initial analyses of the Fast Track program indicated that the program had a significant impact on the proximal outcomes that were assessed in the elementary school phase of the project (see Table 1 for a summary of significant Fast Track outcomes in prior papers). For example, at the end of the 1st grade, in comparison to the high-risk control children, high-risk intervention children displayed significantly greater improvements in behavior problems (increased compliance and prosocial behavior by parent and teacher report) and significantly lower rates of aggressive, oppositional behaviors at school (by teacher report) (Conduct Problems Prevention Research Group 1999). In addition, the intervention children displayed better academic and social skills, with greater word attack skills and higher language arts grades, more positive peer interactions (by observer ratings), higher social preference (by sociometric nominations), and improved social-cognitive and emotion skills (assessed by child interviews). Improvements in parenting were evident on warmth (by observer report), reduced use of physical punishment (by parent report), and improved parental involvement at school (by teacher report). Analyses of program outcomes through the end of elementary school provided evidence, with small effect sizes, for the continuing positive outcomes from Fast Track program. By the end of grade 3, compared to children assigned to the control group, intervention children displayed fewer conduct problems (by teacher and parent report), and parents reported less use of physical punishment and greater improvements in their parenting skills (Conduct Problems Prevention Research Group 2002a, b). Children assigned to the Fast Track preventive intervention were significantly less likely to be identified as clinical cases in person-centered analyses than were children in the control group (63% versus 73%, respectively). In analyses of children’s outcomes in the 4th and 5th grades, Fast Track had a significant but modest influence on children’s rates of social competence and social cognition problems, problems with deviant peers, and conduct problems in the home and community, compared to control children (Conduct Problems Prevention Research Group 2004). However, intervention effects on children’s problem behaviors in school were no longer significant during this time period. Intervention effects faded further during the middle school years, with sustained effects evident only on parent-rated hyperactive behaviors and on youth self-reported delinquent behaviors in the 7th grade (but no other positive intervention effects on 15 other
136
Conduct Problems Prevention Research Group
Table 1 Summary of significant child outcomes from prior Fast Track publications Study
Grade assessed
Behavioral outcomes
Other outcomes
Conduct Problems Prevention Research Group 1999
1
- increased compliance
- more positive peer interactions
- increased prosocial behavior at school and home
- higher social preference
- reduced aggressive, oppositional behavior at school
- improved social-cognitive and emotional skills - higher language arts grades - improved word attack skills
Conduct Problems Prevention Research Group 2002a
3
Conduct Problems Prevention Research Group 2004
4–5
(Conduct Problems Prevention Research Group in press)
6–8
Conduct Problems Prevention Research Group 2007
3, 6, 9
- reduced aggression at school and home
- lower use of special education
- fewer clinical cases
- improved problem solving skills
- reduced conduct problems in home and community
- improved social competence and social cognition - reduced involvement with deviant peers
- reduced hyperactivity (7th grade only)
- more involvement with deviant peers (8th grade only)
- less self-reported delinquent behavior (7th grade only) - less self-reported delinquency (9th grade) - less conduct disorder but only for the highest risk
variables assessed in the middle school years) (Conduct Problems Prevention Research Group in press). It is important to note that an intervention effect on selfreported delinquency, although not evident in 8th grade, did reemerge in the 9th grade (Conduct Problems Prevention Research Group 2007). In addition, in contrast to the findings at the end of elementary school, Fast Track children in 8th grade were more likely than control children to be involved with peers who were in engaging in deviant behavior, and this deviant peer involvement may have mitigated some of the earlier intervention effects. Although we have not found consistent patterns of moderation (by variables such as gender, race, site, and cohort) of intervention effects in the elementary and middle school years, we have found that severity of child’s initial risk score at kindergarten has moderated intervention effects on externalizing psychiatric disorders (Conduct Problems Prevention Research Group 2007). Significant interaction effects between intervention and initial risk level were found after grades 3 and 6, but most strongly after grade 9. Among the highest-risk group (top 3%) in grade 9, assignment to intervention was responsible for preventing 75% of conduct disorder cases, 53% of ADHD cases, and 43% of any externalizing disorder cases. In contrast, the intervention had no impact on the diagnoses of children who were initially at only moderate levels of risk. Similar findings were obtained with an antisocial behavior
Fast track intervention effects on youth arrests and delinquency
137
score, based on the youth’s self-reported delinquency (although there was also a main effect of intervention). 1.3 Purpose of this study This study represents an extension of the series of analyses that have investigated the efficacy of the Fast Track intervention. In this study, we examined the impact of the Fast Track intervention on court arrests based on administrative data and self-report of delinquency through the high school years. Court records of arrests were a key planned outcome for Fast Track since its inception, and had not yet been examined in prior Fast Track outcome papers. Due to the differences in how arrests for juvenile crimes proceed versus arrests for adult crimes (Elrod and Ryder 2005), with greater latitude and variability in whether criminal acts are defined as arrests or not for juvenile crimes, we examined the court records separately for arrests adjudicated as juvenile versus adult crimes. It was hypothesized that youth in the Fast Track intervention condition would have lower levels of the three outcome variables (juvenile arrests, adult arrests, youth self-report of delinquency) than youth in the control condition, and that these condition differences would be evident in analyses of the frequency of antisocial acts and of the onset of antisocial acts.
2 Method 2.1 Participants Schools within the four sites were selected as high risk based on crime and poverty statistics of the neighborhoods that they served. Within each site, the schools were divided into multiple sets matched for demographics (size, percentage free or reduced lunch, ethnic composition), and the sets were randomly assigned to intervention and control conditions. Using a multiple-gating screening procedure that combined teacher and parent ratings of disruptive behavior (Lochman & Conduct Problems Prevention Research Group 1995), all 9,594 kindergarteners across three cohorts (1991–1993) in these 54 schools were screened initially for classroom conduct problems by teachers, using the Teacher Observation of Child AdjustmentRevised (TOCA-R) Authority Acceptance Score (Werthamer-Larsson et al. 1991). Those children scoring in the top 40% within cohort and site were then solicited for the next stage of screening for home behavior problems by the parents, using items from the Child Behavior Checklist (Achenbach 1991a) and similar scales, and 91% agreed (n=3,274). The teacher and parent screening scores were then standardized and combined into a sum score, based on screening a representative sample of approximately 100 children within each site (which also served as a normative comparison) and then summed to yield a total severity-of-risk screen score. Children were selected for inclusion into the study based on this screen score, moving from the highest score downward until desired sample sizes were reached within sites, cohorts, and conditions. Deviations were made when a child failed to matriculate in the first grade at a core school (n =59) or refused to participate (n=75), or to accommodate a rule that no child would be the only girl in an intervention group.
