Coping Power

Study 1

Evaluation Methodology

Design:
Screening for Indicated Intervention: Fourth grade teachers in 17 elementary schools were asked to rate all the children in their classes on a scale of one to five on three items indicating aggressive behavior: 1) verbal aggression, 2) physical aggression, and 3) disruptive behavior. Based on these ratings, the 31% most aggressive male and female children in all classes were identified, producing 473 children in the pool of potential risk subjects who were then eligible for random assignment to indicated intervention and control conditions. Primary caretakers and children were contacted to gather their informed consent to participate in the study. Initial contacts to gain consent were done over the telephone and in home visits. Consent to participate in the study was obtained from the parents of 65% of the children in the pool, with an actual sample size of 245 moderate to high risk indicated subjects. After these subjects were identified, the baseline pre-assessment (Time 1) began, prior to the intervention period.

Universal Intervention Sample: After a summer baseline assessment and after students had been assigned to fifth grade classrooms, the 60 fifth grade classrooms in the 17 elementary schools were randomly assigned to either Universal Intervention (UI) or the Universal Comparison (UC) condition, with each of the 17 schools having at least one UI and UC classroom to control for school effects. Of the 1540 children who were in the fourth grade screening pool, 672 (43.6%) of the UI parents and caretakers who had children on 31 universal classroom rosters were tracked and invited to the UI parent meetings in grade 5. The Coping with Middle School Transitions (CMST) Parent Meetings consisted of 3 sessions scheduled during the fifth grade year, and one session during the sixth grade year. All of the parent/caretakers of the children in the Universal classrooms were invited to attend each of these meetings. The Teacher Meetings consisted of 5 two-hour meetings during the fifth-grade year, with 6 teachers meeting together with a Coping Power staff member.

Indicated Intervention Sample: Within the UI and UC classrooms, the indicated children (approximately 4 per classroom) were further randomly assigned to Indicated Intervention (II) or Indicated Control (IC) cells. Thus, for indicated target analyses there were four cells: Universal Intervention classroom and Indicated Intervention (UI) (N = 61), Universal Intervention classroom and Indicated Control (U) (N = 62), Universal Comparison classroom and Indicated Intervention (I) (N = 59), and Universal Comparison Classroom and Indicated Control (C) (N = 63).

Following condition assignments, the Coping Power indicated and universal interventions began. The Coping Power Child Component was a 16-month program with 22 group sessions scheduled for the fifth grade year, and 12 group sessions for the sixth grade year. Group sessions lasted for 40-50 minutes and consisted of 5-8 children who were co-led by a grant-funded staff school-family program specialist and by a school guidance counselor. Each child also received individual ½ hour sessions at their school, on the average of once every two months. The Coping Power Parent Component consisted of 16 sessions, over the same 16 month intervention period. Eleven parent group sessions were scheduled in the fifth grade year and five sessions were scheduled in the sixth grade year. This intervention was delivered in groups of at least 12 parents or parent dyads with two co-leaders. In most cases only the mother attended, despite efforts to include both parents in the meetings.

Time 2 mid-assessments were collected in the summer between the fifth and sixth grade years and indicated what changes had occurred over the intervention to that point (Time 1 assessments were collected pre-implementation). Time 3 assessments indicated post-intervention effects. Assessment measures were collected from primary caretakers either in their home or in the research offices. Children's "pre" measures were collected in their school settings, at home, or in the research offices. Dependent measures were individually administered to primary caretakers and children by grant staff who were blind to children's condition assignment. Intervention began in the Fall of the fifth grade year and continued through January of the sixth grade year.

Sample Characteristics: The overall sample was comprised of boys-to-girls in a 2-to-1 ratio, and this distribution was very similar across cells (UI: 68% male, I: 66%, U: 63%, C: 68%). The ethnic distribution in the sample primarily consisted of African-American and Caucasian children, with only two Hispanic children and four of other nationalities; the distribution of African-American children across cells was similar (UI: 75% African American, I: 78%, U: 78%, C: 81%). Analyses indicated that the four cells were equivalent in terms of aggressive behavior at screening, demographic factors, and teacher estimated cognitive functioning.

Measures: The following five sets of measures were used in the study:

Substance use: The 14-item child-report CSAP (Center for Substance Abuse Prevention) Student Survey adapted from the California Student Survey was used to measure student's attitudes toward and use of alcohol, tobacco and other drugs. Substance use within the past month was coded as "yes" or "no," and the self-reported information was summed and logged to create a Substance Use score.

Children's self-regulation: The six-item, two subscale Proactive-Reactive Aggressive Behavior Scale developed by Dodge and Coie was used as a measure of self-regulation and aggressive behavior. The scale was completed separately by parents at all three time points during the summer interviews and by teachers at Times 2 and 3 in the end-of-school assessments. Teacher-rated Behavioral Improvement data was also collected in the end-of-school-year assessment at Times 2 and 3 using a 7-point scale. The 16-item Teacher Observation of Classroom Adaption-Revised (TOCA-R) was completed by teachers in order to further measure aggression in all children in the classroom.

In order to measure the internal/behavioral processes involved with self-regulation, a version of the Early Adolescent Temperament Measure with self-reported subscales for fear and activity level was utilized. An abbreviated version of the Dysregulation Inventory was also completed by the child participants in order to measure cognitive dysregulation, affective dysregulation, and behavioral dysregulation. A total Dysregulation score was established by summing the three subscales listed above.

