Developing a Practical Forecasting Screener for Domestic Violence Incidents

• NCJ Number: 223251
• Journal: Evaluation Review: A Journal of Applied Social Research Volume: 29 Issue: 4 Dated: August 2005 Pages: 358-383
• Author(s): Richard A. Berk; Yan He; Susan B. Sorenson
• Date Published: August 2005
• Length: 26 pages
• Publication Format: Article
• Publication Type: Report (Study/Research)
• Language: English
• Country: United States of America

Annotation

This article reports on the development of a short screening tool that deputies in the Los Angeles Sheriff's Department can use in the field in forecasting domestic violence incidents in particular households.

Abstract

The screening tool developed proved to be useful in forecasting future calls to police about domestic violence. The tool accurately forecasted future calls approximately 60 percent of the time and accurately forecasted the absence of domestic violence calls nearly 50 percent of the time. The tool achieved the results with just four predictors from a larger screening instrument. The four predictors are whether the victim reports that there had been more than three previous calls for the police to come to the household; whether the perpetrator is reported to have damaged household property when angry; whether the perpetrator is reported to be unemployed; and whether the perpetrator is reported to have threatened the life of the victim or someone in the victim's family in the past. The screening tool's performance was not significantly improved by adding more predictors. After discussing the caveats for the short screening tool, the authors advise that for households prone to domestic violence, it is possible to develop quick-response threat assessment instruments that can be used successfully in the field by police officers. Such instruments can take into account the relative costs of forecasting errors and then do better than a one-size-fits-all prediction. Also, such instruments can be rigorously evaluated. The data for developing and testing the short screening tool came from just over 500 households to which sheriff's deputies were dispatched in fall 2003. Information on potential predictors was collected at the scene. Outcomes were measured during a 3-month followup. Data were analyzed with modern data-mining procedures in which true forecasts were evaluated. 8 figures, 3 tables, 17 references, and appended long screening instrument