Objectives. This study assessed the effectiveness of an inner-city peer-mentoring program in modifying the attitudes and behaviors involving violence of preadolescent mentees.
Methods. In a case-matched cohort study involving 7- to 13-year-old children, 50 children enrolled in peer mentoring (case subjects) were compared with 75 control subjects. Case subjects were involved before enrollment in the community program in which the intervention occurred; control subjects lived in the same housing project and were matched with case subjects on age, sex, and census tract. A total of 19 community adolescents mentored the case subjects by designing and presenting violence prevention lessons. Two reliable self-report scales, Determining our Viewpoints of Violent Events and Normative Beliefs About Aggression Scale, were used to measure attitudinal change. Teachers completed the Revised Behavior Problem Checklist to assess changes in behavior.
Results. At baseline, the survey scores of the case and control subjects were not different. After the intervention period, the case scores indicated less support for violence than the control scores. Case behavior scores did not change, but control behavior scores worsened.
Conclusions. The data suggest that peer mentoring for younger children may be an important component of efforts to reduce youth violence. A larger multisite trial is warranted. violence prevention, peer mentoring, community-based.
- CGYP =
- Cabrini Green Youth Program •
- CTC =
- Children Teaching Children •
- DOVVE =
- Determining Our Viewpoint of Violent Events •
- NOBAGS =
- Normative Belief About Aggression Scale •
- RBPC =
- Revised Behavior Problem Checklist
Over the last 4 decades, childhood mortality in the United States has decreased significantly. This success was spoiled by the increase in violent deaths between 1968 and 1992.1 Early intervention, focusing, in part, on educating young children, has been recommended as a key component in violence prevention.2 ,3This recommendation resulted in the implementation of many school-based violence prevention programs, although until recently, there has been little evidence that they worked.4 ,5 Another educational approach that may be helpful in preventing violence is based on a community-based peer-mentoring model.
The notion of peer mentoring is based on the observation that the behaviors of children are learned from adults and peers. These learning relationships can promote prosocial behaviors. Studies in other public health areas, such as, AIDS,6 and tobacco use7 prevention, have suggested that peer mentoring contributes to successful health promotion. The use of peer mentoring in violence prevention has begun to be studied.8–10 We hypothesized that community-based peer-mentoring programs can be a useful component of programs that seek to prevent violence, serving to ameliorate the social risk factors that promote violence. To test this hypothesis, we assessed the effectiveness of a peer-mentoring program, developed as part of the established community-based Cabrini Green Youth Program (CGYP) to decrease violence-endorsing attitudes and aggressive behavior in elementary school-aged children. We also examined the effect of serving as peer mentors on the adolescent mentors.
During an 18-month intervention, adolescents between the ages of 14 and 21 years who were in the CGYP's Children Teaching Children (CTC) program designed and presented lessons to teach the younger children who were between the ages of 7 and 13 years. The effectiveness of this intervention was evaluated using a case-matched cohort study design that compared changes in the attitude, behavior, and injury rates of the younger children in CGYP (case subjects) with those of children from the same housing project who were not in CGYP (control subjects), matched on age, sex, and census tract. Changes over time within the case and control groups were also measured.
To assess the effect of peer mentoring on the adolescent mentors, we used a longitudinal cohort design in which each adolescent served as his or her control subject. Changes in the mentors' attitudes toward violence and their self-esteem were measured and described.
The study protocol was approved by the Institutional Review Board at Northwestern University Medical School (Chicago, IL). For each subject, informed consent for the administration of the surveys was obtained from the guardian.
The CGYP was founded in 1984 by a group of Northwestern University medical students with the goal of improving the life opportunities of children living in Cabrini Green, a Department of Housing and Urban Development housing project in Chicago. The geographical area of Cabrini Green covers four census tracts. In the 1990 census, 9732 persons were recorded in these four tracts; 95% were black and 75% lived below the federal poverty level.11Cabrini Green has the reputation of being one of the most violent neighborhoods in Chicago. There are at least three major gangs who control different buildings in Cabrini Green. In the CGYP, role models from within and outside the Cabrini Green community lead activities in education, recreation, health care, unintentional injury prevention, and violence prevention. The Youth Program provides 20 weekly programs for 400 children very efficiently through the use of volunteers and donated space and materials.