Children's social competence: Social behavior in terms of social competence was measured using perceived improvement on the seven items of the Teacher Rating of Children's Social Skills at Times 2 and 3. The three-item Teacher-rated Social Relations Problems subscale of the TOCA-R measured teachers' perceptions of peers' dislike or lack of acceptance of the target child, and subsequent unfriendliness. A perceived social competence score was created by combining the peer and athletic competence scales of the Perceived Competence Scale for Children (PCSC), a self-report measure with 36 items. The perceived academic competence score on the PCSC was used to index school bonding.

Social competence in terms of social cognition was measured using the Attributional Measure adapted from the Lochman and Dodge attribution measure. This measure assessed children's Attributions and Anger to four vignettes of peer provocations and conflicts with authority figures. In the end-of-school-year assessments at Times 2 and 3, teachers rated children's improvement in problem-solving and anger management skills using a seven-point scale ranging from worsening of these skills to great improvement. Social competence was further measured using the Outcome Expectation Questionnaire and the Perceived Peer Behavioral Norms scale.

School bonding: School bonding, academic competence, and parental involvement in the schools were measured by the Kentucky School Bonding measure, the academic subscale from the PCSC, and a 26-item parent-report adapted from the Fast-Track project (Conduct Problems Prevention Research Group), respectively.

Parental supportiveness and parenting practices: The Alabama Parenting Questionnaire (APQ) provided a subscale assessing Harsh Punishment for the parenting practices measure of the study. The Family Relations Scale (FRS) and the Lack of Support Scale were used to examine family relationship characteristics and parental warmth and positive involvement with their children.

Analysis: The analysis plan for intervention effects involved analysis on the substance use outcome variable, and on six sets of variables within the four predictor domains. The six sets of variables included: (1) as part of the self-regulation domain, aggressive behavior and internal or other behavioral processes assessed at Times 1 to 3; (2) also as part of the self-regulations domain, aggressive behavior assessed only at Times 2-3; (3) as part of the social competence domain, social behavior and social cognition assessed at Times 1-3; (4) as part of the social competence domain, social behavior assessed only at Times 2-3; (5) the school bonding domain; and (6) the parenting domain.

Repeated measures ANOVAs were conducted on those domains with single variables and repeated measures MANOVAs were conducted on each of the dependent variables within those domains with multiple variables. When significant or trend effects were obtained, then univariate repeated measures ANOVAs (with variables that had been assessed at the three time points) or univariate analyses (when data was only available for Times 2 and 3) were conducted. An ANCOVA was conducted on the only variable with baseline condition differences (Lack of Parental Supportiveness). Parents in the combined Indicated-Universal group had higher baseline levels of non-supportiveness than the other three groups.

The intervention effects for individual variables that had been part of significant or trend multivariate analyses of variance effects were tested with repeated measures ANCOVAs. The between subject independent variables were intervention status (Indicated or Universal), and the Indicated X Universal (I X U) interaction effect. The within subject independent variable was Time (Baseline, Mid-intervention, Post-intervention). When a significant or trend I X U interaction effect was obtained, two planned contrasts were conducted: the first tested whether the combined intervention cells (I, U, IU) were different from the Control (C) cell, and the second tested whether the combined intervention cell (IU) was different from the other three cells.

Outcomes

Posttest (Lochman & Wells, 2002a): The summer interview battery (child self-report and parent report) was administered to 214 participants (88% of the original sample) across all three time-periods. Teacher measures collected at both mid and post-intervention were available on 187 participants (77% of the original sample), and teacher measures collected as part of a full classroom assessment across the three time points were available on 125 participants (51% of the original sample). There were no significant differences between the attrited and non-attrited participants on the dependent variables of aggression screening score, teachers' estimate of children's cognitive abilities at screening, gender, and race for the summer interview battery, the teacher ratings in years two and three, or the teacher ratings of all students in their classrooms.

Baseline Equivalence: Equivalence analyses were conducted for each dependent measure at baseline using only the subjects who had available data across all three time points. Of the 23 dependent variables, there were significant baseline differences between conditions for only one variable, Lack of Parental Supportiveness. A significant Indicated X Universal interaction effect was found for this variable with the parents in the combined Indicated-Universal cell having higher baseline levels of non-supportiveness than the other three cells. This variable was controlled in subsequent analyses. No additional significant differences were noted at baseline for the intervention conditions.

Implementation: With regard to the Universal Intervention, overall teacher attendance at the Universal Teacher Meetings was 63%. Of the 672 parents randomly assigned to the Universal Parent Meeting Intervention, 141 (21%) attended at least one meeting. With regard to the Indicated Interventions, attendance for the Coping Power Child Component sessions was 84% across the 16 months (85% in fifth grade, 83% in sixth grade). The Coping Power target children received a mean of 6.0 individual child contacts across the 16 months. Attendance rates for the Coping Power Parent Component was 26% (29% in fifth grade, 21% in sixth grade), with 62% of parents attending at least one of the sessions. In addition to their group contacts, the Coping Power target parents received a mean of 6.5 individual contacts from Coping Power staff over the 16 month period.

Substance Use: The three intervention cells as a group displayed significantly lower alcohol, tobacco, and marijuana use than did the control cell.

Self Regulation: Analysis of aggressive behavior and internal/behavioral self-regulatory process resulted in an overall Time X Indicated X Universal interaction effect. Children receiving the Indicated intervention had greater reductions in parent-rated proactive aggressive behavior in comparison to children who did not receive the Indicated intervention. An Indicated X Universal X Time interaction effect was also found for teacher-rated aggressive behavior and the combined intervention produced the greatest reduction over time. Similar effects were found for the parent-rated reactive aggression measure, although these trends were not significant. Children who were in the Universal intervention had significantly lower decreases in fear over time. The Indicated intervention produced a tendency toward a decrease in activity level over time, and there was a non-significant trend toward an interaction effect for the Dysregulation variable.

