Article Text

other Versions


Street conflict mediation to prevent youth violence: Conflict characteristics and outcomes
  1. Jennifer M Whitehill1,2,
  2. Daniel W Webster2,
  3. Jon S Vernick2
  1. 1Harborview Injury Prevention and Research Center, University of Washington, Seattle, Washington, USA
  2. 2Johns Hopkins Center for the Prevention of Youth Violence, Johns Hopkins Bloomberg School of Public Health, Baltimore, Maryland, USA
  1. Correspondence to Dr Jennifer M Whitehill, Harborview Injury Prevention and Research Center, University of Washington, 325 Ninth Avenue, Box 359960, Seattle, WA 98014, USA; jenwh{at}


Background Mediation of potentially violent conflicts is a key component of CeaseFire, an effective gun violence-prevention programme.

Objective To describe conflicts mediated by outreach workers (OW) in Baltimore's CeaseFire replication, examine neighbourhood variation, and measure associations between conflict risk factors and successful nonviolent resolution.

Methods A cross-sectional study was conducted using records for 158 conflicts mediated between 2007 and 2009. Involvement of youth, gangs, retaliation, weapons and other risk factors were described. Principal component analysis (PCA) was used for data-reduction purposes before the relationship between conflict risk components and mediation success was assessed with multivariate logistic regression.

Results Most conflicts involved 2–3 individuals. Youth, persons with a history of violence, gang members and weapons were commonly present. OWs reported immediate, nonviolent resolution for 65% of mediated conflicts; an additional 23% were at least temporarily resolved without violence. PCA identified four dimensions of conflict risk: the risk-level of individuals involved; whether the incident was related to retaliation; the number of people involved; and shooting likelihood. However, these factors were not related to the OW's ability to resolve the conflict. Neighbourhoods with programme-associated reductions in homicides mediated more gang-related conflicts; neighbourhoods without programme-related homicide reductions encountered more retaliatory conflicts and more weapons.

This is an open-access article distributed under the terms of the Creative Commons Attribution Non-commercial License, which permits use, distribution, and reproduction in any medium, provided the original work is properly cited, the use is non commercial and is otherwise in compliance with the license. See: and

Statistics from


Gun violence is a serious threat to health and safety in many communities, especially low-income urban areas.1 Youth are disproportionately involved as both victims and perpetrators of violent crimes.2–4 In the USA, homicide is the leading cause of death for black youth between the ages of 15 and 24, and the second leading cause of death for Hispanic youth in this age group.5 Violent neighbourhoods’ social norms, or the ‘code of the street’, often dictate the use of violence when an individual is threatened or disrespected.6 This environment fosters cycles of retaliatory gun violence and repeated injury.7–9

CeaseFire is a public health programme designed to prevent shootings by changing attitudes, behaviours, and social norms about gun violence.10 The model involves: case-management and mentoring for high-risk youth; interruption of violence through conflict mediation; community mobilisation; public education; and partnership with clergy. The programme maintains a careful relationship with law enforcement, balancing the need for police information about shootings with the need for outreach staff to avoid being perceived by the community as untrustworthy police informants or ‘snitches’, a label that could jeopardise their own safety and their ability to learn of future conflicts before they happen.11 An evaluation of CeaseFire Chicago found programme-related reductions in shootings from 16% to 35% in four out of seven neighbourhoods where it was implemented.11 The Safe Streets programme, a replication of CeaseFire run by the Baltimore City Health Department (BCHD) in four of Baltimore's most violent neighbourhoods, was associated with reductions in gun violence in three of these neighbourhoods, including substantial reductions in homicides in the two neighbourhoods where the programme was fully implemented (McElderry Park and Cherry Hill).12

Of the CeaseFire model components, the most direct way of preventing shootings is the use of specialised outreach workers (OW) to identify and intervene in street conflicts to help resolve them peacefully. The OWs, sometimes called violence interrupters, are recruited for their credibility and influence among high-risk youth in the target communities. They are often former gang members or were previously involved in the drug trade.11 In Baltimore, evaluators noted that the more effective programme sites had three times as many conflict mediations per month compared with programme sites that failed to reduce homicides.12

Communities, in the USA and internationally, are increasingly using conflict mediation as part of their violence-prevention efforts.11–16 It is important to determine the combinations of settings, individuals and risk factors in which mediation can be used effectively. As part of our evaluation of Baltimore's Safe Streets programme, we sought to describe the characteristics of conflicts mediated by OWs, and how they varied across programme neighbourhoods. We also examined associations between conflict risk factors and mediation outcomes in an effort to identify whether certain conflict risk factors are associated with successful mediation.