138
Conduct Problems Prevention Research Group
The outcome was that 891 children (n=445 for intervention and n = 446 for control) participated. Note that these levels of problems are defined relative to other children in these high-risk schools. On the kindergarten Teacher's Report Form of the Child Behavior Checklist (TRF; Achenbach 1991b), which provides national norms, the average Externalizing T-score (available for 88% of the high-risk sample) was 66.4, and 76% of these children scored in the clinical range (T-scores of 60 or higher). Children’s screen scores in kindergarten have been found to be predictive of their externalizing behaviors in first grade (Lochman & Conduct Problems Prevention Research Group 1995) and at the end of elementary school (Hill et al. 2004). The screening score had sufficient sensitivity and specificity in predicting externalizing behaviors 5 years later, at the end of elementary school, that it met criteria for identifying high-risk children for preventive intervention (Hill et al. 2004). The mean age of the participants was 6.5 years (SD=0.48) at the time of identification. Across all sites, the sample was primarily comprised of African American (51%) and European American (47%) participants, with 2% of other ethnicity (e.g., Pacific Islander and Hispanic), and was gender mixed (69% boys). The distribution of the sample by gender, race, site, and intervention status can be found in Table 2. The sample was skewed toward socioeconomic disadvantage: 58% were from single-parent families, 29% of parents were high school dropouts, and 40% of the families were in the lowest socioeconomic class (representing unskilled workers) as scored by Hollingshead (1975). Only 32% of the sample was within the middle-class range (Hollingshead categories 2 and 3), in comparison to rates of up to 75% in these two categories in some community samples (e.g., Reinherz et al. 2006). As can be seen in Table 2, at all but the Nashville site, the control and intervention samples had comparable numbers of African American and European American children. In Nashville, 73% of the intervention children were African-American, but 56% of the control children were African-American. Unfortunately, when sets of schools were randomized to intervention versus control in Nashville, there were more AfricanAmerican children in the intervention schools and more European American children in the control schools. In addition to the high-risk sample, a stratified normative sample of 387 children was identified from the control schools to represent the population-normative range of risk scores (based on teacher ratings only) and was followed over time. These children were not a part of the major analyses but are included here so that outcomes for the intervention group may be contrasted with those of the normative sample to determine whether they reached normative rates. Written consent from parents and oral assent from children were obtained. Parents were paid for completing interviews, and intervention-group parents were paid for group attendance. All procedures were approved by the Institutional Review Boards of participating universities. To improve the precision of the estimates of intervention effects, guard against any departures from randomization, and protect against differential attrition, 11 variables were measured prior to the initiation of intervention and were included as covariates in outcome analyses. These variables measured children’s baseline behavior problems, family demographics and social ecology, children’s cognitive and social skills, and parenting. Instruments and items are described at www. fasttrackproject.org. Previous analyses confirm no statistical difference between the
Fast track intervention effects on youth arrests and delinquency
139
Table 2 Sample by gender, race, site, and intervention status Number of participants African-American Intervention
Non-African-American Control
Norm
Intervention
Control
Norm
Boys Durham
82
75
44
7
4
6
Nashville
64
39
24
22
33
25
Pennsylvania
2
1
1
76
70
48
Seattle
31
29
20
38
45
29
Girls Durham
19
27
43
2
3
7
Nashville
19
26
24
9
18
27
Pennsylvania
1
2
34
39
49
Seattle
18
17
21
18
28
12
Pennsylvania = rural Pennsylvania, Norm = normative sample
intervention and control samples for pre-intervention scores (Conduct Problems Prevention Research Group 2002a, b, in press). For this paper, two data sources were used: administrative records requiring youth consent and the Self-Report of Delinquency (Elliott et al. 1985) measure. Access to juvenile court records required the consent of the youth, yielding missing data, while adult records are public, so complete data were gathered. For juvenile records, 9% of intervention children and 14% of control children did not provide consent. Among those who provided consent for the collection of juvenile arrest records, t-tests were conducted to determine whether the intervention and control groups differed for any of the 11 baseline variables. Only one of 11 tests were significant (WoodcockJohnson letter-word identification scale), a pattern slightly above chance. Twenty-six percent of intervention youth and 29% of control youth did not participate in the follow-up interview by grade 12. Only one of the 11 tests using the youth who participated was significant (parent rating of children’s social competence). We also examined whether there were overall differences between attritted vs. retained youth. Of attritted vs. retained tests for the 11 covariates, two were significant for juveniles whose court records were checked (family satisfaction; friendship satisfaction) and one was significant for children who reported on their delinquency (socioeconomic status). It was concluded that there were no substantial differences between retained versus attritted youth, and that attrition did not lead to meaningful differences in preintervention characteristics of intervention and control groups (complete tables with tests for attrition bias are available upon request). 2.2 Intervention procedures Elementary school phase (grades 1–5) During the elementary school phase of the intervention (grades 1–5), all families were offered parent training with home
140
Conduct Problems Prevention Research Group
visiting, academic tutoring, and child social skills training. Parent and child group interventions were conducted during a 2-h “enrichment program.” These sessions include social skill training “friendship groups” led by educational coordinators for high-risk children (Bierman et al. 1996), parent-training groups for parents led by family coordinators, and guided parent-child interaction sessions (Parent-Child Sharing Time) (McMahon et al. 1996). In the 1st grade, paraprofessional tutors also provided three 30-min sessions to strengthen emergent literacy skills, along with a weekly peer-pairing session to improve friendships with classmates. The enrichment programs were held weekly during grade 1 for 22 sessions, biweekly during grade 2 for 14 sessions, and monthly during grades 3–5 for nine sessions each year. In addition, individual support was provided through home visiting (Dodge 1993) to help parents generalize the skills presented in the group setting and to address individual needs. After grade 1, criterion-referenced assessments were used to adjust the dosage of some components (tutoring, home visiting, and peer coaching) to match family and child need. In addition to indicated interventions, a universal intervention (a modified, grade-level version of the PATHS curriculum that had been published by Kusche and Greenberg 1993) was provided to the classrooms in intervention schools through the elementary school years (grades 1–5), in order to promote social and emotional competence and a more competent and less aggressive social ecology. The universal intervention included weekly teacher consultation for lessons and classroom behavior management (Bierman et al. 1996).
Middle and early high school phase (grades 6–10) There were three standard prevention activities offered to all Fast Track intervention children during middle school: the middle school transition program, parent and youth groups on adolescent topics, and youth forums. Adolescent developmental issues were addressed with four meetings for parents and youth during 6th grade. Parent groups focused on issues such as positive involvement and monitoring, and youth groups focused on issues such as coping with peer pressure. Parents and youth met together in groups to address romantic relationships and sex education, alcohol, tobacco and drugs, and vocational goal setting. In grades 7 and 8, eight Youth Forums based on Oyserman’s (2000) program were held with youth in small groups to address vocational opportunities, budgeting and life skills, job interview skills, and summer employment opportunities. In grades 7– 10, individualized intervention plans were developed and implemented with each youth, based on regular assessments of risk and protective factors, conducted three times during each year. Ratings were made by project intervention staff of intervention children and their families every 4 months in four domains of functioning (parent monitoring and positive involvement; peer affiliation and peer influences; academic achievement and orientation; social cognition and identity development). Based on these ratings, youth and families either received the base level of intervention contact (once per month), or additional contact in interventions related to the targeted domain (e.g., academic tutoring, mentoring, support for positive peer-group involvement, home visiting and family problem solving, and liaisons with school and community agencies) for up to several hours more per month.
