Stalking recidivism and risk assessment

Scientific interest exploded after the crime of stalking was created in the early nineties. It is now generally recognised that stalking can result in significant disruption, distress and, for some victims, physical harm or even death. It is thus noteworthy that, while stalking may be distinctively protracted in nature and features well-established links with violence, very little research has been devoted to the issue of repeat offending (recidivism) among stalkers. Nevertheless, the likelihood of continued or repeat offending remains a key concern for forensic clinicians, courts, police and legislators, who are ultimately responsible for minimising the incidence and impact of stalking. As such, it is important to devise appropriate methods by which this risk can be evaluated and managed. Given the paucity of studies specifically examining recidivism among stalkers, existing findings are in need of both replication and expansion. Accordingly, this research project sought to identify the demographic, psychiatric, and offending-related variables related to recidivism among stalkers. The project also set out to evaluate the two extant stalking-specific risk assessment tools with respect to recidivism. A case-linkage methodology was employed to combine data drawn from state-wide criminal records and forensic mental health records. Participants were clients of a community-based forensic mental health service who were originally recruited as part of other research projects. Each had presented to the clinic for stalking behaviour, either between 2002 and 2007, or between 2010 and 2011. When the index referral related to multiple cases of stalking, each case was evaluated separately. Consequently, analyses were conducted by case (n = 163) rather than participant (n = 146). Offending follow-up data were unavailable for some cases (N = 15); hence, these were excluded from analyses explicitly relating to recidivism outcomes. Altogether, this research project comprised three related studies. The first examined the demographic, historical, clinical and offence-specific variables of stalkers. Analyses were conducted to identify which combinations of these factors best predicted stalking, violent and general recidivism among stalkers. All cases for which recidivism data were available were evaluated (N = 148). Overall, stalking recidivism was best predicted by a model that included traits of personality disorder, criminal history, the presence of an Axis I disorder, index stalking duration of no less than two weeks, and the use of proximity-seeking approach behaviours during the index stalking episode. When considering only the original victim, all of these factors except criminal history remained valuable predictors; however, the combination of criminal history, prior stalking and having stalked a stranger during the index episode was most predictive of stalking recidivism against a new victim. General recidivism was best predicted by a combination of criminal history, stranger stalking victim, Axis I disorder, and traits of personality disorder. Using a broad definition of violence that included stalking and threats, violent recidivism was best predicted by traits of personality disorder, criminal history and Axis I disorder. When this was restricted to actual or attempted physical violence, recidivism was best predicted by the use of diverse stalking behaviours during the index episode and the absence of psychotic disorder. While the first study sought to replicate and extend findings from past research on stalking recidivism, the second and third evaluated the properties, predictive and construct-related validity of the two existing stalking-specific risk assessment tools, the Guidelines for Stalking Assessment and Management (SAM: Kropp, Hart, & Lyon, 2008) and the Stalking Risk Profile (SRP: MacKenzie et al., 2009). While the entire sample (n = 163) was used for most analyses, only those for whom offending data were available (N = 148) were included in predictive validity analyses. Inter-rater reliability (IRR) for the SAM was generally moderate, although this varied depending on the level at which this was evaluated (e.g., conclusory opinions, domain scores or item ratings). Nature of Stalking (N) and Perpetrator Risk (P) domain scores were highly internally consistent, and Lifetime total, domain and item scores were, by and large, modestly related to conclusory opinions for Case Prioritisation, Future Stalking and Serious Physical Harm. Respectively, Case Prioritisation and Future Stalking ratings were modestly and highly accurate for predicting later stalking of index victims. Lifetime total and N domain scores also featured good predictive accuracy, while P scores were only modestly predictive. With respect to construct validity, SAM conclusory opinions were all related to SRP risk ratings in theoretically predictable ways. Like the SAM, IRR for the SRP varied based on the level of evaluation. Agreement was substantial for assignment to motivational typology, Violence and Persistence risk judgements, and moderate for the remaining risk ratings. Agreement was moderate-to-substantial for domain scores, and ranged between fair and substantial for most items. As expected, due to the intended independence of risk factors, internal consistency was generally modest for SRP domain scores. Typically, domain and item scores were associated with the corresponding risk ratings for all stalker types. Risk ratings for stalking Persistence, Recurrence against index victims and Recurrence involving new victims all featured good predictive accuracy. Interestingly, the discriminatory performance of numerically derived domain scores (versus clinicians’ risk judgements) did not differ from chance for any domain or stalker type. Taken as a whole, this research supports the notion that stalkers represent a diverse population, one that poses particular challenges for adequate risk assessment. Indeed, it is clear that an adequate evaluation of stalking-related risks must consider not only past offending, psychopathology and social context, but also specific stalking behaviours and victim selection. Of particular note is the finding that stalking recidivism is associated with somewhat different factors depending on whether the stalker targets their original victim or another. While findings provide support for the utility of the SAM and SRP, it is suggested that both tools would benefit from further validation research and ongoing refinement. Implications of the current results for clinicians, police and legislators are discussed.