Case Subject Recruitment
All CGYP children between the ages of 7 and 13 years were eligible to participate as study case subjects. Based on an intent-to-treat model, case subjects were considered enrolled if they received one or more lessons and completed data at two or more data collection times.
Control Subject Recruitment
The CGYP is open to all children living in Cabrini Green. The control subjects were children who may have chosen not to be in CGYP (or did not know it existed; many have since joined CGYP after participating as control subjects). Because many Cabrini Green families do not have telephones, control subjects were recruited by a study staff person going door-to-door to find children who matched the case subjects on age (within 1 year), sex, and census tract.
At the time the study began, a gang truce had ended, escalating Cabrini Green's violent reputation. As a result, no one was interested in doing the needed door-to-door recruitment. Eventually, violent gang activity quieted down enough for a mother of one of the CGYP children to undertake control enrollment. Yet, there were still gang-related obstacles. Because the recruiter lived in a building that was controlled by one gang, she was associated with that gang, although she did not actively participate in it. This association limited her safe access to buildings in Cabrini Green that were controlled by rival gangs. Relatives who lived in buildings controlled by members from another gang extended her access to more areas of Cabrini Green and this led to successful enrollment of case-matched control subjects. We enrolled more control subjects than case subjects to allow for anticipated greater attrition.
Adolescent Subject Recruitment
All the adolescents who participated in the CGYP's CTC program served as peer mentors.
During CTC, the adolescent mentors designed lessons to teach younger children about violence prevention. Information was provided through skits, games, and rap music. A total of 12 activities were produced over the 18-month period. Table 1 lists the lessons with brief descriptions.
Survey Tools and Procedures
The scales used to measure study outcomes are described in Table 2. These scales were administered at baseline, midstudy (9 months), and at the end of the study (18 months after baseline). To measure violence knowledge and attitudes, two violence surveys, Determining Our Viewpoint of Violent Events (DOVVE)12 and Normative Belief About Aggression Scale (NOBAGS),13 were given to the younger children. The surveys were administered by a research assistant to the case subjects and control subjects. To assess behavioral change, children were evaluated by their teachers using the Revised Behavior Problem Checklist (RBPC).14 The families of the control group were paid $10 for completing all the surveys. The CGYP children's families did not receive money but were compensated through CGYP activities.
Mentors also completed surveys at baseline, midstudy, and at the end of the intervention, including the DOVVE survey to measure changes in attitudes toward violence as well as the Piers-Harris Children's Self-Concept Scale15 to assess self-esteem. Anticipating that quantitative measures might fail to detect real differences because there were relatively few mentors, research assistants also maintained a qualitative log of the attitudes and behaviors of the adolescents. The mentors were paid $4.50/hour and worked ∼5 hours/week.
A secondary outcome measure was to compare injuries between the two groups, but the incidence was too low for meaningful analysis. Therefore, this information is not presented.
Data were entered using Epi Info 6.2 (Centers for Disease Control and Prevention, Atlanta, GA). Analysis was performed using Statistical Package for the Social Sciences (SPSS), PC+ (SPSS, Inc, Chicago, IL) and SAS Institute (Cary, NC). ANOVA was used for comparison of continuous variables, χ2 was used for categorical variables, and the Wilcoxin Rank-Sum test was used for nonparametric data. Statistical significance was set atP ≤ .05.
All available subject data were included even if data were not available for all three data collection times. At time 1, 13 case subjects entered the study. Their responses were included as baseline data (time 0), and their time 2 responses as time 1 data. Surveys that were not completed by the children or teachers were not included in the analysis.
Case-Matched Cohort Study
Table 3 lists the number of the forms that we received for each survey instrument at each measurement period.