Fast track intervention effects on youth arrests and delinquency
141
Intervention participation Participation was defined as attendance at one or more group sessions––96% of parents and 98% of children participated during grade 1. Of these families, 79% of parents and 90% of children attended at least 50% of all sessions. In grade 2, 88% of parents and 92% of children participated, with 79% of parents and 87% of children attending at least 50% of all sessions. In grade 3, 80% of parents and 86% of children participated, with 78% of parents and 84% of children attending at least 50% of all sessions (Conduct Problems Prevention Research Group 2002b). The proportion of families unable to participate in the intervention increased modestly across the years, which was primarily due to moves out of the area. In the last year of the group sessions (grade 6), 43 of the 445 intervention families (10%) did not participate but had still received the majority of the services in previous years. Intervention fidelity was ensured by manualization of all components, regular cross-site training and communication, weekly staff training, and ongoing clinical supervision. Outside interventions were neither encouraged nor discouraged and were assumed to occur at the same rate for intervention and control groups. The control condition was a “treatment as usual” comparison that included regular school prevention programs to the extent that schools chose to use them. 2.3 Outcome measures There are two measures of criminal activity over time: administrative court records and self-report of delinquency. Court records Juvenile and adult arrest information was collected from the court system in the child’s county of residence and surrounding counties through age 19. A court record of arrest indicates any crime for which that youth was arrested and adjudicated, with the exception of probation violations (which were inconsistently reported in courts across the four sites) and referrals to youth court diversion programs for very young first-time offenders (starting at age 11). Other offenses leading to youth diversion programs were included as long as there was an identified arrest in the records. We obtained records of arrests rather than of charges; if an individual was charged but then the charge was dropped, we do not have a record of the charge. Court-record conviction data were also obtained, and indicated that 65% of arrests led to convictions. However, the conviction and arrest data were so highly correlated (.94 for males, .91 for females) that it would have been duplicative to analyze conviction data as well as arrest data. By local law, youth were typically seen in juvenile court through age 15 in Durham, NC, and through age 17 at the other three sites, although in the year prior to the transition to adult court, increasing numbers of youth were being sent to adult court with considerable variation across the sites (5% offending at age 15 are sent to adult court in Durham, 6% at age 17 in Nashville went to adult court, 27% in rural Pennsylvania, and 57% in Seattle). Across all four sites, relatively few adolescents age 15 or younger were sent to adult court (8% in Durham, 2.5% in Nashville, 4% in rural Pennsylvania, and 6% in Seattle). The data collected from the courts included a description of the offense, the date of the offense, the adjudication date for the arrest, and the outcome of the arrest. To capture both frequency and severity of the crimes
142
Conduct Problems Prevention Research Group
for which youth were arrested, we created a lifetime severity weighted frequency of juvenile and adult arrests (Cernkovich and Giordano 2001). Each offense for each arrest was assigned a severity score ranging from 1 to 5, using a cross-site coding manual developed from the entire list of offenses coming from all four sites. Severity level 5 included all violent crimes such as murder, rape, kidnapping, and first-degree arson. Severity level 4 contained crimes involving serious or potentially serious harm and included assault with weapons and first-degree burglary. Severity level 3 crimes reflected medium severity, such as simple assault felonious breaking and entering, possession of controlled substances with intent to sell, and firesetting. Severity level 2 included low-severity crimes such as breaking and entering, disorderly conduct, possession of controlled substance, shoplifting, vandalism, and public intoxication. Severity level 1 involved status and traffic offenses. We then summed the severity level of the most severe offense from each arrest from grade 6 through grade 12 (separately for adult and juvenile arrests). Self-Report of Delinquency The Self-Report of Delinquency measure (Elliott et al. 1985) has been used in numerous prior studies including The National Survey of Youth and the Pittsburg Youth Study. It was administered in the current study from grades 7 through 12 and captured the number of times in the past year the respondent committed 34 different offenses. Offenses range from lying about your age to get something to attacking someone with the intent to hurt. Following earlier use of the measure (e.g., Elliott et al. 1985), the items in each grade were capped at three to avoid creating an extremely skewed distribution. To create an annual scale capturing both frequency and severity of delinquency, each item was multiplied by a weight capturing the severity of the crime (the same weighting structure used for arrests) and the 34 weighted items were summed. The final outcome measure sums the products for all items across grade 7 through 12 to create a lifetime measure of delinquency. 2.4 Statistical approach and treatment of missing data To account for sporadic missing data, we used PROC MI in SAS v9.1 to impute 100 data sets. The imputation model included all outcome variables as well as indicators for race, gender, cohort, site, 11 continuous pre-intervention covariates, initial severity-of-risk score, indicators for whether the youth’s mother and father were arrested in grades 3, 4, and 6 through 11, parent report of days youth spent incarcerated in the past year for grades 8–12, and days sentenced to jail based on administrative records in each grade (grades 6–12). The pre-intervention covariates were identified based on theoretical and empirical research on the antecedents of conduct problem behaviors. The severity-weighted juvenile arrests, severityweighted adult arrests, and the severity-weighted self-reported offenses for each grade were also included in the imputation model. One hundred imputations should be ample given the degree of missing information in these analyses (Schafer and Graham 2002). Given the low rate of attrition and few differences between attritted and continuing youth, the assumption that data are missing at random is plausible. Missing data rates for individual variables in the model ranged from 0 to 49%. Data were missing for more than 30% of the sample on only six variables. These variables included a report of whether the child’s father was arrested in grades 6–12.
Fast track intervention effects on youth arrests and delinquency
143
The imputation was performed separately for the intervention and control conditions to preserve interactions between intervention and other variables. Because imputation from a covariance matrix, as PROC MI uses, preserves only the estimated variances and covariances when generating imputed data sets, other effects, such as interactions between intervention status and other variables in the model, tend to be artificially weakened by the conventional imputation process. By imputing separately by intervention status, we allowed the covariances among the other variables to differ by intervention, which is the definition of an interaction. We assessed the impact of intervention on delinquent behavior by analyzing each of the three cumulative measures of delinquency. The means by intervention status are found in Table 3, and the means broken down by site can be found in Table 4. The three measures of delinquency are: severity-weighted juvenile arrests based on court records (juvenile arrest index), severity-weighted adult arrests based on court records (adult arrest index), and severity-weighted self-reported offenses (selfreported offense index). The normal distribution of the self-reported offense index allowed the use of a standard linear regression model. For the arrest indices, however, the large proportion of the sample with no juvenile or adult arrests required an alternative modeling technique. Consequently, we divided the arrest outcomes into categories. For juvenile arrests, the categories were 0, 1–4, 5–8, and 9 or more. This division yields meaningful categories based on our severity scale. The 0 category captures non-offenders, the 1–4 category captures minor offenders, the 5–8 category captures