The repeated measures MANOVA of teacher-rated measures of children's problem behaviors at Times 2 and 3 revealed a main effect for the Indicated intervention, and a Measure X Time X Indicated interaction effect. Children in the Indicated intervention had increasing levels of teacher-rated behavioral improvement over time, and had a trend toward greater reductions in teacher-rated proactive aggressive behavior.

Social Competence: Analysis of Social Behavior revealed a trend for a Time X Universal interaction effect, and a significant Measure X Time X Indicated X Universal interaction effect. Children in the Universal intervention showed significant reductions in their teacher-rated social relationship problems, and hence had become better accepted by their peer group. A significant Indicated X Universal X Time interaction effect of the Child Social Competence variable, and the significant IU planned contrast, indicated that children who had received both interventions displayed greater increases in their perceived social competence than did children in the other cells. A trend emerged for an interaction effect for Indicated X Universal X Time for the Anger variable, and the IU combined intervention cell tended to have less of an increase in anger over time in comparison to the other cells. For the social skills variables measured only at Time 2 and 3, children in the Indicated intervention displayed significant improvement in teacher-rated social skills and tended to have increasing levels of teacher-rated Problem-Solving and Anger-Management skills.

School Bonding: There were no significant effects found for school bonding variables.

Parenting: Because of noted baseline differences between conditions for the Lack of Supportiveness variable, an ANCOVA was conducted using the Time 3 levels of Lack of Supportiveness as the dependent variable. The covariance analysis yielded a trend for the Indicated intervention effect over time for the Lack of Supportiveness variable. No intervention effects were found for the harsh parenting measure.

One Year Follow-up (Lochman & Wells, 2003): The summer interview battery was administered to 201 participants (83% of the original sample) at the One-Year Follow-up. Teacher measures collected as part of a full classroom assessment at the One-Year Follow-Up were available on 160 participants (66%) of the original sample). The latter measure was not collected when children moved out of their original schools, and some teachers did not complete all of the teacher measures. There were no significant differences between attrited and non-attrited participants in terms of self-reported data or teacher ratings of all students in their classrooms. Thus, the sample available for the analyses at the One-Year Follow-Up (three years after the original baseline assessment) is representative of the original sample at baseline.

Delinquency: The ANOVA for participants' self reported delinquency indicated that Intervention children had significantly lower levels of delinquent behavior than did the Control children. None of the interaction effects for Intervention were significant.

When the planned Intervention cell contrasts for the two Coping Power conditions were conducted, the CP condition had significantly lower delinquency rates than did the C (control) condition, and the CPCL (Coping Power plus Classroom Intervention) condition had lower delinquency rates than the C condition. There was no difference between the classroom intervention alone and the control condition.

Substance Use: The ANCOVA for substance use indicated that Intervention children had lower substance use rates at the One-Year Follow-Up than did the control children. The interaction effects for Intervention X Age and Intervention X Screen Status indicated significant group differences on these variables (i.e., the CP condition had greater effects on substance rates for older children than for younger children, and had greater effects on children with moderate initial screening scores than on children with higher initial screening scores, relative to the control condition).

Because of the significant interactions (noted above), planned contrasts were done separately for older and younger children, and children with moderate and higher screen scores. Older children in the CP condition had significantly lower substance use rates than did the C condition, but there was no significant difference between younger children in the CP and C conditions. Older children in the CPCL condition had significantly lower substance abuse rates than did the C condition; although there was no significant difference between younger children in the CPCL and C conditions, the effect size for the change in substance use for these 2 conditions was only slightly smaller than for the older children (effect size: .50). It should also be noted that the number of participants in each of the above cells is relatively small, resulting in lower power for usual significance tests.

Moderate risk children in the CP condition had significantly lower substance abuse rates than did the C condition, but there was no significant difference between high-risk children in the CP and C conditions. When the planned contrast was conducted between the Classroom Only and Control conditions, the intervention effect was also statistically significant.

School Behavior: The ANCOVA for teacher-rated aggressive behavior at school indicated that Intervention boys had greater behavioral improvement by the One-Year Follow-Up than did the Control boys. None of the interaction effects involving Intervention were significant, indicating that the Coping Power program had similar effects across the various potential moderator variables.

When the planned Intervention cell contrasts for the two Coping Power conditions were conducted for School Aggression, the CPCL condition tended to have lower school aggression rates than did the C condition, but the CP condition did not differ from the C condition. When the planned contrast was conducted between the Classroom Only and the Control condition, the Intervention effect was not significant.

Three-Year Follow-Up (Lochman et al., 2013): Along with the additional two years of data it examined, this study of the indicated treatment sample added geocodes for residence and measures of neighborhood characteristics. The authors hypothesized that the Coping Power program would have stronger effects in less disadvantaged and more socially organized neighborhoods, where problems won't overwhelm the influence of the intervention.

Attrition. Retention rates for parent and child assessments were 91% at time 2, 88% at time 3, 83% at time 4, 82% at time 5, and 86% at time 6; retention rates for teacher assessments were 84% at time 2, 69% at time 3, 64% at time 4, 65% at time 5, 60% at time 6, and 68% at time 7.

