Background: Stalking is often viewed as a precursor to violence, but determining which stalkers might attack is a difficult task. This study overcomes shortfalls in previous investigations by adopting a pseudo-prospective design and examining potential risk factors for different types of stalker.
Method: Demographic, behavioural and diagnostic information was collected from stalkers referred to a community forensic mental health service (n=211). Potential risk factors for stalking violence were identified using odds ratios and chi2 tests, and entered into logistic regression models. Model utility was assessed using receiver operating characteristic curves.
Results: Amongst Rejected ex-intimate stalkers, violence was best predicted by previous violence, making threats and being employed (area under the curve=0.75), while for stalkers with other motives and relationships to the victim, being aged less than 30 years, substance use at the time of stalking and prior violence best predicted stalking violence (area under the curve=0.80).
Conclusions: Stalkers at increased risk of violence can be accurately identified by examining motivational and relationship type in conjunction with specific relevant risk factors. Previous violence is a particularly important risk factor, as are threats amongst ex-intimate stalkers. Approach behaviours and psychosis were shown to be less useful in predicting violence.