Table 3 Imputed group means (standard deviations) through grade 12, by intervention status Intervention
Control
Normative
n=445
n=446
n=387
Mean
SD
Mean
SD
Mean
SD
Juvenile arrest index
3.18
0.29
3.27
0.28
2.02
0.26
Adult arrest index
1.89
0.05
1.82
0.04
1.21
0.04
Self-reported offense index
54.99
4.22
55.15
4.06
44.93
4.13
Severity 1/2 arrests
0.61
0.06
0.69
0.06
0.53
0.06
Severity 3 arrests
0.47
0.05
0.54
0.05
0.30
0.04
Severity 4/5 arrests
0.17
0.03
0.15
0.02
0.07
0.02
Severity 1/2 arrests
0.47
0.00
0.42
0.00
0.28
0.00
Severity 3 arrests
0.22
0.00
0.22
0.00
0.13
0.00
Severity 4/5 Arrests
0.11
0.00
0.11
0.00
0.08
0.00
Juvenile arrests
Adult arrests
Self-reported offenses Severity 1/2 offenses
19.50
1.12
20.13
1.15
17.35
1.04
Severity 3 offenses
3.99
0.31
4.26
0.33
3.69
0.36
Severity 4/5 offenses
4.29
0.33
4.25
0.28
3.35
0.33
144
Conduct Problems Prevention Research Group
Table 4 Means by site and intervention status Intervention
Control
Normative
n
Mean
SD
n
Mean
SD
n
Mean
SD
Durham
110
2.16
0.43
109
2.87
0.48
100
1.56
0.41
Nashville
114
5.79
0.68
116
4.73
0.60
100
3.72
0.67
Pennsylvania
113
1.89
0.45
112
2.41
0.45
98
0.94
0.29
Seattle
108
2.81
0.64
109
2.99
0.65
89
1.79
0.59
Durham
110
0.35
0.07
109
0.41
0.07
100
0.29
0.07
Nashville
114
1.15
0.15
116
1.03
0.15
100
0.95
0.17
Pennsylvania
113
0.53
0.12
112
0.84
0.14
98
0.51
0.12
Seattle
108
0.40
0.10
109
0.44
0.11
89
0.34
0.12
Durham
110
0.26
0.06
109
0.47
0.08
100
0.20
0.06
Nashville
114
0.75
0.10
116
0.81
0.12
100
0.58
0.12
Pennsylvania
113
0.36
0.08
112
0.37
0.08
98
0.09
0.04
Seattle
108
0.48
0.11
109
0.51
0.12
89
0.34
0.10
Durham
110
0.18
0.05
109
0.22
0.05
100
0.14
0.05
Nashville
114
0.32
0.08
116
0.15
0.04
100
0.10
0.04
Pennsylvania
113
0.04
0.02
112
0.06
0.03
98
0.01
0.01
Seattle
108
0.13
0.04
109
0.18
0.05
89
0.04
0.03
Durham
110
3.14
0.35
109
3.24
0.28
100
2.52
0.24
Nashville
114
1.93
0.17
116
1.22
0.07
100
1.38
0.23
Pennsylvania
113
1.09
0.06
112
1.05
0.14
98
0.14
0.01
Seattle
108
1.44
0.13
109
1.82
0.12
89
0.72
0.05
Durham
110
0.61
0.01
109
0.57
0.01
100
0.57
0.01
Nashville
114
0.33
0.01
116
0.28
0.00
100
0.23
0.01
Pennsylvania
113
0.33
0.01
112
0.20
0.01
98
0.01
0.00
Seattle
108
0.62
0.02
109
0.63
0.01
89
0.30
0.01
Durham
110
0.39
0.01
109
0.36
0.01
100
0.21
0.00
Nashville
114
0.24
0.01
116
0.16
0.00
100
0.20
0.01
Pennsylvania
113
0.17
0.00
112
0.20
0.00
98
0.04
0.00
Seattle
108
0.10
0.00
109
0.17
0.00
89
0.08
0.00
Durham
110
0.20
0.01
109
0.25
0.00
100
0.20
0.01
Nashville
114
0.15
0.00
116
0.07
0.00
100
0.09
0.00
Juvenile arrests Index
Severity 1/2
Severity 3
Severity 4/5
Adult arrests Index
Severity 1/2
Severity 3
Severity 4/5
Fast track intervention effects on youth arrests and delinquency
145
Table 4 (continued) Intervention n
Mean
Control SD
n
Normative Mean
SD
n
Mean
SD
Pennsylvania
113
0.01
0.00
112
0.03
0.00
98
0.00
0.00
Seattle
108
0.06
0.00
109
0.09
0.00
89
0.01
0.00
Durham
110
51.41
7.62
109
49.73
6.89
100
46.39
9.32
Nashville
114
76.19
9.83
116
59.45
7.91
100
56.45
8.78
Self-reported offenses Index
Pennsylvania
113
28.41
6.00
112
42.25
8.27
98
17.33
4.94
Seattle
108
64.07
9.08
109
69.23
9.64
89
60.74
8.37
Durham
110
15.72
1.91
109
16.26
1.87
100
14.31
1.84
Nashville
114
22.98
2.50
116
20.88
2.07
100
20.47
2.15
Pennsylvania
113
13.75
1.80
112
17.77
2.32
98
9.58
1.24
Seattle
108
25.69
2.50
109
25.61
2.61
89
25.81
2.69
Durham
110
4.07
0.65
109
3.76
0.63
100
3.95
0.85
Nashville
114
5.91
0.73
116
5.20
0.63
100
5.34
0.86
Pennsylvania
113
1.92
0.37
112
3.21
0.71
98
1.51
0.31
Seattle
108
4.05
0.65
109
4.83
0.65
89
3.94
0.65
Durham
110
4.44
0.66
109
4.16
0.55
100
3.92
0.82
Nashville
114
6.55
0.76
116
5.11
0.57
100
4.43
0.74
Severity 1/2
Severity 3
Severity 4/5
Pennsylvania
113
1.86
0.33
112
2.99
0.52
98
1.27
0.28
Seattle
108
4.29
0.70
109
4.73
0.58
89
3.77
0.57
Pennsylvania=rural Pennsylvania
moderate offenders, and the final category captures severe and violent offenders. Among youth in the 1–4 category, 57% were only arrested for severity 1 or 2 offenses (36% were arrested for a severity 3 offenses and 6% were arrested for a severity 4 offenses). Among youth in the 5–8 category, the worst offense for which the majority (58%) were arrested was a severity 3 offense, whereas 13% were arrested for several severity 1 or 2 offenses and 29% were arrested for a severity 4 or 5 crime. Finally, among those in the 9 or more category, 54% were arrested for at least one severity 4 or 5 offense (44% were arrested for multiple offenses that included at least one severity 3 offense and only 2% were arrested for only severity 1 and 2 offenses). The proportion of the sample in each of these categories by intervention status and site are displayed in Table 5. For adult arrests, the categories were 0, 1–3, 4–6, and 7 or more. Similar to the juvenile categories, the 0 category captures non-offenders, the 1–3 category captures minor offenders, the 4–6 category captures moderate offenders, and the final
146
Conduct Problems Prevention Research Group
Table 5 Percentage of sample in each juvenile and adult arrest category, by site Intervention
Control
Normative
0
58.34
52.3
67.66
1–4
15.76
20.59
17.05
5–8
13.42
13.58
8.24
9+
12.48
13.54
7.06
0
66.25
54.06
68.89
1–4
12.55
19.31
15.87
5–8
14.68
14.54
11.95
9+
6.51
12.08
3.29
0
33.66
36.91
53.84
1–4
20.34
23.74
22.84
5–8
22.03
21.15
7.32
9+
23.97
18.21
16
0
67.73
55.23
71.73
1–4
14.91
24.34
20.26
5–8
10.56
9.96
5.36
9+
6.8
10.46
2.65
Juvenile arrest index Full sample
Durham
Nashville
Pennsylvania
Seattle 0
66.5
63.89
77.31
1–4
15.09
14.65
8.33
5–8
6.04
8.27
8.26
9+
12.37
13.19
6.1
0
67.87
68.61
76.49
1–3
13.93
13.68
12.66
4–6
9.44
8.30
6.46
7+
8.76
9.42
4.39
0
53.64
52.29
53.00
1–3
16.36
15.60
23.00
4–6
15.45
12.84
14.00
7+
14.55
19.27
10.00
0
73.68
75.86
81.00
1–3
6.14
8.62
7.00
4–6
10.53
10.34
7.00
Adult arrest index All
Durham
Nashville
Fast track intervention effects on youth arrests and delinquency
147
Table 5 (continued) Intervention
Control
Normative
9.65
5.17
5.00
0
76.99
83.93
95.92
1–3
13.27
8.93
3.06
4–6
5.31
2.68
1.02
7+
4.42
4.46
0.00
0
66.67
61.47
76.40
1–3
20.37
22.02
17.98
4–6
6.48
7.34
3.37
7+
6.48
9.17
2.25
7+ Pennsylvania
Seattle
category captured severe and violent offenders. Among youth in the 1–3 category, 77% were only arrested for severity 1 or 2 offenses. Among youth in the 4–6 category, the worst offense for which 42% of the youth were arrested was a severity 3 offense (28% were arrested for only severity 1–2 crimes and 30% were arrested for at least one severity 4–5 offense). Among youth in the highest category, the worst offense for which the majority (53%) were arrested was a severity 4/5 offense (only 5% were arrested for only severity 2 offenses and 42% were arrested for at least a severity 3 offense along with additional arrests). Again, the proportion of the sample in each of these categories by intervention status and site are displayed in Table 5. The three outcomes examined here were significantly correlated, but at relatively modest levels. The correlations between juvenile arrests with adult arrests, juvenile arrests with self-reported delinquency, and self-reported delinquency with adult arrests were .29, .32, and .25, respectively, all significant at the p<.0001 level. To assess the impact of intervention on these categorical variables capturing arrest activity, we wanted to use an ordered logit framework. The ordered logit approach was first introduced by McKelvey and Zavoina (1975). It is akin to the logit model which estimates the probability of dichotomous output. The ordered logit allows for an outcome with more than two ordinal categories and estimates the probability of being in a higher category relative to a lower category. This model assumes proportional odds—the impact of a covariate on the probability of being in the 1–4 juvenile arrest index category relative to the 0 category is the same as the impact of being the 9+ category relative to the 4–8 category. For the juvenile arrest index, we rejected this assumption for 12 of the 100 imputed data sets; however, the results do not change when those data sets were excluded from the model. Since the majority of the data sets fail to reject the assumption, we present the ordered logit results. For adult arrests, we rejected the proportional odds assumption for a majority of the data sets. Consequently, we used a stereotype logit model to assess the impact of intervention of adult arrest activity. The stereotype model was first introduced by Anderson (1984) in response to the limitations of the proportional odds assumption. As in the ordered logit, only one set of coefficients/slopes are estimated, which reflects the impact of
148
Conduct Problems Prevention Research Group