Study sites

Four Safe Streets neighbourhoods in Baltimore were included in this analysis. McElderry Park was the first neighbourhood to implement the programme, beginning in June 2007. Later, additional outreach staff were hired to expand this site into two adjacent neighbourhoods—Ellwood Park (added in February 2008) and Madison-Eastend (added in December 2008). Another site began operation in the South Baltimore neighbourhood of Cherry Hill in November 2008. Each neighbourhood was assigned a staff of four full-time equivalent OWs. Table 1 shows demographic characteristics and crime statistics for the four Safe Streets neighbourhoods, and Baltimore as a whole.

Table 1

Safe streets site characteristics

Data collection

The Safe Streets sites routinely collect information about mediations for programme monitoring purposes using a tool developed by CeaseFire Chicago called the Conflict Mediation Form (CMF). The CMF is typically completed within 24 h of the mediation. BCHD provided the research team with de-identified copies of all CMFs completed between the programme's inception in June 2007 and December 2009.

A dataset was constructed by abstracting information from the CMFs. Variables abstracted included: neighbourhood; conflict date; mediation date; number of OW(s) mediating conflict; number of people involved in the conflict; the way(s) in which Safe Streets learned about the conflict; and reasons for the conflict (gang, personal altercation, narcotics, domestic violence, child abuse, robbery, retaliation, unknown or other). Also recorded were seven risk factors indicating whether one or more parties was thought to: (1) be gang/crew involved; (2) have a history of violence; (3) be between ages 14 and 25 years; (4) have recently been released from prison; or if the situation (5) was related to high risk street activity (ie, drug-selling); (6) was in retaliation for a shooting in the past 90 days; or (7) involved a weapon. The type(s) of mediation conducted (phone, one-on-one, small group, diplomacy between groups, and/or conducted by a third party who was not an OW); the OW's perceived likelihood that the conflict, unmediated, would have led to a shooting; and the perceived outcome of the mediation (conflict resolved, resolved temporarily, ongoing or unknown) were also abstracted. A conflict was considered ‘resolved’ if, at the time the CMF was completed, the OWs believed the situation was fully resolved and would lead to no further violence. ‘Resolved temporarily’ meant immediate violence had been avoided but the parties had not reached a final agreement to avoid future violence related to this incident.

The form also contained an open-ended section in which OWs describe their actions. A majority of CMFs contained information in this field addressing the cause of the conflict. One author (JMW) read this section and coded a variable for the conflict trigger. This allowed us to identify the specific causes of conflicts and gain a deeper understanding of the meaning behind the closed-ended categories ‘personal altercation’ and ‘retaliation’. A codebook was compiled and reviewed by a second author (DWW) to verify that codes were appropriate and fit with prior research on causes of street conflicts. The list of all codes was reviewed and condensed, forming categories into which similar codes could be grouped.19 From the text, we also coded binary variables indicating whether alcohol was involved, whether a Safe Streets case-management client was involved, and the sex of the people involved.

Data analysis

Descriptive statistics were used to assess frequencies and proportions for binary and categorical variables within the four Safe Streets neighbourhoods, and for the programme as a whole. A χ2-test was used to examine the relationship between gang-motivated conflict and having more than 10 people involved.

