Purpose: To develop and test a scale to measure children's and adolescents' recent exposure to violence in three specific domains: at home, school, and in the neighborhood.

Methods: Recent exposure to violence was assessed by a 26-item scale (R-VES). This self-administered scale was designed to measure specific acts of violence that occurred within the past year: threats, slapping/hitting/punching, beatings, sexual abuse, and knife/gun-related violence. For the first three types of violence, separate items were included to capture the site where the violence occurred: at home, at school, or in the neighborhood. Reports on sexual abuse and knife/gun violence were not site specific. Respondents were asked to report on both violence they had experienced directly, and violence they had personally witnessed over the past year. A four-point Likert scale ranging from "never" (a score of 0), "sometimes" (1), "often" (2), to "almost every day" (3), was used to assess the frequency of exposure to each type of violence. The scale was administered as an anonymous self-report questionnaire to two cohorts of youth: high school students (grades 9-12; N=3,745) and elementary and middle school students (grades 3-8; N=2,411) during usual school hours.

Results: To examine the factor structure of the R-VES, the 26 items were first analyzed by four exploratory factor models. Each of these models applied a different combination of extraction and rotation methods. All models revealed that a six-factor solution was appropriate, with a total variance explained by the six factors of approximately 56%. Using factor loadings .35 or larger as an indication of the grouping of related items, we found that items loading on four of the six factors were consistent across all models. However, the patterns for two items were not clear (slapped/hit/punched in neighborhood; threatened in neighborhood), and thus the exploratory factor analysis and content examination suggested two conceptual models. To see which model would best fit the data, we ran confirmatory factor analysis. All indices of goodness-of-fit (c², GFI, and the incremental fit index) indicated that one model (Model 1) was slightly better than the other. We next focused on the factor-factor relationships and the item-factor relationships, within the global factor pattern of Model 1. A nested model test revealed that: the six factors were correlated with each other and the correlated-factor model better fitted the data better than the independent-factor model. The final R-VES model (correlated factors and near congeneric) reached a satisfactory fit: its goodness-of-fit and incremental fit indices both exceeded .90. All factor loadings were .85 or higher, indicating that all item-factor relationships were strong. All scales reached an acceptable level of reliability (alpha= .72-.80) except for the Sexual Abuse scale (.51).

Implications for practice: The importance of careful screening and assessment of children/adolescents for community violence exposure is now well recognized since the relationship between violence exposure and numerous psychological, emotional and behavioral sequelae has been documented by empirical studies. The use of this non-proprietary scale by clinicians and researchers could aid our understanding and treatment of violence-exposed youth.