Changes in Attitudes Toward Violence
Figure 1 presents the results of the DOVVE survey when we compared the case subjects and the control subjects. The higher the DOVVE score, the greater the exposure to violence and/or acceptance of violence. At baseline, the case subject and control subject scores were not statistically different from one another (case score: 4.4 ± 3.1; control score: 4.0 ± 2.9;P ≤ .56); this remained true at midstudy (case score: 4.1 ± 2.7; control score: 4.4 ± 3.0; P ≤ .64). However, at the end of the study, the case subjects and the control subjects were statistically different in their responses with case scores being lower, ie, better (3.3 ± 2.3 vs 5.5 ± 3.1; P ≤ .006).
On the NOBAGS survey tool, a higher score is associated with greater acceptance of violence. The overall NOBAGS scores of case subjects and control subjects were not statistically different at baseline (1.69 ± .58 vs 1.74 ± .48; P ≤ .61). At midstudy, the case subjects had a statistically lower (ie, better) score than the control subjects (1.62 ± .48 vs 1.93 ± .48;P ≤ .0008). After the peer-mentoring intervention, the case subjects continued to have a statistically lower score than the control subjects (1.48 ± .39 vs 1.83 ± .39;P ≤ .001).
Neither the DOVVE nor the NOBAGS scores changed statistically within groups from time 0 to time 2.
Changes in Behavior
Teachers were blinded to case subject and control subject status. The RBPC that they completed has six subscales; all the subscales produced similar results. For simplicity, only the results of the conduct disorder scale are presented, which measures, among other things, aggressive behavior. A high score indicates worse behavior. At baseline, the case subjects and control subjects had comparable scores (57.3 ± 13.7 vs 58.4 ± 11.8; P ≤ .79). However, after the peer-mentoring intervention, there was a statistically significant difference between the scores of the two groups (56.9 ± 12.0 vs 69.9 ± 10.2; P ≤ .0001), reflecting an increase in the scores of the control subjects, ie, their behavior worsened. However, this result must be interpreted with caution, because there was a limited response from the teachers in completing the RBPC with very modest continuity of children for whom forms were received between time 0 and time 2; only 14 case subjects and 6 control subjects were measured at both times.
Because DOVVE contains questions that examine both exposure to violence and attitudes toward violence and there is no reason to suspect that this intervention would have any effect on exposure to violence, an analysis was performed that eliminated the 5 exposure questions and examined only the 10 questions that explored attitudes toward violence. This analysis paralleled that for the DOVVE-15; at baseline, the case subject and control subject scores were not statistically different (case score: 3.1 ± 2.7; control score: 2.9 ± 2.3; P ≤ .71) in their attitudes toward violence and this remained true at midstudy (case score: 2.7 ± 2.1; control score: 3.2 ± 2.4). At the end of the study, the case subjects and the control subjects were statistically different in their responses, with the case scores being lower, ie, better (2.6 ± 1.8 vs 3.8 ± 2.4; P ≤ .04).
The relationship between the number of interventions received and the scores of the participants on the DOVVE and NOBAGS was also examined. To examine this relationship, participants were divided into two groups, those who received 5 or more interventions and those who received fewer than 5 interventions. The DOVVE scores changed significantly between time 0 and time 2 (P ≤ .03); however, it must be noted that only 3 participants received fewer than 5 lessons. There was no statistically significant change noted in NOBAGS scores between time 0 and time 2 between the two groups.
Because there was significant attrition among the case subjects, baseline DOVVE scores were examined between those lost to follow-up and those who completed the intervention. There was no statistical difference between the two groups (lost to follow-up: 4.8 ± 1.8; finished study: 4.4 ± 3.1; P ≤ .73). Baseline NOBAGS scores were examined in a similar fashion. As with DOVVE, there was no statistical difference between the two groups (lost to follow-up: 1.64 ± .59; finished study: 1.69 ± .58;P ≤ .81). Also, because of attrition, baseline RBPC scores were compared between those lost to follow-up and those who completed the intervention. No difference was noted between those who were lost to follow-up and those who completed the intervention (57.8 ± 17.5 vs 56.4 ± 13.2; P ≤ .84).