Measures. The study focused on four outcome measures that partially overlap with measures used in earlier studies. The four measures included the:

• Teacher Observation of Classroom Adaption-Revised (TOCA-R) for aggression,
• Teacher Observation of Classroom Adaption-Revised (TOCA-R) for academics,
• reduction in aversive treatment subscale of the Outcome Expectations Questionnaire, which measures expectations of the children that aggressive behavior will lead to good outcomes, and
• lack of support subscale from the Family Relations Scale, which measures low parental warmth and low positive involvement with children.

Two neighborhood measures included 1) a scale based on nine items measuring parents' baseline perceptions of belonging to and support from the neighborhood, and 2) a disadvantage scale based on combining census measures such as poverty, unemployment, and owner-occupied housing for the subjects' neighborhood of residence.

Analysis. Hierarchical Linear Modeling (HLM) was used with time as the first level, student as the second level, and neighborhood as the third level. The model included the condition indicator at level 2 and allowed the level-3 neighborhood measures to moderate the condition effect. The authors stated, "Missing data is estimated within the HLM analyses," but offered no detail on the procedure.

Differential Attrition. In logistic regressions with attrition (yes or no) as the outcome, only 5 of 65 tests of possible attrition were significant (i.e., tests were done for 5 child characteristics by 6 parent-child assessment points plus 5 child characteristics by 7 teacher assessment points). The five reaching significance revealed no systematic pattern of bias across time.

Results. The Coping Power program significantly improved the growth curve of the intervention group relative to the control group for all four outcomes: teacher-rated aggression, teacher-rated academics, expectations of benefits from aggression, and lack of parental support. The benefits held with controls for neighborhood disadvantage and neighborhood social organization.

Tests of moderation showed that neighborhood characteristics significantly interacted with the program in two of eight tests, but in the opposite direction. On one hand, the program did more to reduce lack of parental support in better neighborhoods with low disadvantage. On the other hand, the program did more to reduce teacher-rated aggressiveness in worse neighborhoods with low social organization. The lack of a consistent pattern of moderation suggests that the program works across diverse neighborhoods.

Study 2

Evaluation Methodology:

Design: This publication reports on an efficacy study conducted prior to Study 1 described above. Fourth- and fifth-grade boys were screened and selected for potential involvement in the study and were retained in the high-risk pool if they surpassed cutoffs on detailed behavioral measures on the Aggression subscale from the Achenbach Teacher Report Form, as rated by teachers and parents. Eighty-four fourth and fifth grade teachers in 11 elementary schools completed the screening process for Cohort 1 and 86 fourth and fifth grade teachers in 12 elementary schools completed the screening process for Cohort 2. The sample (N = 183) were in the top 22% of boys in teachers' ratings of children's aggressive and disruptive behaviors. Among the three risk components, there were no significant differences in ethnicity, grade level, cognitive ability, or screen score among those who consented to participate (59% of those approached) and those who did not. The 183 boys in both cohorts were randomly assigned to the child intervention only condition (CI) (N = 60), the child plus parent intervention condition (CPI) (N = 60), or the control condition (C) (N = 63). Both cohorts were assessed at three times: pre-testing (Time 1), post-intervention (Time 2), and a one-year follow-up (Time 3). Time 3 assessments were collected two summers after intervention (when boys had completed either sixth or seventh grade).

Intervention occurred in grades 5 and 6. Thirty-three structured group sessions were completed by 29 children groups across the 2 cohorts (fifth and sixth grades). Sixty parents from the 2 cohorts were assigned to 11 parent groups, and each of these groups met for 16 sessions over the 15 month intervention period. Attendance rates at child group sessions was 83% and attendance at parent group was 49%.

The child component consisted of a 1.25 year intervention conducted in the school setting. There were 33 weekly group sessions which lasted 40-60 minutes per session. Eight intervention sessions were held during the first intervention year and 25 were held in the second year, with sessions being somewhat shorter in the second year due to shorter class periods. Group sessions consisted of four to six boys and were co-led by a school-family program specialist and a school guidance counselor. Each boy also received approximately 12 individual ½ hour sessions.

The parent component consisted of 16 sessions, over the same 15 month intervention period. This intervention was delivered in groups of four to six single parents and/or couples and was led by two grant staff persons. Groups began with weekly meetings that were gradually reduced over the year to biweekly and then monthly sessions. Child care and transportation were provided to those in need. Parents also received stress management training in two of the sessions. The content of the child and parent components were synchronized so that the social-cognitive and problem-solving skills were introduced at the same time.

Sample Characteristics: Because funding was not sufficient to obtain a large enough sample to accurately analyze gender differences, only boys were selected for inclusion in the study. A normative sample of 63 boys was representative of the non-risk portion of the population. These boys were in the lower 78% assessed for aggression-disruptiveness. This sample was 64% fourth graders and 36% fifth graders and 47% African American and 53% White. 55% of those receiving one of the two intervention groups were fourth graders, 45% were fifth graders, and 38% were White.

Measures: Time 3 outcomes were multisource. Measures used in the analysis included self-report information (delinquency, via the National Youth Survey), and teacher and parent ratings (school behavior improvement and substance use, respectively). The substance use outcomes included cigarette, alcohol, and marijuana use and were measured using the National Youth Survey as well as parental reports. Teacher ratings of children's behavioral improvement covering the follow-up year were blind raters. There were four Time 3 outcome measures and nine Time 2 intervention change variables. There were two measures of social cognitive processes: the Attributional Measure and the Outcome Expectation Questionnaire, and two measures of schematic beliefs: the Multidimensional Locus of Control scale and the Object Representation inventory. Parenting processes were assessed with the Alabama Parenting Questionnaire, providing subscales for assessing Inconsistent Discipline and Maternal Involvement. The measures targeted intervention change variables that could serve as variables mediating intervention-outcome relations.

