Bridging Disciplinary Boundaries (January 11 - 14, 2007)

Purpose: The purpose of this presentation is two-fold. First, it will describe 6-month follow-up findings from a quasi-experimental study of a school-based program intended to prevent aggressive behavior in third-grade children. Second, it will demonstrate the use of concatenated cluster designs, where power is increased in small longitudinal group trials by cross-classifying clusters – in this case, third and fourth grade observations of students over three sequential cohorts.

Based on social information-processing (SIP) theory, a skills-training program was developed and delivered to third-grade students (52% male) in two schools comprised of African American (19%), Latino (40%), non-Latino White (36%), and other (5%) children. Entering third grade in 2000, 136 children received a routine health curriculum. Entering third grade in 2001, 136 children received the Making Choices (MC) program, delivered by program specialists who visited schools once per week. Entering third grade in 2002, 149 children received MC plus (MC+) an intervention designed to involve teachers and parents in program-related activities. The teachers in 2002-3 used supplemental activities to infuse MC training in routine classroom activities. In addition, parents of children in MC+ received a monthly newsletter that included home-application exercises, and they were invited to a series of five evening programs designed to strengthen parenting skills.

Methods: For each cohort, teacher ratings of student behavior were collected in the Fall and Spring of the third grade and at 6-month follow-up in the Fall of the fourth grade. Outcomes include cognitive concentration, social competence, social contact, social aggression, authority acceptance, and overt aggression. Collected in the Spring of the third grade and at follow-up, SIP-related skills were assessed using analogue measures for encoding, hostile attribution, goal formulation, and response decision-making.

Controlling for Fall and Spring scores on each outcome, two-level hierarchical linear models were used to estimate follow-up effects. Level-1 regresses student follow-up scores from each measure on covariates – Fall pretest score, Spring post-intervention score, race/ethnicity, and gender. Level-2 is comprised of cross-classified classrooms concatenated across the third and fourth grades. It regresses the student-level intercept on indicators for MC and MC+. This design controls for the potential reduction of standard errors that may occur as a result of the clustering of children in classrooms.

Results: No significant differences across cohorts were observed at pretest. When compared to the comparison cohort, children in classrooms receiving the interventions had fewer behavioral problems at follow-up. Both MC and MC+ had significant effects on children's social and overt aggression. Post-intervention SIP scores suggest that the programs significantly improved children's encoding and goal-setting skills.

Implications for Practice: Consistent with earlier reports, these follow-up data suggest that strengthening the information-processing skills of children reduces overt and social aggression. Using a cross-classification design, the study demonstrates an important methodological innovation for estimating standard errors when the number of clusters in group randomized and other group-based designs is small. Substantively, the MC program appears to hold the potential to disrupt developmental trajectories that lead from early aggressive behavior to later drug involvement and delinquency.