the covariate on the probability of the most serious arrest activity relative to no arrest activity. To estimate the impact of the covariates on the probability of different categories (i.e., minor arrest activity relative to no arrest activity), the estimated coefficients are multiplied by an adjustment factor also estimated by the model, that is, for each comparison, all the slopes are adjusted by the same percentage. In addition to these global measures of delinquent behavior, we examined an additional set of outcomes, the number of arrests and offenses by severity level. Specifically, we examined nine additional outcomes aggregated through grade 12: number of severity 1 and 2 juvenile arrests, number of severity 3 juvenile arrests, number of severity 4 and 5 juvenile arrests, number of severity 1 and 2 adult arrests, number of severity 3 adult arrests, number of severity 4 and 5 adult arrests, number of severity 1 and 2 self-reported offenses, number of severity 3 self-reported offenses, number of severity 4 and 5 self-reported offenses.1 These additional outcomes were analyzed within a negative binomial framework as they are counts. Given the non-linear nature of the negative binomial model, the estimates have been transformed into the percentage change in the expected number of arrests for intervention relative to control youth. The percentage change in the expected number of arrests is calculated using the following formula: exp(estimate)1 (Long and Freese 2001). For each outcome, the model included indicators for intervention, gender, race, site, cohort, 11 continuous pre-intervention covariates, whether the youth’s father or mother were ever arrested, and whether the youth was incarcerated for any part of the past year in grades 8 through 12. Standard errors were clustered by kindergarten school to account for the fact that randomization occurred at the school level. For each outcome, we assessed whether the intervention effect was moderated by gender, race, cohort, site, or initial severity of risk. All statistically significant moderated intervention effects are described in the Results section and a complete set of results is available upon request. Finally, we assessed the impact of intervention on the onset of any juvenile arrest, a severity 1 or 2 juvenile arrest, a severity 3 juvenile arrest, a severity 4 or 5 juvenile arrest, any adult arrest, a severity 1 or 2 adult arrest, a severity 3 adult arrest, a severity 4 or 5 adult arrest, any self-reported delinquent offense, a severity 1 or 2 self-reported offense, a severity 3 self-reported offense, and a severity 4 or 5 selfreported offense. To do so, we estimated discrete time survival analyses. These models included indicators for intervention, gender, race, site, cohort, 11 continuous pre-intervention covariates, whether the youth’s father or mother were ever arrested, and time and time squared. Standard errors were clustered by kindergarten school to account for the fact that randomization occurred at the school level. If the intervention effect was statistically significant, we then examined whether the intervention effect varied across time. In addition, we assessed whether the intervention effect was moderated by gender, race, cohort, site, or initial severity of risk. All statistically significant moderated intervention effects are described in the Results section.
1
We estimated a revised imputation model for these auxiliary analyses that included the number of arrests and offenses in each severity category and grade rather than the severity indices at each grade.
Fast track intervention effects on youth arrests and delinquency
149
3 Results 3.1 Impact of intervention on arrest activity As seen in Table 6, we found a statistically significant intervention effect on juvenile arrest activity based on the juvenile arrest index. Among intervention youth, the odds of being in a higher juvenile arrest activity group were only 71% of the odds for control youth (OR=0.71, p=0.05). Examining the number of juvenile arrests by severity level revealed that intervention decreased the expected number of severity 3 arrests by 24% (p=0.04) As seen in Table 3, the average number of severity 3 juvenile arrests among control youth was 0.54 through the end of high school, whereas the average for intervention youth was only 0.47. As seen in Table 6, the main effect of intervention on the probability of being in the most severe adult arrest category relative to having no adult arrests was not statistically significant. While the results in the table compare the probability of being in the most severe adult arrest category relative to never being arrested, we calculated the impact of intervention when comparing the probability of being in each arrest category relative to all other categories. The intervention did not have a statistically significant impact on any of these group comparisons. The auxiliary analysis of the number of adult arrests in the three severity groups showed no significant main effect of intervention. However, the interaction between intervention and screen score was negative and significant (p=0.04) indicating that intervention was associated with a lower expected number of severity 4/5 adult arrests among youth with higher initial risk levels. To better understand this interaction, we determined the largest sub-sample for which the intervention effect was significant. We accomplished this by selecting sub-samples based on the initial risk percentiles from the normative population. The intervention effect was
Table 6 Intervention effects on severity indices and number of arrests/offenses by severity type Severity
Number of arrests/offenses by severity (neg. binomial)b
Indexa
Severity 1/2
Severity 3
Change
Change
OR/Est
p-value
p-value
Juvenile arrests
0.71
0.05
–11%
0.34
Adult arrests
0.93
0.77
27%
0.10
Self-reported offenses
–1.56
0.78
–5%
0.58
–11%
Severity 4/5 p-value
Change
p-value
–24%
0.04
–10%
0.63
0%
0.99
–16%
0.47
0.32
–11%
0.24
a
Ordered logit estimated for the juvenile arrest index (OR reported), stereotype logit estimated for the adult arrest index (OR comparing the probability of being in the highest adult arrest category to lowest category/no arrests reported), and a standard linear regression estimated for self-reported offenses (regression estimate reported)
b
Percentage change in the expected number of arrests/offenses reported. Models also include 11 preintervention covariates; indicators for gender, race, site, and cohort; an indicator for whether mother has ever been arrested; an indicator for whether father has ever been arrested; and indicators for whether youth spent any time in jail during the past year for grades 8–12. The complete set of results is available on request
150
Conduct Problems Prevention Research Group
significant for the sub-sample with initial risk scores at or above the 13th percentile, corresponding to 52% of the intervention/control sample. Among those with initial risk levels at or above the 13th percentile of the normative population (denoted higher-risk youth), intervention decreased the expected number of severity 4/5 arrests by 47% (p=0.05). Among the higher-risk youth, the average number of severity 4/5 adult arrests for intervention youth was 0.09 (SD=0.04) while the control group average was 0.15 (SD=0.04). Among youth in the moderate-risk group (initial risk levels below the 13th percentile of the normative population), the impact of intervention was not statistically significant (p=0.33). We assessed the impact of intervention on the self-reported delinquency index within a standard linear regression model. As seen in Table 6, the main effect of intervention on self-reported delinquency index was not statistically significant nor was it significant for the number of self-reported offenses by severity level. 3.2 Impact of intervention on onset of arrest and delinquency Next, we used discrete survival analysis to estimate intervention effects on the onset of arrest and delinquent offenses. Survival analysis estimates the probability of an arrest/ offense in a given grade given no previous arrest/offense. This method controls for the data censoring that naturally occurs due to the fact that some youth have not been arrested by grade 12 but still may be arrested in the future. We employed a discrete time survival analysis model using logistic regression based on juvenile arrest information from grades 6 through 12, adult arrests from grades 9 through 12, and self-reported offenses from grades 7 through 12. We used the same set of imputed data sets described above. For each of these three data sources, we also examined the onset of different severity arrests/offenses: severity 1/2, severity 3, and severity 4/5. As seen in Table 7, intervention decreased the probability of any juvenile arrest among youth who were not previously arrested (OR=0.77, p<.05), thus indicating that intervention delayed the onset of arrests. No significant intervention effect was found for the onset of a severity 1/2, severity 3, or severity 4/5 juvenile arrest.