Analysis: Intervention effects were tested with ANOVAs and ANCOVAs using general linear models. Dependent variables were the Time 3 outcome variables. When baseline scores were available, they were used as a covariate in the ANCOVA; when baseline scores were not available, ANOVAs were conducted. The initial analysis for each dependent variable examined main effects for the four independent variables of Intervention, Ethnic Status, Grade level during intervention, and Screening Status Level, and the three interaction effects for Intervention X Ethnic Status, Intervention X Grade, and Intervention X Screening Status. If a significant or trend main effect for Intervention existed, then two planned comparisons were conducted to examine any differences between the Child Intervention (CI)and Control condition (C) and the Child plus Parent Intervention (CPI) to the Control condition (C).

Outcomes

There was 12.7% attrition in the Normative group by Time 3, with no significant differences between attrited and non-attrited participants on aggression, race, or cognitive abilities. There was 30%, 31%, and 27% attrition for intervention boys, parents, and teacher reports, respectively. More whites attrited than non-whites. Baseline equivalence analyses conducted for all individuals with Time 3 data, rather than the original sample, revealed no significant differences between conditions on cognitive abilities, ethnicity, grade level, and the dependent variables.

Delinquency: The results from the MANCOVA on self-reported covert delinquency revealed that the Intervention condition had greater reductions in delinquent behavior than did the Control group. When the planned Intervention cell comparisons were conducted, however, the CI condition was not significantly different from the C condition, while the CPI condition produced significantly greater reductions in delinquency in comparison to the C condition. There were no significant intervention effects on overt delinquency.

Substance Use: There were no intervention results for self-reported substance use. The ANOVA for parent-reported alcohol and marijuana use indicated that Intervention boys had lower rates of substance use at the Time 3 follow-up than did the Control boys. A significant interaction effect for Intervention X Ethnic Status was also significant, indicating that intervention white boys had lower rates of substance use than did white boys in the C condition. The substance use rates for intervention minority boys and control minority boys were relatively similar, and were both lower than the cells for white boys. A planned ANOVA using Income Level in place of Ethnic Status revealed a trend toward significance for the interaction effect for Intervention X Income Level with Intervention effects on substance use evident for higher income youth but not for lower income youth. When the planned Intervention cell comparisons were conducted, the CI condition was not significantly different from the C condition, while the CPI condition produced significantly greater reductions in parent-rated substance use than did the C condition.

School Behavior: The ANOVA for teacher-rated behavioral improvement at school indicated that Intervention boys demonstrated greater improvement during the Follow-Up year than did the Control boys. The interaction effect for Intervention X Ethnic Status was also significant, indicating that white boys in the CI condition had higher ratings of behavioral improvement at school than did white boys in the C condition. The ratings for minority boys in the CI and C conditions were relatively similar, and both were between the ratings for the white CI and C boys. When the planned Intervention cell comparisons were conducted, the CI condition had greater behavioral improvement than did the C condition, and the CPI condition had greater behavioral improvement than did the C condition.

Intervention Effects on the Targeted Intervention Change Variables at Post Intervention: These intervention effects were tested with ANCOVAs using general linear models with the Time 2 targeted intervention change variables as dependent variables.

Social-Cognitive Processes: The ANCOVA revealed that the Intervention boys tended to have greater reductions in attributional biases and anger than did the boys in the Control condition. When the planned Intervention cell comparisons were conducted, the CPI condition was not significantly different from the C condition, but the CI condition produced a greater reduction in attributional biases and anger than did the C condition. The interaction effect for Intervention X Grade was significant, indicating that boys in the Intervention in the fourth-fifth grades had greater changes in their expectations that aggression would not have good outcomes than did boys in the Control condition. The planned Intervention cell comparisons indicated that both the CI and the CPI conditions tended to have less optimistic expectations that aggression would lead to useful tangible outcomes in comparison to the C condition.

Schema: Intervention boys had greater increases in their internal locus of control for producing successful outcomes than did the Control boys. None of the interaction effects with Intervention were significant. The planned Intervention cell comparisons indicated that both the CI and the CPI conditions had greater increases in internal locus of control in comparison to the C condition. The Intervention boys demonstrated greater improvements in person perception than did the Control condition. Again, none of the interaction effects with Intervention were significant. The planned Intervention cell contrasts indicated that the CPI condition had greater improvements in person perception in comparison to the C condition, while the improvements for the CI condition were not significant.

Parenting Behaviors: No Intervention effects were obtained for parent or child reported parental involvement, or for parent reports of inconsistent discipline.

Normative Comparisons for Intervention Outcome Effects: In order to determine if the intervention children had moved to within a normative range at follow-up, two types of comparisons were made. First, ANOVAs were conducted to determine if aggressive children in either the intervention or control groups were significantly different from normative children on the outcomes that had been found to be significantly affected by intervention. Second, normative comparisons were conducted to determine if the normative and intervention means were close enough to be considered clinically equivalent. At the one-year follow-up, ANOVAs indicated that the aggressive control group had higher levels of delinquency and weaker levels of behavioral improvement at school in comparison to the normative group. ANOVAs for the aggressive boys in the intervention condition, however, indicated that these boys were not significantly different from the normative boys in their rates of self-reported delinquency, parent-reported alcohol and marijuana use, or in teacher-rated behavioral improvement at school. Thus, aggressive boys in the intervention group moved from a non-normal range to a normative range by the one-year followup. Equivalency testing, using one-tailed z-tests, was conducted using one standard deviation around the normative mean as the range of closeness. These tests indicated that the aggressive intervention boys were significantly within normal limits in their rates of delinquent behavior, their parent-reported levels of substance use, and their school behavioral problem improvements.