Longitudinal Study of Adolescent Mentors
The adolescents' mean score on the DOVVE survey decreased from 6.5 ± 2.9 to 5.1 ± 3.8 (P ≤ .12) over the period of the study. The change was not statistically significant, probably because of small sample size; 19 adolescents completed two assessments and only 12 adolescents completed all three data collection points.
There also was no statistically significant difference between the mean Piers-Harris Self-Concept Scale scores at baseline and at the end of the study (62.9 ± 11.1 vs 65 ± 9.7; P ≤ .7).
However, the study research assistants noted that over the 18 months that the adolescents served as peer mentors, the adolescents developed leadership abilities, became more self-confident, and learned skills that they could use for future employment. One year after the end of the study, of the 19 adolescents, 11 are active in the peer-mentoring program, 2 are in college, 1 is working full-time, and 5 (26%) have dropped-out of the CGYP.
Study Findings and Significance
The results of this study indicate that, compared with matched control children, school-aged children who received the CTC peer-mentoring lessons avoided an increase in attitudes that support violence and may also have avoided an escalation of aggressive behaviors. Such changes could lead to reduced injury and jail time in the future. The small number of mentors made it difficult to measure change in them, yet they demonstrated a suggestive decrease in their violence-related attitudes and displayed increased self-esteem to the research assistants.
These results need to be understood in the broad context of violent injury prevention efforts. The prevention of violence must be addressed with two components. One component addresses the factors that contribute to the incidence of violent interactions, such as society and community problems. Variables that affect the incidence of violence include the media,16 ,17 poverty, crowded housing, residential mobility, and weak local social structure, eg, low participation in community life, poor schools, and poor coping repertoire.18 The second approach addresses those factors that contribute to the lethality of violence, particularly the easy availability of firearms (especially handguns). Prevention of violence will need to involve specific interventions; simultaneous efforts in decreasing the incidence and the lethality of violence would be expected to have the greatest effect. This study evaluates an intervention that works to decrease the incidence of violence by improving youth coping skills through the use of community mentors who model positive behaviors.
Changes in morbidity and mortality are the most robust outcome measures of a violent injury prevention program. Assessment of attitudinal change is the least sensitive indicator, because it does not guarantee a reduced injury rate.19 However, in a longer process of change, attitudinal change may be the first thing altered and can lead to behavioral change, which, in turn, may lead to fewer injuries. Indeed, for violence, attitude changes are probably necessary, although not sufficient, to reduce injury morbidity and mortality.
We chose to focus our evaluation on younger children, because we wanted to try to prevent violence (primary prevention). We thought intervening before the children entered the age in which they are likely to get heavily involved in violence was preferable to intervening after the fact. However, injury-based evaluation for this age group is difficult, because violent incidence rates are still low. As a result, our 18-month study was too small and too short to have the statistical power to show a difference in the violent injury rates of our mentees.
It was not easy to know the best way to measure the impact of the peer-mentoring intervention. Few tools suitable for this group of children who made up the case subjects and control subjects were evaluated at the time the study was designed.20 We developed and evaluated the reliability of a new tool, DOVVE, to measure exposure to violent incidents and attitudes toward violence that were not assessed by NOBAGS. RBPC, which has been used widely in many studies, was used by the teachers to evaluate school behavior. The utility of this tool was limited by our difficulty in getting teachers to complete the forms.
Because the study used an existing program, the CGYP, there were no costs associated with starting CTC as a new program. The largest CTC expense was the stipend for the adolescents, $4.50/hour with a $.50/hour raise yearly. The approximate cost for our 19 adolescents to work with 50 CGYP children for 48 weeks at $4.50 an hour was $20 000. This is only slightly above the average charge of one pediatric gun injury hospitalization in Illinois.21
Implications for Future Research
Violence prevention efforts must target the communities that need these services the most. However, evaluation of these efforts can be difficult. In Cabrini Green, children tend to move frequently among family members, making it difficult to obtain follow-up surveys. During the period of the study, mobility was increased, because the City of Chicago began to tear down high-rise buildings in Cabrini Green (to replace them with low-rise housing). This led to relocation of families to different areas of Chicago and led to additional study attrition.
