RT Article
T1 Dashing hopes?: the predictive accuracy of domestic abuse risk assessment by police
JF The British journal of criminology
VO 59
IS 5
SP 1013
OP 1034
A1 Turner, Emily
A1 Medina, Juanjo
A1 Brown, Gavin
LA English
YR 2019
UL https://krimdok.uni-tuebingen.de/Record/1671911636
AB The Domestic Abuse, Stalking and Honour Based Violence (DASH) form is a standardized risk assessment implemented across most UK police forces. It is intended to facilitate an officer’s structured professional judgment about the risk a victim faces of serious harm at the hand of their abuser. Until now, it has been an open question whether this tool works in practice. Here, we present the largest scale European study, making the case that the risk assessment tool is underperforming. Each element of the DASH questionnaire is, at best, weakly predictive of revictimization. Officer risk predictions based on DASH are little better than random and a logistic regression model that predicts the same outcome using DASH only provides modest improvement in performance.
K1 Domestic  abuse
K1 Risk assessment
K1 Police
K1 Predictive policing
K1 Machine learning
K1 Häusliche Gewalt
K1 Polizeiarbeit
K1 Maschinelles Lernen
K1 Vorausschauende Polizeiarbeit
K1 Techniken
DO 10.1093/bjc/azy074