Test of the Theoretical Model (Lochman & Wells, 2002b): Path analyses were performed to test the hypothesis that the intervention effects on Time 3 outcomes would be mediated through changes from Time 1 to Time 2 in targeted intervention change variables. The analyses examined the two intervention cells combined together (N = 120) in comparison to the control condition (N = 63). Path analyses models examined the relationships between the exogenous variables and the endogenous variables (i.e., that the changes in Time 3 outcomes would be mediated through changes from Time 1 to Time 2 mediating variables).

Delinquency Outcome: Goodness of fit indices for the various models of delinquency outcomes indicated that the null model provided a poorer fit than either the base or the mediation models. The base model is one in which the intervention has direct effects on Time 3 delinquency and Time 2 mediating variables. The mediation model is one in which paths from each of the Time 2 mediating variables are added to the base model to determine if the addition of these mediation paths will produce a nonsignificant path from intervention to Time 3 delinquency. Consistent with the prior ANCOVA results, the base model intervention produced significant reductions in the delinquency outcome and had significant effects on outcome expectations, internal locus of control, and parental inconsistency; and tended to improve person perceptions. For the mediation model, the path from intervention to the Time 3 delinquency outcome changed from being significant in the base model to being nonsignificant in the mediation model, showing that the delinquency reductions were a result of changes in the theorized mediating mechanisms. In this model, Time 3 delinquency was significantly predicted by lower levels of Time 2 parental inconsistency and tended to be predicted by lower levels of hostile attributions.

Substance Use Outcome: Goodness of fit indices for the various models for the parent-rated substance use outcome indicated that the null model provided a significantly poorer fit than both the base model and the mediation model. In the base model, intervention tended to predict lower levels of Time 3 substance use, significantly predicted better outcome expectations and more internal locus of control, and tended to predict lower parental inconsistency. Although the mediation model did not provide a significantly better fit than the base model, the critical path for the mediation test indicated that the effect of intervention is at least partially mediated by the addition of the five Time 2 targeted intervention change variables despite the fact that none of the five Time 2 variables alone predicted the Time 3 substance use outcome.

School Behavior Outcome: Goodness of fit indices again indicated that in the case of the various models for the teacher rated school behavior outcome the null model provided a significantly poorer fit than both the base and mediation models. For the base model, intervention predicted significantly greater improvement in Time 3 school behavior during the follow-up year, predicted significantly more internal locus of control and less parental inconsistency, and tended to predict better outcome expectations and person perception. The mediation model did not provide a significantly greater fit to the data than the base model. With regard to the critical path for the mediation test, the effect of intervention appears to be at least partially mediated by the addition of the five Time 2 targeted intervention change variables in the mediation model. Although none of the five Time 2 variables alone significantly predicted the Time 3 school behavior outcome, lower levels of Time 2 hostile attributions tend to predict improved Time 3 school behavior.

Study 3

Description: This study examined the dissemination of the Coping Power prevention program as implemented in a field trial in 57 schools within five school districts. Existing school staff (school counselors) were trained to use the Coping Power program with high-risk children at the time of transition to middle school. This study was designed to examine whether the intensity level of training influences the intervention outcomes. Counselors at the schools were randomly assigned to one of three conditions: Coping Power-training plus feedback (CP-TF), Coping Power basic training (CP-BT), and a comparison group.

CP-TF provided more intensive training and had four training components. First, the school counselors received a total of three initial workshop training days in the fall prior to the beginning of intervention. Second, the school counselors participated in monthly ongoing training sessions in which the trainers provided concrete training for upcoming sessions, debriefed previous sessions, and conducted problem solving concerning barriers and difficulties involved in the implementation of the program. Third, individualized problem solving concerning barriers and difficulties in the implementation of the program was available to counselors in the CP-TF condition through a technical assistance component. This component included access by the implementation staff to an email in which they could raise implementation concerns and problems and through which they could receive trainers' responses. It also included a telephone hotline through which trainers were available for consultation about these same concerns.

Fourth, the trainers reviewed the rate of completion of session objectives and provided individualized supervisory feedback through written and telephone contacts with the school counselors to enhance the intervention integrity. Counselors in this condition received from their trainer a monthly letter, followed up with a phone call when serious concerns with implementation were evident. In addition to noting the objectives that had been fully, partially, or not met based on the trainers' review of session audiotapes, the trainers provided qualitative feedback on the enthusiasm of children's or parent's involvement in the session, the ability of counselors to stimulate discussion and elaborate and clarify material while still staying on the topic, the counselors' ability to engage students in positive ways, and the counselors' use of appropriate monitoring and consequences during sessions.

CP-BT had two training components. First, the school counselors received a total of three initial workshop training days in the fall, prior to the beginning of the intervention. Second, the counselors participated in monthly ongoing training session in which the trainers provided concrete training for upcoming sessions, debriefed previous sessions, and conducted problem solving concerning barriers and difficulties involved in program implementation. These sessions were conducted separately for CP-BT counselors but were equivalent to the ongoing monthly training sessions for the CP-TF condition.

Evaluation Methodology:

Design: Pre-intervention assessment began during the summer after third grade. At school, data were gathered from teachers in the spring of the pre- and postassessment years. The Coping Power program was delivered during the fourth and fifth grade years. Two annual cohorts of schools and children were recruited. Cohort 1 had 13 schools with 118 children (CP-TF: 33, CP-BT: 50, comparison: 35). Cohort 2 had 44 schools with 413 children (CP-TF: 135, CP-BT: 133, comparison: 145). Postintervention assessments were collected from children and caretakers in the summer after the fifth grade year, two years after baseline data collection.