We examined whether certain conflict risk factors might be more important than others for predicting a successful mediation. Because there might be close correlations between some of the risk factors, we conducted a principal component analysis (PCA) of 10 risk factors (the seven conflict risk factors described above, an indicator variable for any mention of retaliation, the number involved in conflict, and the likelihood of it leading to a shooting) using varimax rotation.20 This data-reduction tool allowed us to use only a few variable clusters (components) to explain the maximum amount of variance in the risk factors. The Kaiser–Meyer–Olkin (KMO) measure of sampling adequacy was used to determine whether the correlation was high enough to warrant PCA. KMO=0.60 was used as the criterion.21 ,22

We dichotomised the outcome variable to reflect whether or not the OWs considered the conflict fully resolved without serious violence. This variable was included in a multiple logistic regression model to examine the relationship between the risk factor components derived from the PCA and mediation outcome. Because two neighbourhoods had relatively small numbers of mediations, we pooled the data for the regression analysis. Statistical analysis was conducted using Stata 10.23 The Institutional Review Board at the Johns Hopkins Bloomberg School of Public Health approved this study.


There were 167 conflicts mediated during the study period; CMFs were available for 158 (95%) of these conflicts. Outreach staff in Cherry Hill and McElderry Park mediated the most conflicts, averaging 3 and 2.5 per month, respectively, compared with an average of one conflict mediation per month in Ellwood Park and Madison-Eastend. Across all the neighbourhoods, a majority of conflicts (54%) involved two or three individuals. The median number involved was four in Ellwood Park, three in McElderry Park and Cherry Hill, and two in Madison-Eastend. Two-thirds (68%) of conflicts were mediated by one OW. Table 2 shows the frequency and the number of individuals involved in the conflicts. No clear pattern was observed in the correlation between the number of individuals in conflict and the number of OW mediators. A large majority (91%) of conflicts required only one mediation session.

Table 2

Frequency and size of conflicts mediated by safe streets outreach staff

Reasons for conflict

The issues that precipitated conflicts did not vary significantly across neighbourhoods, and thus we grouped the data for this analysis. In the closed-ended section, a ‘personal altercation’ was noted as the reason for 65% of conflicts. Other reasons for conflicts included gangs (29%), retaliation (15%), narcotics (11%), and domestic violence (10%) (table 3). The percentage of gang-motivated conflicts with more than 10 individuals (37%) was significantly greater than the percentage of non-gang conflicts of that size (18%) (χ2 = 6.63; p value=0.010).

Table 3

Reasons for conflicts

The open-ended section provided additional details on conflict triggers for 105 conflicts (67%). Within this group, there were 57 cases for which personal altercation was listed as a reason and a specific trigger was identified; the most common triggers described included disrespect (19%), disputes over territory (16%), theft (14%) and disagreements about money (13%). In 14 cases of retaliation where specific triggers were described, the conflicts were related to disputes over territory (28%), money (21%) or physical confrontations when at least one party was punched, hit, kicked or assaulted in some manner (21%).

Risk factors and people involved

Nine out of 10 conflicts involved at least one individual between ages 15 and 24 years (89%), and at least one with a history of violence (90%); nearly half (47%) involved someone who was recently released from prison. Almost 60% of conflicts were related to drug selling. The proportion of conflicts involving gang members was notably lower in Madison-Eastend than in the other neighbourhoods. Many more conflicts in Ellwood Park involved a weapon (81%) compared with the other neighbourhoods where 50%–61% of the conflicts involved weapons. Conflicts in Ellwood Park and Madison-Eastend were more frequently related to retaliation than in the other neighbourhoods (see table 4).

Table 4

Conflict characteristics by programme neighborhood

OWs reported that, without their intervention, a shooting was ‘very likely’ in 57% of the incidents. Coding of the open-ended field revealed that nearly 20% of conflicts involved both women and men, and 7% involved only women. The proportion of conflicts involving a Safe Streets programme client—a high-risk youth whom OWs mentored and directed to services—varied across the programme sites, accounting for 16% of conflicts in McElderry Park and 5%–8% in the other neighbourhoods.

There was a ‘middling’ level of sampling adequacy for the PCA (KMO=0.73) based on Kaiser's benchmarks.21 Four factors together explained 64% of the total variance. These factors were comprised of: (1) five items measuring the general risk level of the individuals in conflict (age 15–24 years, gang involvement, history of violence, recent release from prison and high-risk street activity), (2) two items indicating whether the conflict was related to retaliation, (3) two items reflecting the likelihood of a shooting if the conflict is not mediated (OW's assessment and the presence of a weapon) and (4) one item indicating the number of people involved (table 5). Based on the results of the PCA, three factor score variables were created for the first three factors above and included for use in subsequent analyses. A fourth variable measured the number of people involved in the conflict.

