Abstract
Research on risk assessment for domestic abuse has focused primarily on the predictive validity of specific tools; less attention has been paid to implementation of risk tools by practitioners. This paper presents findings from a mixed methods study in England and Wales. Multi-level modelling reveals an 'officer effect' whereby victims' responses to the Domestic Abuse, Stalking and Harassment and Honour-Based Violence (DASH) risk assessment are influenced by the specific officer that completes the assessment. Specifically, this officer effect is strongest in relation to questions intended to capture elements of controlling and coercive behaviour, and least apparent in relation to identifying physical injuries. We further present findings from field observations and interviews with first response officers that corroborate and help explain the officer effect. We discuss implications for the design of primary risk assessments, victim safeguarding, and the use of police data for predictive modelling.