Fifty-seven public schools in five schools systems in north central Alabama were randomly assigned to one of three levels of training conditions. Assignment was stratified so that each school system had at least one comparison, one CP-TF, and one CP-BT school. Random assignment occurred at the level of the school counselor. Sixteen of the participating schools shared a school counselor. These eight pairs of schools were yoked for the purpose of assignment, so that schools that shared a counselor were assigned to the same condition. Yoking of schools was another level of stratification, and yoked pairs were randomly assigned to condition, with at least two yoked school pairs per condition. Assignment was stratified to ensure that the final number of schools in each condition was equal (19).

A teacher-rating approach was used to identify at-risk students who were eligible for the indicated intervention. During screening, third-grade teachers were asked in the spring to complete the Teacher Report of Reactive and Proactive Aggression for all the children in their classes. Based on these ratings, the 30% most aggressive children across all classes were determined. Because children in the upper 2nd percentile were believed to be more likely to already have psychiatric diagnoses and to be engaged in severe antisocial behavior, these children were excluded. The selection criterion (30%) was based on the distribution of teacher ratings across all of the third-grade classes, rather than identifying the 30% most aggressive children in each class.

Of the 3,838 children screened, the scores of approximately 1,422 children fell within the range for inclusion in the study. Of the children who were eligible to participate, 752 were not contacted due to wrong or disconnected phone numbers or less frequently because the maximum number of children to be included from their schools (10) was reached and included in the study before they could be contacted. Contact to schedule interviews was made with a total of 670 potential participants. Of these potential participants, 531 (37% of the 1422 eligible) agreed to participate and were assessed at baseline. Of the total sample, 183 children were in CP-BT schools, 168 children were in CP-TF schools, and 180 were in comparison schools. The aggression screener scores of the 531 participants (17.4) were not different from the screener scores of the 891 children who met the screening range to be included but were not contacted or did not agree to participate (16.9). Data was available on 94% of this sample through postintervention. Differential attrition for completers vs. attriters within the three conditions was tested. Three of 21 tests were significant. Comparison attriters had higher externalizing problems, lower social skills according to parents, and lower expectations that aggression would lead to good outcomes.

Sample: Of the total sample, 183 children were in CP-BT schools, 168 children were in CB-TF schools, and 180 were in comparison schools. The students in the sample were 65% male and 35% female. 84% were African-Americans, 14% were Caucasian, and 2% were of another race/ethnicity.

There were 15 counselors in the CP-TF condition, 17 counselors in the CP-BT condition, and 17 in the comparison condition. The 49 counselors in the three conditions were equivalent in terms of years of experience and in race. 51% of the counselors were African American (47% in CP-TF, 53% in CP-BT, 53% in comparison). Only two (4.1%) of the 49 counselors were male - one in the CP-TF group and one in the comparison group. The counselors in the CP-TF had an average of 11.9 years of experience (SD=7.2). Those in the CP-BT group had an average of 10.1 years of experience (SD=7.6) and those in the comparison group had an average of 9.4 years of experience (SD=5.9). 79% of the counselors were trained at the master's degree level, with the remainder having bachelor's or doctoral degrees (CP-TF: 15 masters; CP-BT: 1 bachelor's, 13 master's, 3 PhDs; comparison: 10 master's, 6 PhDs, 1 unknown).

Measures: Child delinquency, substance use, and behavioral outcomes were measured using questions from the National Youth Survey and the Behavior Assessment for Children (BASC). The National Youth Survey measures provide self-report information on children's substance use and delinquent behaviors. The BASC is a behavior problem checklist completed for this project by children's teachers and by parents. The BASC contains scales assessing both clinical problems and positive traits, with items rated from 0 to 3. This study calculated four composite scores for the parent and teacher reports. The Externalizing Composite (Parent) and Externalizing Composite (Teacher) were derived from the BASC Aggression, Conduct Problems, and Hyperactivity subscales for both parents (possible raw score range: 0-99) and teacher (possible raw score range: 0-111) reports. The BASC Social Composite (Parent) encompassed parent reports on the social skills, leadership, and adaptation subscales and the BASC Social/Academic Composite (Teacher) includes these same subscales as well as the Study Skills subscale.

An adaptation of the Outcome Expectations Questionnaire (OEQ) was used to measure child and parent mediating processes. These analyses were conducted to determine if intervention effects also occurred on potential mediating processes that the intervention targeted. The OEQ consists of twelve brief vignettes in which subjects are asked to imagine that they are performing a behavior toward a specific classmate and then to indicate their level of confidence that a particular consequence would ensue. For this study, the OEQ's Reduction of Aversive Treatment subscale was also used (with possible scores ranging from 1 to 4, with higher scores indicating more certainty that aggression would lead to desirable outcomes).

Parents' use of inconsistent discipline practices was assessed through parent self-report on the Inconsistent Discipline subscale of the Alabama Parenting Questionnaire (APQ). The researchers utilized the Inconsistent Discipline subscale consisting of six items rated on a 5-point Likert scale (item scores are averaged; thus the subscale score range is 1 to 5).