Prevention efforts in at-risk communities can be dangerous for staff. Several times over the course of this study, data collection was interrupted because of increased gang violence activity, including increased gunfire. During one visit to a control child's apartment, one of the research assistants found a dead body in the building's lobby.
It took over 1 year of the study to build a relationship with the teachers (in addition to adding a small monetary incentive) before many teachers were interested in participating in the study by completing the forms.
Despite problems in conducting the study and its limitations, the results presented are encouraging. Therefore, this peer-mentoring approach warrants replication. A multicenter study that is called for by these results will need to be designed with the real world constraints of community-based work in mind. The constraints must not prevent continuing efforts to assess how the limited program resources should be applied to the greatest benefit of the children they serve.
This research was supported by Grant MCH-174002 from the Emergency Medical Services for Children Program (Washington, DC).
We appreciate the statistical analysis support provided by Hsiao-Ting Chiu, MS and Edwin Chen, PhD. We also thank the children and families who live in Cabrini Green, their teachers and counselors, and the Cabrini Green Youth Program staff.
- Received April 27, 1998.
- Accepted December 13, 1998.
Reprint requests to (K.S.) Children's Memorial Hospital, 2300 Children's Plaza No 62, Chicago, IL 60614. E-mail:
- ↵Dowd MD. Consequences of violence: premature death, violence recidivism, and violent criminality. In: Hennes HM, Calhoun AD, eds. The Pediatric Clinics of North America: Violence Among Children and Adolescents. Philadelphia, PA: WB Saunders Co; 1998;45:333–340
- Webster DW
- ↵Gabriel RM, Hopson T, Haskins M, Powell KE. Building relationships and resilience in the prevention of youth violence. Am J of Prev Med. 1996;12:5:48–55. Supplement
- ↵Kelder SH, Orpinas P, McAlister A, et al. The Students for Peace project: a comprehensive violence-prevention program for middle-school students. Am J Prev Med. 1996;12:5:22–30. Supplement
- ↵Wiist WH, Jackson RH, Jackson KW. Peer and community leader education to prevent youth violence. Am J Prev Med. 1996;12:5:56–64. Supplement
- ↵US Bureau of the Census. 1990 Census of Population and Housing: Summary Tapes 1A and 3A. Washington, DC: US Bureau of the Census; 1990
- Sanguino S,
- Marcelle D,
- Sheehan K,
- DiCara J,
- LeBailly S,
- Christoffel K,
- ↵Huesmann LR, Guerra NG, Miller L, Zelli A. The Normative Beliefs About Aggression Scale. Chicago, IL: University of Illinois at Chicago; 1989
- ↵Quay HC, Peterson DR. Revised Behavior Problem Checklist. Odessa, FL: Psychological Assessment Resources, Inc; 1983
- ↵Piers EV, Harris DB. The Piers-Harris Children's Self-Concept Scale. Los Angeles, CA: Western Psychological Services; 1969
- ↵Willis E, Strasburger VC. Media violence. In: Hennes HM, Calhoun AD, eds. The Pediatric Clinics of North America: Violence Among Children and Adolescents. Philadelphia, PA: WB Saunders Co; 1998;45:319–331
- ↵Center for Communication and Social Policy, University of California, Santa Barbara. National Television Violence Study, II. Thousand Oaks, CA: Sage; 1998
- ↵Reiss AJ, Roth JA. Understanding and Preventing Violence. National Research Council. Washington, DC: National Academy Press; 1993
- ↵The National Committee for Injury Prevention and Control. Injury Prevention: Meeting the Challenge. New York, NY: Oxford University Press; 1989;5:70
- ↵Centers for Disease Control and Prevention. Violence Prevention Evaluation Workshop. The evaluation of youth violence interventions: measurement of knowledge, attitudes, psychological and behavioral factors. Atlanta, GA; 1994
- ↵Violent Injury Prevention Center. Children's Memorial Medical Center. Mortality in Chicago, 1980–1994: Chicago Department of Public Health in Data and Policy Center Report. Chicago, IL: Children's Memorial Medical Center; 1995
- Copyright © 1999 American Academy of Pediatrics