Table 7 Intervention effects on onset of arrest/offense based on discrete time survival analysis Any
Severity 1/2
Severity 3
Severity 4/5
OR
p-value
OR
p-value
OR
p-value
OR
p-value
Juvenile arrest
0.77
0.04
0.80
0.15
0.83
0.18
0.90
0.65
Adult arrest
0.98
0.89
1.18
0.28
0.80
0.26
0.85
0.46
Self-reported offense
0.98
0.84
1.03
0.81
0.91
0.37
0.82
0.05
Models also include time and time squared; 11 pre-intervention covariates; indicators for gender, race, site, and cohort; an indicator for whether mother has ever been arrested; an indicator for whether father has ever been arrested; and indicators for whether youth spent any time in jail during the past year for grades 8–12. The complete set of results is available on request
Fast track intervention effects on youth arrests and delinquency
151
The main effect of intervention was not significant for the onset of any adult arrest or the onset of severity-specific adult arrests. However, the interaction between site and intervention was statistically significant. Further analyses revealed an iatrogenic effect of intervention on the onset of a severity 4/5 adult arrest in Nashville. In Nashville, the odds of a severity 4/5 adult arrest among intervention youth with no previous 4/5 severity arrest were 1.98 that of control youth (p=0.03). An analysis of the other three sites individually revealed intervention effects in the expected direction but none were significant at the 0.05 level. When the remaining sites were analyzed together, a trend emerged, indicating that among intervention youth with no previous severity 4/5 adult arrest the odds of a severity 4/5 adult arrest were lower than control youth (OR=0.64, p=0.09). Although the intervention did not statistically significantly impact the onset of any self-reported delinquent offense, the intervention significantly decreased the probability of a severity 4/5 offense given no previous severity 4/5 offenses (OR=0.82, p=.05).
4 Discussion The current analyses of the long-term impact of the Fast Track intervention on youths’ delinquent behavior indicate that intervention influenced overall courtrecord juvenile arrest activity (as measured by the juvenile arrest index, the odds of intervention youth being involved in court-recorded arrests was only 71% of the rate of control youth), the frequency of moderate-severity juvenile arrests (the rate for intervention youth was 24% less than for control youth), and the onset of a juvenile arrest (the odds of intervention youth starting juvenile arrests was only 77% of the onset rate for control youth). There is also evidence that intervention delayed the onset of severe self-reported delinquent behavior (the odds of high-risk intervention youth having onset of self-reported delinquent behavior was only 82% of the onset rate for control youth). Intervention effects on the frequency of highseverity adult arrests (through age 19) were moderated by children’s baseline levels of problem behavior whereas the onset of high-severity adult arrests was moderated by site. Juvenile versus adult crimes The Fast Track intervention had a main effect on court records of the rates of moderately-severe juvenile arrests, but only with higher-risk youth on rates of severe arrests for adult offenses. The positive effects on juvenile arrest with the entire sample possibly reflect the impact of intervention while it was underway, and these effects may have dissipated for moderate-risk youth after intervention was completed in the 10th grade, resulting in less impact on adult arrest records for moderate-risk youth. However, we believe that the failure to obtain intervention effects across the full sample on court records of adult crimes may be partly due to the timing of our data collection for adult crimes. While we have juvenile crime data collected over a number of years of the youths’ lives, the adult crimes are primarily, with only a relatively few exceptions of juveniles being remanded to adult court, collected just during the last 2 years of data collection when youth were between 18 and 19. As a result, as can be seen in Table 3, there are fewer
152
Conduct Problems Prevention Research Group
adult arrests detected at this early point in adulthood. The small numbers of adult arrests make it difficult to sensitively test for intervention effects. It will be critical to determine whether intervention effects are apparent in future years as the participants move further into young adulthood. Another possible explanation for the relatively differential intervention effects on juvenile arrests is that the nature of juvenile and adult arrests differ in certain qualitative ways. In some jurisdictions, courts and police can be somewhat flexible in determining what qualifies as a juvenile arrest, but there is less latitude for adult arrests (Elrod and Ryder 2005). Thus, it might be that the intervention might be having its clearest effects on the somewhat more flexibly defined moderate-level juvenile arrests, and primarily during the time that the intervention remains actively in effect. For adult crimes, the intervention effect on court-ordered arrests appears stronger for the rate of more severe criminal behavior displayed by the half of the sample who had displayed higher risk at baseline. We should note that there was a significant interaction between site and the variable assessing onset of serious court-recorded adult arrests, with the intervention at one site having significant negative effects, and the other sites tending to have positive intervention effects. Similar patterns across sites are evident in nonsignificant ways across the other two outcomes of self-reported delinquency and court-record juvenile arrests. There tended to be effects of intervention at three of the sites—Durham, Pennsylvania, and Seattle—but an iatrogenic effect at Nashville. The iatrogenic effect in Nashville is likely driven by African-American males based on an examination of the means. Among African-American males in Nashville, 19% of the intervention youth were arrested for a severity 4/5 adult crime, whereas only 8% of similar control males were arrested. Among European males in Nashville, none of the intervention males were arrested for a severity 4/5 adult crime but 6% of similar control males were arrested. Neither control nor intervention females in Nashville were arrested for a severity 4/5 adult crime during the time period analyzed. Self-reported delinquency Despite the finding of intervention effects on court records of the rates of arrests, intervention did not influence youths’ reports of the rates of their delinquent behavior, although there was an intervention effect on the onset of severe delinquent behavior. There are several possible interpretations that can be made of this disparity in findings. One conclusion is that one source of data more accurately portrays the impact of the intervention than the other. Another is that each provides some truth about the situation and that the evidence for a prevention impact is mixed. If one is to choose between sources, self-report has the potential advantage of reflecting more delinquent activity than the arrest data, since not all delinquency comes to police attention and not all delinquents are apprehended. On the other hand, as was noted in the Introduction, police-arrest data are not subject to the potential bias in reporting that could be operating with youth self-reports, since police had no knowledge of the intervention and which youth were a part of it. However, courtrecorded arrest data have their own limitations. In contrast to the variation in intervention findings for the rates of self-report versus court-recorded arrests, intervention effects were evident for both sources of criminal behavior when the onset of criminal behaviors was examined. The Fast Track intervention delayed the onset of juvenile court-record arrests, and delayed the
Fast track intervention effects on youth arrests and delinquency
153
onset of the most severe forms of self-reported criminal behavior. These similar results across the two sources support the validity of these findings. Relation of the arrest and self-report outcomes to earlier Fast Track outcomes Earlier examinations of the outcomes of Fast Track in elementary school had found consistent modest positive effects of intervention for all four sites on children’s behavior and social skills through the fifth grade (Conduct Problems Prevention Research Group 1999, 2002a, 2004), but more limited effects during middle school (Conduct Problems Prevention Research Group, in press). Intervention youth reported lower engagement in self-reported delinquency in some years (7th and 9th grades) but not in other years (e.g., 8th grade). The present findings indicate that the previously obtained Fast Track intervention effects on youths’ rates of self-reported delinquency were maintained across this age period through the 12th grade for the onset of serious levels of delinquent behavior, but not for milder offenses. Although the reduction in Fast Track effects on milder offenses over time are not conclusively clear, it appears that developmental changes in the youth, and reductions in the intensity of the intervention over time, likely contribute to this pattern of findings. However, there is a positive intervention effect on overall juvenile court-recorded arrest activity as measured by the juvenile arrest index and the frequency of moderate severity juvenile arrests. Similarly, delayed onset of more severe selfreports of delinquent behavior indicates that certain aspects of intervention effects are evident in both sources of data. These results support the trend finding obtained with the Montreal Delinquency Program (MDP) (Boisjoli et al. 2007), and the significant effects of the Seattle Social Development Project on age 21 lifetime rates of court charges (Hawkins et al. 2005), of the Linking the Interest of Families and Teachers program on onset of police arrest (Eddy et al. 2003), and of a nurse home visitation program on arrests, convictions, and violations of probation of children of the subset of mothers who were unmarried and were from low-SES families (Olds et al. 1998). Fast Track is one of the relatively few prevention studies to demonstrate significant preventive effects on youth arrests, and is the only prevention program targeted on high-risk children, other than the MDP program, to have an effect on arrests. While these effects are modest, because of the large costs that accrue due to juvenile antisocial behavior (e.g., Foster et al. 2005), Fast Track appears to be costeffective in reducing mental health service utilization (Jones et al. in press) and for the subgroup of youth who were at highest risk in initial screening (Foster et al. 2006), and future cost-effectiveness research can determine if this is apparent for arrest outcomes for highest-risk youths as well. It will be critically important to determine whether intervention has any longer-term effects on adult arrests as the sample ages into young adulthood (see Appendix).