Analysis: Hierarchical linear modeling (HLM), with counselor as the second level in a mixed model design, was used to evaluate if intervention conditions influenced children's externalizing behavior problems (as rated by teachers, parents, and children) and proximal processes that were targeted by the intervention (positive social and academic behaviors as rated by parents and teachers, children's outcome expectations for aggressive behavior, and parents' consistency of discipline). To obtain estimates of effect sizes, the continuous variables were standardized (via z scores) and assigned values (0.05 and +.05) to dummy-coded treatment conditions so that beta weights could be interpreted as in a standard linear regression model context. Full maximum likelihood estimation was used for all models.

The form of the HLM model consisted of two equations: a within- and a between-units model. The within-units are students as Level 1, and the between-units are counselors as Level 2. The within-unit model represented the postintervention score for student i under counselor j as a function of various students in the intervention group, baseline score of the dependent variable, and random error. The intercept for each variable was allowed to vary among counselors.

There were three conditions (CP-TF, CP-BT, and comparison) and three comparisons of interest in the study. The three comparison were CP-TF versus comparison, CP-BT versus comparison, and CP-TF versus CP-BT. The first dummy variable (TRT1) compared the CP-TF group and the comparison group. TRT1=1 if participants were in the CP-TF group, and TRT1=0 if participants were in the CP-BT or comparison group. The second dummy variable (TRT2) compared the CP-BT groups and the comparison group. TRT2=1 if participants were in the CP-BT group, and TRT2=0 if participants were in the CP-TF or comparison group. Continuous postintervention dependent variables scores and dependent variable baseline scores were standardized in these analyses (M=0, SD=1) in these analyses. The model had fixed intervention effects, and the Wald test was used to test comparisons between the two intervention conditions.

Outcomes

The study tested the following three behavioral hypotheses:

1. Students of counselors who received training plus feedback in Coping Power (CP-TF) will have lower rates of externalizing behavior problems according to teacher and parent reports than will comparison children by the end of the intervention period.
2. Students of counselors who received CP-TF will have proximal improvements in positive skills targeted by the intervention (social and study skills, expectations for consequences of aggressive behavior, consistent parenting) than will comparison children by the end of the intervention period.
3. CP-TF will have stronger effects on these outcomes than will CP-BT, indicating the importance of training plus feedback for implementation of prevention programs.

Results:

Hypothesis One: At postintervention, the CP-TF condition had preventative effects on rates of BASC Externalizing Composite according to teachers, as the CP-TF children had lower scores at postassessment than did the comparison children, controlling for the preintervention scores. The CP-TF children maintained their levels of teacher-rated BASC externalizing behaviors in the two years between the pre- and postassessment, while the comparison children increased their teacher-rated BASC externalizing behavior over that time. Children in the CP-TF condition had significantly lower rates of BASC externalizing problems according to parents and lower rates of self-reported assaultive behaviors relative to the comparison condition. Both the CP-TF and comparison children had reductions in parent-rated BASC externalizing behaviors over time and increases in self-reported assaultive behaviors over time, but the CP-TF children's changes were significantly better than the comparison children.

Hypothesis Two: The test for intervention effects indicated that at postintervention, the CP-TF condition had higher rates of teacher-rated BASC positive social and academic behaviors in comparison to the comparison condition. Although both groups of children demonstrated improvements in teacher-rated BASC academic and social skills over time, the CP-TF children had greater improvements in academic and social skills in the school setting over the intervention period than did the comparison children. The CP-TF children had lower self-reported expectations that aggression would lead to positive outcomes in conflicts with others than the comparison group, who had increased expectations that aggressive behaviors would lead to good outcomes. There were no intervention effects evident for parent-rated BASC positive social behaviors or for parents' self-reports of their inconsistent discipline.

Hypothesis Three: While the CP-TF children maintained their levels of teacher-rated BASC externalizing behaviors over time, the BASC externalizing behaviors of the CP-BT children became worse, according to teacher reports. The CP-BT children had larger increases in self-reports of assaultive behaviors over time than did the CP-TF children. While the CP-TF children had decreases in their expectations that aggression would lead to good outcomes for them, the CP-BT children had increases in their expectations of the utility of aggression.

Two-Year Follow-Up: The follow-up study (Lochman et al., 2012) focused specifically on academic functioning among the subjects, including those identified as having special education needs. The authors argued that improvements in social skills should aid academic progress of both regular and special education students. As in the previous article, the design compared two versions of the program, one involving basic training of counselors and one involving intensive training. With the intervention implemented for students in 4^th and 5^th grade, this two-year follow-up examined subjects through seventh grade.

Measures. The study gathered data from school records for 3^rd through 7^th grade. Language arts and math grades were averaged within years and ranged from 0 to 100. Another measure distinguished students receiving special education services for emotional disturbance, other health impairment, specific learning disability, or developmental delay. About 30% of the sample received the special education services.

Analysis. Multilevel growth curve models using HLM estimated level-1 changes in language arts and math grades from 3^rd to 7^th grade. Participation of students in special education was measured at level 2, while the school-based conditions were measured at level 3.

Baseline Equivalence. There were no significant baseline differences in language arts grades, but there were significant differences in math grades, with the Coping Power basic training condition having higher scores than the control condition.

Attrition. For the 531 subjects, data from school records were obtained for 66% of the five assessment points. Those with missing data for three or more of the assessments were compared to those with less missing data. The two groups did not differ significantly on race, gender, baseline screening score, or condition assignment.

Follow-Up. Language arts grades declined over the study period for all groups. However, the intensive training intervention significantly moderated the decline relative to the control group. The intensive training intervention failed to influence the change in math grades, and the basic training intervention influenced neither language arts nor math grades.

Tests found that the influence of the intensive training program did not differ significantly by receipt of special education services. Thus, the intervention similarly helped students with as well as without special needs.