تاثیر یک برنامه پیشگیری از خشونت جهانی مبتنی بر مدرسه در بزهکاری خشونت آمیز: مزایای متمایز برای جوانان با سابقه بدرفتاری

Abstract Objective Child maltreatment constitutes a strong risk factor for violent delinquency in adolescence, with cumulative experiences of maltreatment creating increasingly greater risk. Our previous work demonstrated that a universal school-based violence prevention program could provide a protective impact for youth at risk for violent delinquency due to child maltreatment history. In this study we conducted a follow-up to determine if participation in a school-based violence prevention program in grade 9 continued to provide a buffering effect on engaging in acts of violent delinquency for maltreated youth, 2 years post-intervention.

Introduction Child maltreatment interferes with healthy development and contributes to a range of negative psychological and physical health outcomes (Kim, Cicchetti, Rogosch, & Manly, 2009). Children who experience abuse are more likely to be diagnosed with a psychological disorder and to show difficulties with early attachment, emotional regulation, peer relationships, school adjustment, and pro-social behaviors (Cyr et al., 2010 and Wekerle and Wolfe, 2003). Child maltreatment history raises the risk of numerous health-risk behaviors such as smoking, substance use, and early and promiscuous sexuality (Dube et al., 2006, Edwards et al., 2007 and Hussey et al., 2006), and substantially increases risk for delinquent and violent behaviors (Lansford et al., 2007 and Maas et al., 2008). The relationship between child maltreatment and delinquency suggests that severe or cumulative experiences of child maltreatment are particularly detrimental with respect to future conduct (Maas et al., 2008 and Margolin et al., 2009). It is clear that experiences of child abuse and maltreatment show a dosage effect. That is, a greater number of abuse experiences and types exert an increasingly more detrimental impact. The Adverse Childhood Experiences Study (Felitti et al., 1998) utilized a sample of more than 8,500 adults accessed through a large HMO and found a dosage–response relationship between experiences of childhood maltreatment and other adverse effects (such as parental incarceration, mental health problems, and/or substance abuse, parental separation or divorce), and a wide range of physical and mental health outcomes including heart disease, cancer, and liver disease. In some areas of health, experiences of abuse eclipse associations previously attributed to certain causes. For example, the negative psychosocial sequelae and fetal deaths commonly attributed to adolescent pregnancy seem to result from underlying adverse childhood experiences rather than adolescent pregnancy per se (Hillis et al., 2004). The mechanisms by which adverse childhood experiences lead to poor pregnancy and birth outcomes remains to be clarified, and may include mechanisms such as low self-efficacy, high stress, and difficulties with anger, all of which could interfere with accessing good prenatal care. Beyond physical health outcomes, multiple forms of abuse had a cumulative effect in predicting anger and hostility among a sample of 554 youth (Teicher, Samson, Polcari, & McGreenery, 2006). Poly-victimization has been shown to predict higher rates of psychiatric disorders in a nationally representative sample of more than 2,000 youth in the United States aged 2–17 (Finkelhor, Ormrod, & Turner, 2007), as well as higher rates of delinquency, independent of diagnoses of depressive, posttraumatic, and substance use disorders (Ford, Elhai, Connor, & Frueh, 2010). At the extreme end of the dosage continuum, some researchers and policy makers have contended that severe, early, chronic experiences of abuse and neglect may lead to the profound damage in psychological development that, in turn, sets the context for the perpetration of severe and lethal violence against others (Garbarino, 1999 and Karr-Morse and Wiley, 1997). Child maltreatment extends well beyond the cases known to child protection agencies and 21.9% of youth in aforementioned nationally representative sample have experienced some form of child maltreatment and/or witnessed domestic violence (Finkelhor, Ormrod, & Turner, 2009). Accordingly, universal prevention has been touted as an essential yet under-utilized component of a comprehensive intervention and prevention strategy (Chaffin and Schmidt, 2006 and Hammond et al., 2009). In addition to universal prevention initiatives provided to all individuals regardless of risk level, selected and indicated prevention are also important for maltreated children who may require specialized programming (Wolfe et al., 2003). The Fourth R is a universal approach to the prevention that focuses on the underlying relationship dynamics common to dating violence, unsafe sex, substance use, and peer violence (Wolfe, Jaffe, & Crooks, 2006; see www.youthrelationships.org for more program information). The cornerstone of the program is a 21-session classroom-based curriculum that emphasizes knowledge, awareness, and skill development pertaining to developmentally relevant issues of personal safety in relationships, sexual health, and substance use. The focus on skill development through role play exercises that break complex skills down into manageable segments is a particularly unique aspect. As part of a comprehensive, universal prevention strategy, the lessons are taught by trained classroom teachers and fulfill state/provincial curriculum requirements. Other components include a Youth Action Committee that conducts school-wide awareness campaigns and prevention activities, extensive teacher training, and parent information. The Fourth R has been demonstrated to provide a protective effect on delinquency for maltreated youth at post-test. Using data from our cluster randomized controlled trial of the Fourth R with students from 20 schools, we found student perception of safety (aggregated at the school level) predicted lower rates of delinquency over and above all individual-level predictors at the end of grade 9 (Crooks, Scott, Wolfe, Chiodo, & Killip, 2007). Furthermore, being a Fourth R intervention school buffered the impact of child maltreatment history on the development of violent delinquency. This school-level moderating effect among youth with maltreatment backgrounds was unexpected because schools had implemented the prevention program for only 1 year. In addition to this benefit for youth with maltreatment histories, the Fourth R conferred positive benefits on behavior among the whole sample. At 2 year follow-up, the Fourth R was found to significantly reduce dating violence and increase condom use among students, relative to peers attending control schools who received regular health class (Wolfe, Crooks, Chiodo, & Jaffe, 2009). Reductions in substance use were not found among the full sample. In their call for research identifying potentially protective contexts, Foster and Brooks-Gunn (2009) identified our previous study of predictors of violent delinquency as one of the few examples to date to investigate school level effects as a moderator of the impact of child maltreatment. Due to this scarcity of research identifying contextual influences on maltreated youth, the question of duration of the protective impact on the subsample of youth with self-identified histories of child maltreatment is paramount. Therefore, the purpose of this study is to examine the stability of these effects 2 years later. That is, does dosage of child maltreatment have the same exponential relationship to perpetration of violent delinquency for youth by the end of grade 11 as it did at the end of grade 9? Also, does participation in a comprehensive violence prevention program in grade 9 continue to provide a buffering effect for maltreated youth at the end of grade 11?

Results On average, 9.6% of grade 11 students across our sample of 20 schools were classified as having engaged in violent delinquency; however, rates varied considerably by school (ranging from 2% to 32%). In the 3 least delinquent schools, 5% or fewer grade 11 students were classified as engaging in violent delinquent acts (2%, 4%, 5%). In contrast, the proportion of violent delinquent students at 3 schools exceeded 16%, reaching a rate of 32% in 1 school. We began our analysis of the effect of individual and school-level variance on acts of violent delinquency by estimating an unconditional model within HLM. This model predicts the outcome within each level 1 unit (i.e., student) with just 1 level 2 parameter—the intercept (in this case, the adjusted average engagement in violence delinquency across schools). It also allows for partitioning variance into within- and between-school components. Results indicated an adjusted average engagement in violent delinquency of 9.45 across schools, 1/[1 + exp(2.26)]. The estimated variance of the intercept term was modest (.04), but indicated that school-level variation in rates of violent delinquency was worth modeling; χ2(19) = 23.91, p = .20. The next step of the analysis added student-level predictors (i.e., childhood maltreatment, grade 9 delinquency, gender) to the model. Effects for all variables except childhood history of maltreatment were fixed (i.e., constrained to be the same across all schools) and centered