Table 5

Principal components of conflict risk

Outcomes and associations

OWs were usually able to resolve the conflicts in which they intervened, thus avoiding violence. A higher percentage of conflicts were reported as fully resolved in McElderry Park (71%) and Madison-Eastend (93%) as compared with Ellwood Park (57%) and Cherry Hill (49%). In Cherry Hill and Ellwood Park, approximately one-third of conflicts were reported to be temporarily resolved, meaning that violence was prevented but OWs were uncertain about whether conflict might re-erupt. Overall, OWs were able to fully resolve 65% of the conflicts without serious violence and bring about a temporary resolution in 23%. Only 3% of conflicts remained ongoing after the mediation attempt (table 4).

None of the PCA components showed a statistically significant relationship with the OWs’ ability to fully resolve the conflict (table 6). The point estimate was largest for the impact of individual risk (OR=0.79, p=0.137), though it was not statistically significant.

Table 6

Adjusted and unadjusted associations between risk score and successful conflict resolution*


To our knowledge, this is the first analysis of conflict mediation records from a street outreach programme. The results provide insight into the circumstances and people involved in situations that might turn deadly without intervention by skilled OWs. Most conflicts encountered by OWs were between two or three individuals. However, one in four conflicts involved more than 10 people. These larger conflicts were more likely to be gang-related than not. Youth, persons with a history of violence, persons recently released from prison, and weapons were common. Females were involved in nearly one-quarter of the conflicts. PCA reduced the risk factor items on the CMF to four dimensions of conflict risk: the risk-level of individuals involved, whether the conflict is related to retaliation, the number of people involved, and the likelihood of a shooting, but these factors did not predict mediation outcomes.

The role of gangs

Nonviolent resolutions of larger conflicts may have a greater impact on overall rates of violence in the community if the resolution includes a truce between gangs or groups.24 ,25 In this study, the number of people involved in the conflict was not associated with the OWs’ ability to resolve the conflict, suggesting it was not necessarily harder to mediate larger conflicts.

Neighbourhood variation

The Safe Streets neighbourhoods are demographically similar to each other; all have a lower median income, a higher unemployment rate, and a greater proportion of black residents than the Baltimore city-wide average. All were among the most violent neighbourhoods in the city based on the number of homicides and shootings in the years before the programme began,12 though Madison-Eastend had a higher number of shootings per capita. The variation in certain conflict risk factors across neighbourhoods may provide some insight into the mixed effects of the programme on homicides and shootings found in the Safe Streets evaluation.26 Gangs were involved in a smaller percentage of conflicts that were mediated in Madison-Eastend (50%) than in other neighbourhoods (70%–86%). An intense feud between gangs selling drugs precipitated an incident in which 12 people were shot at a cookout in Madison-Eastend in the summer of 2009.27 ,28 The feud allegedly involved kidnapping of a gang leader's family, murders and retaliatory violence during the period when Safe Streets was implemented. The intensity of this ongoing conflict may have made it difficult or impossible for OWs to mediate the very violent, gang-related conflicts in this area. Madison-Eastend and Ellwood Park experienced more violence during programme implementation than did McElderry Park and Cherry Hill. This may explain why roughly half the mediated conflicts in these neighbourhoods were related to retaliation, compared with approximately one-third of the conflicts in the other two neighbourhoods. Future studies should examine specific mediation strategies for addressing retaliation. Future attention is also needed to understand whether differences in programme implementation, such as staffing or community engagement, may have contributed to these observed differences in the types of conflicts mediated, and ultimately, the programme outcomes.