Disclosure Drs. Bierman, Coie, Dodge, Greenberg, Lochman, and McMahon are the developers of the Fast Track curriculum and have a publishing agreement with Oxford University Press. Dr. Greenberg is an author on the PATHS curriculum and has a royalty agreement with Channing-Bete, Inc. Dr. Greenberg is a principal in PATHS Training, LLC. Dr. McMahon is a coauthor of Helping the Noncompliant Child and has a royalty agreement with Guilford Publications, Inc.; he is also a member of the Treatments That Work Scientific Advisory Board with Oxford University Press. The other authors have no financial relationships to disclose.
154
Conduct Problems Prevention Research Group
Appendix 9,594 Assessed for Eligibility at School 6,320 Excluded 5,994 Not Meeting Inclusion Criteria 326 Refused to Participate 3,274 Assessed for Eligibility at Home 2,383 Excluded 2,249 Not Meeting Inclusion Criteria 75 Refused to Participate 59 Did Not Matriculate to 1st Grade 891 Randomized 445 Assigned to Receive Intervention 445 Received Intervention as Assigned
446 Assigned to Control Sample
137 Parents Lost to Follow-up in Grade 12 57 Explicit Refusal 3 Passive Refusal 2 Moved Out of Fast Track Area 30 Contact Information Unavailable 12 Failure in Contact Attempts 2 Physically/Mentally Dsabled 3 Deceased 1 Interviewer Error 18 Child is Living Independently 8 In State Custody 1 Child ran away
157 Parents Lost to Follow-up in Grade 12 54 Explicit Refusal 6 Passive Refusal 2 Moved Out of Fast Track Area 46 Contact Information Unavailable 21 Failure in Contact Attempts 2 Other 2 Deceased 15 Child is Living Independently 9 In State Custody
111 Youths Lost to Follow-up in Grade 12 44 Explicit Refusal 4 Passive Refusal 3 Moved Out of Fast Track Area 33 Contact Information Unavailable 10 Failure in Contact Attempts 2 Physical/Mental Disability 7 Family dropped out 5 Deceased 1 Equipment Malfunction 1 In State Custody 1 Run away
130 Youths Lost to Follow-up in Grade 12 44 Explicit Refusal 5 Passive Refusal 4 Moved Out of Fast Track Area 47 Contact Information Unavailable 22 Failure in Contact Attempts 6 Family dropped out 1 Deceased 1 Child is Living Independently
49 Youth Missing Juvenile Records 49 Youth Refused Consent
72 Youth Missing Juvenile Records 72 Youth Refused Consent
445 Included in Analysis - using Multiple Imputation 0 Excluded from Analysis
446 Included in Analysis - using Multiple Imputation 0 Excluded from Analysis
Consort flow-chart
References Achenbach, T. M. (1991a). Manual for the child behavior checklist and 1991 profile. Burlington: University of Vermont Department of Psychiatry. Achenbach, T. M. (1991b). Manual for the Teacher’s Report Form and 1991 Profile. Burlington: University of Vermont Department of Psychiatry. Anderson, J. A. (1984). Regression and ordered categorical variables. Journal of the Royal Statistical Society Series B, 46, 1–30. Bierman, K. L., Greenberg, M. T., & the Conduct Problems Prevention Research Group. (1996). Social skills training in the Fast Track program. In R. DeV Peters & R. J. McMahon (Eds.), Preventing childhood disorders, substance abuse, and delinquency (pp. 65–89). Thousand Oaks: Sage. Boisjoli, R., Vitaro, F., Lacourse, E., Barker, E. D., & Tremblay, R. E. (2007). Impact and clinical significance of a preventive intervention for disruptive boys: 15-year follow-up. British Journal of Psychiatry, 191, 415–419. Brandt, D. E. (2006). Delinquency, development, and social policy. New Haven: Yale University Press.
Fast track intervention effects on youth arrests and delinquency
155
Butts, J. A., & Snyder, H. N. (2007). Where are juvenile crime trends headed? Juvenile and Family Justice Today, 16–20. Cairns, R. B., Cairns, B. D., Neckerman, H. J., Gest, S. D., & Gariepy, J. L. (1988). Social networks and aggressive behavior: Peer support or peer rejection? Developmental Psychology, 24, 815–823. Cernkovich, S. A., & Giordano, P. C. (2001). Stability and change in antisocial behaviour: The transition from adolescence to early adulthood. Criminology, 39, 371–410. Conduct Problems Prevention Research Group. (1992). A developmental and clinical model for the prevention of conduct disorder: The FAST Track program. Development and Psychopathology, 4, 509–527. Conduct Problems Prevention Research Group. (1999). Initial impact of the Fast Track prevention trial for conduct problems: I. The high-risk sample. Journal of Consulting and Clinical Psychology, 67, 631–647. Conduct Problems Prevention Research Group. (2002a). Evaluation of the first three years of the Fast Track prevention trial with children at high risk for adolescent conduct problems. Journal of Abnormal Child Psychology, 30, 19–35. Conduct Problems Prevention Research Group. (2002b). The implementation of the Fast Track program: An example of a large-scale prevention science efficacy trial. Journal of Abnormal Child Psychology, 30, 1–17. Conduct Problems Prevention Research Group. (2004). The effects of the Fast Track program on serious problem outcomes at the end of elementary school. Journal of Clinical Child and Adolescent Psychology, 33, 650–661. Conduct Problems Prevention Research Group. (2007). The Fast Track randomized controlled trial to prevent externalizing psychiatric disorders. Journal of the American Academy of Child and Adolescent Psychiatry, 46, 319–333. Conduct Problems Prevention Research Group. (in press). The effects of the Fast Track program on disruptive behavior, deviant peer relations, and social skills during the middle school years. The Journal of Early Adolescence. Dishion, T. J., Patterson, G. R., & Griesler, P. C. (1994). Peer adaptations in the development of antisocial behaviour: A confluence model. In L. R. Huesmann (Ed.), Aggressive behaviour: Current perspectives (pp. 61–95). New York: Plenum Press. Dodge, K. A. (1993). Home visiting manual for the Fast Track program. Nashville: Unpublished. Dodge, K. A., Dishion, T. J., & Lansford, J. E. (2006). Deviant peer influences in programs for youth: Problems and solutions. New York: Guilford Press. Eddy, J. M., Reid, J. B., Stoolmiller, M., & Fetrow, R. A. (2003). Outcomes during middle school for an elementary school-based preventive intervention for conduct problems: Follow-up results from a randomized trial. Behavior Therapy, 34, 535–552. Elliott, D. S., Huizinga, D., & Ageton, S. S. (1985). Explaining delinquency and drug use. Beverly Hills: Sage. Elrod, P., & Ryder, R. S. (2005). Juvenile justice: A social, historical, and legal perspective (2nd ed.). Boston: Jones and Bartlett. Farrington, D., Loeber, R., Stouthamer-Loeber, M., Van Kammen, W. B., & Schmidt, L. (1996). Selfreported delinquency and a combined delinquency seriousness scale based on boys, mothers, and teachers: concurrent and predictive validity for African-Americans and Caucasians. Criminology, 34, 493–517. Farrington, D. P., Jolliffe, D., Loeber, R., & Homish, D. L. (2007). How many offenses are really committed per juvenile court offender? Victims & Offenders, 2, 227–249. Foster, E. M., Jones, D., & Conduct Problems Prevention Research Group. (2005). The high costs of aggression: Public expenditures resulting from conduct disorder. American Journal of Public Health, 95, 1767–1772. Foster, E. M., Jones, D., & Conduct Problems Prevention Research Group. (2006). Can a costly intervention be cost-effective? An analysis of violence prevention. Archives of General Psychiatry, 63, 1284–1291. Hawkins, J. D., Kosterman, R., Catalano, R. F., Hill, K. G., & Abbott, R. D. (2005). Promoting positive adult functioning through social development intervention in childhood. Archives of Pediatrics and Adolescent Medicine, 159, 25–31. Hill, L. G., Lochman, J. E., Coie, J. D., Greenberg, M. T., & Conduct Problems Prevention Research Group. (2004). Effectiveness of early screening for externalizing problems: Issues of screening accuracy and utility. Journal of Consulting and Clinical Psychology, 72, 809–820. Hollingshead, A. A. (1975). Four-factor index of social status. Unpublished manuscript. Yale University: New Haven.
