Type:	Journal Article
Title:	Developing a practical forecasting screener for domestic violence incidents
Other Titles:	Evaluation review : a journal of applied social research
Authors:	He, Yan Berk, Richard A Sorenson, Susan B
Keywords:	Policy;Screening
Year:	2005
Publisher:	Sage Publications
Notes:	This article reports on the development of a short screening tool that deputies in the US Los Angeles Sheriff’s Department could use in the field to help predict future domestic violence incidents and their seriousness. Forecasts of domestic violence calls for particular households could assist police at the scene to decide what action to take. Current instruments on threat assessment were found to be too long, and some were not designed as predictive instruments. A very short forecasting screener was needed for sheriff’s deputies to use in the field. Data came from over 500 households to which sheriff’s deputies had been dispatched. Information on potential predictors was collected at the scene. Outcomes were measured during a 3-month follow-up. Data mining procedures were used to analyse data for evaluating forecasts. Relative costs of false positives (incorrectly forecasting future calls) and false negatives (incorrectly forecasting no future calls) were taken into account and the tool correctly forecasted 60% of the time for future calls for service, and 50% of the time for future calls of domestic violence misdemeanours and felonies. It concludes that it is possible to forecast future calls for law enforcement assistance, and that this can be done with just 4 predictors from the larger screening instrument. The 4 predictors are: (a) whether the victim reports that there have been more than 3 police calls to the household before; (b) whether the perpetrator is reported to damage household property when angry; (c) whether the perpetrator is unemployed; and (d) whether the perpetrator is reported to have threatened the life of the victim or someone in the victim’s family in the past. For establishing subsequent probable cause that a domestic violence offence has occurred, only 3 predictors are needed. Based on the CART (classification and regression trees model), these 3 predictors of a new domestic violence offence, in order of increasing likelihood, are: (a) whether the police are reported to have been called in the past; (b) whether the perpetrator is reported to be unemployed; and (c) whether the violence is reported to be getting worse.
URI:	https://anrows.intersearch.com.au/anrowsjspui/handle/1/14166
ISSN:	0193-841X
Appears in Collections:	Journal Articles

Items in ANROWS library are protected by copyright, with all rights reserved, unless otherwise indicated.