Conflict triggers and participants

The reasons for the conflicts in which Safe Streets OWs intervened are generally consistent with those identified in the literature as triggers for gun incidents: perceived disrespect, drug territory disputes and gang activities were among the most frequently cited.29 The reasons for conflicts and the risk factors for people involved were similar to those noted in the evaluation of CeaseFire Chicago.11

OWs indicated involvement of Safe Streets clients in a relatively small proportion (12%) of conflicts, though this figure should be seen as the lower bound of the true value because client involvement was ascertained from open-ended items and not directly asked on every CMF completed. The small number may reflect that the programme's case-management efforts are succeeding at keeping these youth out of trouble. Or, the programme may be failing to engage as clients the very highest-risk youth that end up in conflicts. The open-ended section of the CMF also provided some insight into the prevalence of females in street conflicts. Because young males are predominantly the victims and perpetrators of gun violence, descriptions of violent conflicts rarely focus on the involvement of women.2 ,5 ,30 ,31 The role of females in urban youth violence is an area that warrants future exploration.

Limitations and strengths

The process of mediating a street conflict involves assessing multiple factors about the situation, the actors involved, and the actions that will likely bring about a nonviolent resolution. The information on the CMFs does not represent the entirety of the interaction when reduced to check boxes and an open-ended section.11 For example, it was unknown to researchers what type of weapon was involved in each scenario, and whether certain high-risk street activities referred to drug-selling or, perhaps, something else.

The reliance on self-reported data creates the possibility of social desirability bias if OWs exaggerate the danger of conflict mediations and their ability to resolve them peacefully. However, there is no viable alternative to capturing data about these very risky street situations. All CMFs are reviewed both by the OW's direct supervisors and the programme overseers at BCHD; we believe this system helps foster accurate reporting.

The data presented here depict conflicts mediated in one city implementing the CeaseFire model and cannot necessarily be generalised to other locations. Additionally, this analysis was limited to reports from conflicts in which OWs intervened. Therefore, it is not able to shed light on conflicts in which a shooting was likely or had occurred, but in which OWs did not intervene.

Despite these limitations, the study results demonstrate previously unknown patterns in conflict mediations likely to be useful for programmes seeking to mediate conflicts involving urban youth. In Baltimore neighbourhoods with programme-related reductions in homicides, a high proportion of mediated conflicts involved gang members, suggesting that mediation of gang conflicts may be especially important for preventing lethal violence. Examination of conflict data for other cities implementing the programme is needed to facilitate comparison and advance our understanding of conflict mediation as an evidence-based violence-prevention strategy.

What is already known on the subject

  • An increasing number of communities are using street outreach, including violence interruption and conflict mediation, to address the problem of gun violence.

  • Conflict mediation is a key component of CeaseFire, an effective public health programme for gun violence prevention.

What this study adds

  • Conflicts mediated by CeaseFire-trained outreach workers (OWs) in Baltimore frequently involved youth, gangs, weapons and persons with a history of violence.

  • Despite this combination of risk factors, OWs were able to prevent gun violence in two-thirds of conflicts mediated.

  • More gang-related conflicts were mediated in neighbourhoods where homicides were reduced.


We gratefully acknowledge the Baltimore City Health Department staff who provided us with this data, and the outreach worker, T Carter, for his insights on how conflict data is recorded in the field. We also thank S Frattaroli, K Smith, P Leaf and F Rivara for their comments on the manuscript.


View Abstract


  • Contributors JM Whitehill, DW Webster and JS Vernick conceived of and designed the study. JM Whitehill and DW Webster acquired the data. JM Whitehill conducted the analyses and led the writing of the article. DW Webster and JS Vernick contributed data interpretation and critical manuscript revisions. All authors gave final approval of the version to be published.

  • Funding This work was supported, in part, by the Centers for Disease Control and Prevention grant number U49CE000728.

  • Competing interests None.

  • Provenance and peer review Not commissioned; externally peer reviewed.

  • Data sharing statement The study database was compiled by the authors from data maintained by Baltimore City Health Department (BCHD). It cannot be shared by the authors with anyone without approval from BCHD.

Request permissions

If you wish to reuse any or all of this article please use the link below which will take you to the Copyright Clearance Center’s RightsLink service. You will be able to get a quick price and instant permission to reuse the content in many different ways.