around the grand mean. The slope for childhood maltreatment was left random to allow examination of between-school variation in the association between childhood maltreatment and violent delinquency and to investigate school-level predictors of such variation. Results showed that sex of student, time 1 delinquency, and maltreatment history predicted individual variation in the chances of engaging in violent delinquency (details reported in final model). With these individual-level predictors, the residual variance associated with the intercept was .07; χ2(19, n = 1,450) = 26.02, p = .13. Residual variance associated with maltreatment was modest at .12, χ2(19) = 28.15, p = .08 indicating that additional modeling was warranted. In an attempt to replicate our previous findings at follow-up, a level 2 model was specified with the 3 individual-level predictors (gender, grade 9 delinquency, and childhood history of maltreatment) and with 2 school-level predictors (perceived safety, intervention status) added to both the intercept and to the freed child maltreatment term (see Table 1). This allowed us to estimate both school-level predictors of violent delinquency (adjusting for individual-level characteristics) and to determine whether the relationship between childhood maltreatment and subsequent violent delinquency varied depending on the characteristics of the school in which youth were enrolled (again, adjusting for individual-level characteristics). Due to the relatively low variance associated with maltreatment between schools, models were estimated with and without freeing this parameter. Because no significant differences were found between models the fixed model was retained [deviance = 275.01, χ2(2) = 2.64, ns]. Results showed significant effects for both individual and school-level variables. At level 1 and controlling for all other factors, boys were 4.67 times more likely to engage in violent delinquency than were girls. Table 1. Individual, school and cross-level predictors of violent delinquency. Variable Coefficient (B) Standard error (SE B) Odds ratio (95% CI) t-ratio Intercept −2.81 0.13 .06 (.04–.08) −20.25** Level 1: Model of Student predictors Gender 1.54 0.25 4.67 (2.85–7.65) 6.12** Maltreatment 0.38 0.10 1.46 (1.21–1.78) 3.86** Violent delinquency (grade 9) 0.54 0.07 1.72 (1.49–1.99) 7.46** Level 2: Model of School predictors of intercept Intervention Condition −0.09 0.22 .91 (.56–1.48) −0.40 Safety −0.36 0.11 .70 (.56–.87) −3.39** Cross-level interaction Condition × Maltreatment −0.41 0.19 .66 (.46–.97) −2.11* df = 1,429, 19. * p < .05. ** p < .01. Table options Childhood maltreatment was also significant, with each additional type of maltreatment increasing the relative odds of engagement in violent delinquency by 46. Unsurprisingly, violent delinquency in grade 9 predicted violent delinquency in grade 11. Odds of violent delinquency in grade 11 were 72% higher for youth who had reported such acts in grade 9, holding all other variables constant. There were 2 significant effects at the school level. First, rates of violent delinquency varied according to school-level perceived safety. Specifically, each standard deviation increase in perceived safety was associated with a 30% decrease in the odds of perpetrating violent delinquency, holding all other variables constant. In other words, Given the same individual profiles or risk, students were more likely to engage in violent delinquency if they were attending a school perceived by the entire student body as an “unsafe” climate than if they were attending a school perceived as a “safe” climate. Second, a cross-level interaction was found between childhood maltreatment and intervention condition. As shown in Fig. 1, the risk of violent delinquency associated with cumulative types of child maltreatment was lower in intervention schools than in control schools. Solving for the equations, we see that for a child in the control group, each additional form of child maltreatment yields a 46% higher chance of engaging in violent delinquency [exp(.38 + 0)]. In comparison, the effect of an increase in forms of child maltreatment on violent delinquency for a child in an intervention school is negligible delinquency (reduce risk by 3%); [exp(.38 − .41)]. With these predictors in the model, the amount of variation (.0007) was reduced by 99% to a non-significant level [χ2(17) = 10.36, p > .50], indicating that no further modeling was warranted. Cross level interaction of child maltreatment and intervention condition on ... Fig. 1. Cross level interaction of child maltreatment and intervention condition on probability of violent delinquency (two years post-intervention).