156
Conduct Problems Prevention Research Group
Jones, D. E., Godwin, J., Dodge, K. A., Bierman, K. L., Coie, J. D., Greenberg, M.T. et al. (in press). The impact of the fast track prevention trial on health services utilization by youth at risk for conduct problems. Pediatrics. Kirk, D. S. (2006). Examining the divergence across self-report and official data sources on inferences about the adolescent life-course of crime. Journal of Quantitative Criminology, 22, 107–129. Kusche, C. A., & Greenberg, M. T. (1993). The PATHS (Promoting alternative thinking strategies) Curriculum. Deerfield: Channing-Bete Co. Lahey, B. B., Loeber, R., Burke, J. D., & Applegate, B. (2005). Predicting future antisocial personality disorder in males from a clinical assessment in childhood. Journal of Consulting and Clinical Psychology, 73, 389–399. Lochman, J. E., & the Conduct Problems Prevention Research Group. (1995). Screening of child behavior problems for prevention programs at school entry. Journal of Consulting and Clinical Psychology, 63, 549–559. Lochman, J. E., & Wells, K. C. (2003). Effectiveness study of Coping Power and classroom intervention with aggressive children: outcomes at a one-year follow up. Behavior Therapy, 34, 493–515. Lochman, J. E., & Wells, K. C. (2004). The coping power program for preadolescent aggressive boys and their parents: Outcome effects at the one-year follow-up. Journal of Consulting and Clinical Psychology, 72, 571–578. Long, J. S., & Freese, J. (2001). Regression models for categorical and limited dependent variables with Stata. College Station: Stata Press. McKelvey, R. D., & Zavoina, W. (1975). A statistical model for the analysis of ordinal level dependent variables. Journal of Mathematical Sociology, 4, 103–120. McMahon, R. J., Slough, N., & the Conduct Problems Prevention Research Group. (1996). Family-based intervention in the Fast Track Program. In R. DeV Peters & R. J. McMahon (Eds.), Preventing childhood disorders, substance use, and delinquency (pp. 90–110). Thousand Oaks: Sage. Olds, D., Henderson, C. R., Cole, R., Eckenrode, J., Kitzman, H., Luckey, D., et al. (1998). Long-term effects of nurse home visitation on children's criminal and antisocial behavior: 15-year follow-up of a randomized controlled trial. Journal of the American Medical Association, 280, 1238–1244. Oyserman, D. (2000). School-to-Jobs facilitators manual: After-school program. Ann Arbor: University of Michigan. Pellegrini, A. D. (2003). Perceptions and functions of play and real fighting in early adolescence. Child Development, 74, 1522–1533. Reinherz, H. Z., Tanner, J. L., Berger, S. R., Beardslee, W. R., & Fitzmaurice, G. M. (2006). Adolescent suicidal ideation as predictive of psychopathology, suicidal behavior and compromised functioning at age 30. American Journal of Psychiatry, 163, 1226–1232. Schafer, J. L., & Graham, J. W. (2002). Missing data: Our view of the state of the art. Psychological Methods, 7, 147–177. Thornberry, T. P., & Krohn, M. D. (2001). The development of delinquency: An interactional perspective. In S. O. White (Ed.), Handbook of youth and justice (pp. 289–305). Dordrecht: Kluwer. Vitaro, F., Brendgen, M., Pagani, L., Tremblay, R. E., & McDuff, P. (1999). Disruptive behavior, peer association, and conduct disorder: Testing the developmental links through early intervention. Development and Psychopathology, 11, 287–304. Werthamer-Larsson, L., Kellam, S. G., & Wheeler, L. (1991). Effects of first grade classroom environment on shy behavior, aggressive behavior, and concentration problems. American Journal of Community Psychology, 19, 585–602. West, D. J., & Farrington, D. P. (1973). Who becomes delinquent? Second report of the Cambridge Study in delinquent development. Oxford: Crane. Karen L. Bierman, Ph.D. is a distinguished professor of child-clinical psychology and director of the Child Study Center at The Pennsylvania State University. Her research program focuses on socialemotional development and children at risk, with an emphasis on the design and evaluation of school- and community-based programs that promote social competence, school readiness, positive intergroup relations, and that reduce aggression and violence. Dr. Bierman served as the principal investigator for the Pennsylvania site of the Fast Track Program. John D. Coie, Ph.D. is Professor Emeritus in Psychology and Neuroscience at Duke University. He has studied the development of antisocial behavior and children’s relations with peers. He continues to be interested in prevention of antisocial behavior and disorder and the promotion of children’s competence in community settings.
Fast track intervention effects on youth arrests and delinquency
157
Kenneth A. Dodge, Ph.D. is the William McDougall Professor of Public Policy and Director of the Center for Child and Family Policy at Duke University. He studies the development and prevention of violence in children and families. Mark T. Greenberg, Ph.D. is the Bennett Chair of Prevention Research and Director of the Prevention Research Center at Pennsylvania State University. He studies developmental processes and interventions focused on improving the competence of children and families. John E. Lochman, Ph.D., ABPP is Professor and Doddridge Saxon Chairholder in Clinical Psychology at The University of Alabama, and he directs the Center for Prevention of Youth Behavior Problems. His research examines social and social-cognitive risk factors related to children’s aggressive behavior, examines intervention efficacy, and examines variables that affect the dissemination of interventions in real-world settings. Robert J. McMahon, Ph.D. is Professor and Director of the Child Clinical Psychology Program in the Department of Psychology at the University of Washington. His primary research and clinical interests concern the development, assessment, treatment, and prevention of conduct problems and other problem behavior in youth, especially in the context of the family. Dr. McMahon served as the principal investigator for the Seattle site of the Fast Track Program. Ellen E. Pinderhughes, Ph.D. is Associate Professor in the Eliot-Pearson Department of Child Development at Tufts University. Her research examines family socialization processes among children and youth at-risk for problems in their development.