Principal Investigator: David Grossman
Level of Intervention: Universal
Target Population: elementary-aged children (grades 1-3); similar curriculum available for other age groups.
References: Grossman, Neckerman, Koepsell, Liu, Asher, Beland, Frey, & Rivera (1997)
Theory (Risk & Protective Factors Targeted):
Second Step targets early and persistent antisocial (aggressive) behavior and seeks to promote prosocial behavior as reflected by competence in peer interactions and friendships and in interpersonal conflict resolution skills. It seeks to improve interpersonal problem-solving skills by training children in cognitive processing (identifying the interpersonal problem and generating non-aggressive solutions).
Description of Intervention:
The Second Step curriculum consists of 30 lessons, 35 minutes each, taught once or twice per week in a classroom setting. The program is designed to teach anger management, empathy, and impulse control. Lessons consist of photograph lesson cards accompanied by a scenario that forms the basis for discussion and role-plays. A video-based parents guide provides familiarization with the program and encourages the reinforcement of skills at home. The home intervention component was not available at the time of the study and thus was not assessed in the evaluation.
The study assessed 790 second and third grade, primarily white students from 12 elementary schools (6 matched pairs) in Washington state. The students were 53% male.
Randomized controlled trial with 12 elementary schools from 4 school districts in King County (Seattle) Washington. Schools were used as the unit of randomization to reduce potential contamination between intervention and control groups. Schools were selected based on: no previous experience with Second Step or similar curricula; at least 4 classrooms of 2nd and 3rd graders; faculty and principal approval of the curriculum and evaluation; and school agreement not to introduce other related programs during the study. Schools and classrooms were also selected based on their perceived willingness to deliver the curriculum and facilitate the evaluation. Schools were paired according to school district, proportion of students eligible for free or reduced-cost lunches, and proportion of minority students. After matching, schools in each pair were randomly assigned to intervention or control groups. 49 classrooms (equal numbers of 2nd and 3rd grade) participated. Active parental consent was used.
Data were collected at pre, 2 weeks post, and 6 month followup, and included parent and teacher ratings as well as coded observations.
Teacher ratings included 2 self-administered instruments for all enrolled subjects the School Social Behavior Scale and the Teacher Report From (an adaptation of the Child Behavior Checklist, using only the delinquency and aggression subscales.) Two items on the TRF related to extreme behavior were eliminated at the request of school administrators.
Parent ratings included 2 self-administered instruments for all enrolled subjects CBCL, using the 2 subscales analogous to the aggression and delinquency scales of the TRF, and the Parent-Child Rating Scale (P-CRS). A combination of monetary incentives, mailings, and telephone calls were used to increase parent response.
Direct observations were recorded for 12 randomly selected subjects from each classroom (588 total students) using the Social Interaction Observation System (modeled after the State-Event Classroom Observation System). Observations were done by two observers spending 2 to 4 days in each classroom. Entries were made in 30 ten-second intervals for a total of 5 minutes. Approximately 60 minutes of observation time was conducted for each subject during each of the first two data collection points, and approximately 45 minutes for each at the third. Coded behavior categories included prosocial and neutral, verbal negative, and physical negative. Prosocial and neutral behaviors were grouped together because observers had difficulty differentiating between the two categories in the field. Observations were conducted in the classroom, cafeteria, and playground. Approximately 10% of the observations were intentionally sampled simultaneously by both members of the observer team. Using the interclass correlation coefficient, the interobserver agreement mean K value across all time points was .92 for neutral/prosocial behavior, .50 for physical negative, and .45 for verbal negative. Observers were trained in a 4 to 6 week session and were blinded to the study purposes, hypothesis, and research design. Observers, teachers and students were all blinded to student assignment to the observation group. Observers were instructed to have minimal contact with teachers or other school staff. To minimize effects of observers presence in the classroom, a 2-hour habituation period was used upon observers first entry to the classroom, during which time no observation data was collected.
Data collected at 2 weeks post treatment showed reductions in physical aggression (p=.03) and increases in prosocial behavior (p=.04) based on observations, while levels of aggression increased in control group. No significant effects on parent or teacher-rated behavior were found.
Six month followup data showed physical aggression in the classroom remained significantly (p=.03) reduced; other previously identified outcomes did not retain significance.
Strengths & Limitations of Evaluation:
Teachers from intervention schools participated in a 2-day training conducted by the program developers. Two investigators monitored and rated the quality of program implementation twice during the intervention period using a numerical rating. K-score for interobserver reliability was .60.
Participating subjects (those with parental consent and completed parent and teacher surveys) were similar to the total eligible population by sex and SES (as measured by subsidized school lunch eligibility), but were more likely to be white (79% vs 75%). The intervention and control groups were similar in age, sex, teacher reported behavior problems, household composition, family SES, and classroom size. A larger proportion of control subjects were in special education and were African-American, while a higher proportion of intervention subjects were Asian-American.
The study design was sound. While coded observations showed significant effects, there were no differences between groups on parent or teacher ratings. Another (non-randomized) controlled study of the same curriculum in middle schools found no statistically significant improvement (Orpinas, Parcel, McAlister, & Frankowski, 1995).