Wyman and colleagues (2010) used a cluster randomized-assignment research design to evaluate the Sources of Strength program in 18 schools in two phases. The first phase, conducted from 2007 to 2008, included 6 metropolitan schools in Cobb County, Georgia. The second phase, conducted from 2008 to 2009, included 12 predominantly rural schools (8 in New York and 4 in North Dakota). After blocking the schools by state and region and matching by size, 1 school from each pair was randomly assigned to the intervention group (to begin training immediately); the other school was assigned to the control group (on a wait list to begin training 5 months later). The metropolitan schools were larger and had more African American and Hispanic students than did the rural schools. No school withdrew or altered its assigned status.
Peer leaders were recruited using identical, standardized procedures across all schools before assignment to intervention and control groups. Nomination forms submitted by staff members, administrators, and students identified a list of potential student leaders, which was further reviewed to select a diverse group of student leaders representing as many student friendship groups as possible. Each school in phase 1 recruited approximately 2 percent of all students, and each school in phase 2 recruited approximately 10 percent of all students. A total of 720 students were nominated (10 to 54 per school), of which 496 (68.9 percent) were enrolled with parents’ permission and youths’ assent. Of those students enrolled, 453 completed two assessments. The first assessment was conducted before any schools were randomly assigned to a condition (baseline).The second assessment was conducted 4 months later (approximately 3 months after messaging activities began at the intervention schools, but before peer-leader training started in the wait-list control schools).
There were 268 peer leaders in intervention schools and 185 peer leaders in control schools. Peer leaders in the intervention and control schools were largely equivalent in gender and race/ethnicity; however, peer leaders in the intervention schools were 0.5 years younger than those in the control schools. Peer leaders in each intervention school completed at least three of the four messaging steps; participation in messaging ranged from 59 percent to 100 percent across schools. Thirty-six staff members (in four schools in phase 2) were interviewed after the messaging phase to check program fidelity. Approximately 97.2 percent reported observing or receiving intervention-specific messages, and 88.9 percent of those named as trusted adults reported that they had been contacted by a peer leader, as the intervention had intended.
The outcome measures of the study covered three constructs: 1) suicide perceptions and norms, 2) social connectedness, and 3) peer-leader behaviors. Suicide perceptions and norms were measured using the following scales: Help for Suicidal Peers, a 4-item scale assessing perceptions that adults help suicidal students in their school; Reject Codes of Silence, a 6-item scale assessing attitudes toward overcoming secrecy barriers to engage adults for suicidal peers; and Maladaptive Coping, a 4-item scale assessing perceptions such as suicide acceptability. Social connectedness was measured using the following scales: Help Seeking From Adults at School, 4-item scale assessing intentions to seek help, expectations about receiving help, and perceived peer norms about seeking help from adults at school; Sources of Strength Coping, a 9-item scale assessing the student’s status on eight resources that include family, friends, adult mentors, and services; and School Engagement, a 4-tem scale assessing attitudes and participation related to school. Peer leaders were also asked to list the names of adults to whom they would turn for help for a suicidal friend (Trusted Adults; only completed in phase 2 schools). Peer-leader behaviors were assessed using two questions that assessed the frequency of peer leaders’ partnering with adults to help peers (Referred Distressed Peers to Adults; 2 items) and the frequency of supportive behaviors (Support to Peers; 2 items). There were no significant differences at baseline for any of the measures between peer mentors in the intervention and control groups. Peer leaders who completed posttest measures had higher baseline scores on School Engagement than those for whom posttest data were not able to be collected. Peer leader attrition, however, was not related to condition.
Intervention effects on peer leaders were tested using a 2-level, linear mixed-effects model, which included individual youths' gender, grade, age, race/ ethnicity, and baseline scores as random factors at level 1, and intervention condition and state location for the school as fixed factors at level 2. Schools were included in each model as random factors because randomization occurred at the school level. All analyses used an intent-to-treat approach by including all enrolled peer leaders regardless of their level of participation.