Tracking the Accuracy of Assessing High-Risk Offenders for Intimate Partner ViolenceA 10-Year AnalysisBy Acting Sergeant Maria Wright, London Police ServiceSince 2003, the police response to Intimate Partner Violence (IPV) in Ontario has evolved following two inquests involving IPV fatalities and the subsequent forming of the Ontario Domestic Violence Death Review Committee (ODVDRC). As a result, police have come under increased scrutiny to improve responses to IPV, especially lethal IPV. By 2012, the London Police Service’s IPV Unit wasprimarily using the Ontario Domestic Assault Risk Assessment (ODARA) to identify high-risk IPV offenders and provide assistance to victims.My research seeks to track the criminal behaviour of offenders that have been forecasted by police to be at risk of committing high-harm IPV in London, Ontario. It poses the question of whether or not risk assessments are valid or if there should be enhancements made that may help identify the “power few” (Sherman, 2007) high-harm offenders. The point of this research is not to challenge that the ODARA is a valid tool to predict recidivism. Rather, the purpose is to spark efforts to improve the accuracy of forecasting high-harm IPV and potentially save lives and reduce crime harm. This research illuminates the need for police services to carefully examine the accuracy of forecasting models and begin using police data more effectively to identify high-harm IPV in the name of public safety. Forecasted High-Risk Offenders After a 730-Day Follow Up Flagged High-Risk Offenders Charged Vs Not Charged with a Crime after 730 DaysResearch QuestionsKey Research Question: How accurate is the current process of predicting high-risk IPV offending in London, Ontario, in a list of 1,314 identified offenders?What other possible predictors could be used to enhance the accuracy of targeting high-risk IPV?Sub Questions:1. How were the 1,314 persons identified as high risk (HR) – by whom, with what pipeline of candidates to assess, with what assessment procedures?2. Among HR offenders, in the 730 days after their identification as HR, how many were :• Not charged for any crime category [false positive 1] • Not charged for IPV but for some other category [false positive 2] • Charged for IPV but not in a highharm category [false positive 3] • Charged for IPV in a high-harm category [true positives] • How many were flagged in each category as having suicidal tendencies?3. Among all perpetrators of the highest high-harm IPV in 2009-19, how many were not on the list of 1,314? How many of these individuals were flagged as having suicidal tendencies? (false negatives) 4. Among all 9,035 total perpetrators of IPV in 2009-2019, how many were not on the forecasted list of 1,314 offenders and did not commit a high-harm offence? (true negatives) FindingsThe safeguarding practices conducted by the IPV Unit at the London Police Service involves an assessment of every offender after they are charged with an IPV crime, primarily using the ODARA risk assessment tool. As a result of the risk assessments, 1,314 offenders were flagged as high risk between 2009 and 2019.In order to understand the accuracy of the risk assessments, a clear definition of “high risk” needed to be established. For the purposes of this study, 15 IPV crimes were identified based on the Canadian Crime Severity Index to cause the most harm. The post-flag offending was examined for a 730-day period for each of the 1,314 offenders that were classified as “high risk.” As a result of this research, it was found that 11 per cent of offenders that were forecasted as high risk went on to commit one or more of the 15 high-harm crimes within 730 days.In contrast, 35 per cent of offenders that were forecasted as high risk were not charged with any crimes in the 730-day follow-up period, as shown in Figure 1.Figure 2 illustrates that the current model of forecasting high-risk offenders accurately predicts general recidivism in 65 per cent of cases. Total Number of False Negative IPV High Risk Flags 2009-2019 Number of Non-High-Harm Offenders Flagged vs Not Flagged as High RiskIn order to provide a fulsome understanding of the forecasting model, this research examined the total number of offenders that committed one or more of the 15 identified high-harm crimes between 2009 and 2019. It was found that 2,398 of a total of 9,035 IPV offenders committed one or more high-harm IPV crimes between 2009 and 2019. Interestingly, 71 per cent of the 2,398 offenders were not flagged as high risk at any time between 2009 and 2019. In other words, the false negative rate is 71 per cent in forecasting high-harm IPV offenders. This finding is illustrated in Figure 3.Figure 4 indicates that 91 per cent of the total IPV offenders between 2009 and 2019 were true negatives, meaning they were correctly forecasted as not high risk and were not flagged.Recent research has advanced the theory of suicidal ideations as an indicator of lethal IPV offending. This study has considered whether or not a suicidal tendencies flag may assist in identifying the most harmful IPV offenders. Although the findings were retrospective, support for this theory was generated by the finding that over three times as many offenders that were charged with murder or attempted murder had a suicidal tendencies flag compared to offenders that did not commit any crimes after being forecasted as high risk. This finding should be considered with caution given that the total sample of offenders with a suicide flag was small, coupled with the fact that it is not clear if an offender received the flag prior to or after being charged with an IPV crime.DiscussionThis study brings into question the long-held notion that IPV involves a cycle of violence and ongoing repeat victimization (Walker, 1984). This is illustrated by the finding that 69 per cent of offenders predicted to be high risk were not charged with any further IPV crimes in 730 days after being identified. This finding is similar to other research in this area (Bland & Ariel, 2015; Chambers-McClellan, 2002; Sherman & Berk, 1984). This knowledge can be utilized by police services in understanding the value of using resources to focus on identifying the perpetrators that are likely to re-offend rather than anticipating that all offenders will re-offend.In terms of increasing frequency and escalation, many offenders (89 per cent) did not escalate in the sense of committing an identified high-harm crime, and 35 per cent were not charged with any crimes. It should be noted that the lack of escalation found in this study may be a result of the targeting impact that police and social services have on predicted high-risk offenders. This is due to the fact that all of the victims in this study received additional supports and outreach compared to the general population of IPV victims. Overall, these findings bring into question the theory of escalation and adds credence to the desistance literature by indicating that even offenders predicted to be the most dangerous desist in criminal offending.OVERALL, THESE FINDINGS BRING INTO QUESTION THE THEORY OF ESCALATION AND ADDS CREDENCE TO THE DESISTANCE LITERATURE BY INDICATING THAT EVEN OFFENDERS PREDICTED TO BE THE MOST DANGEROUS DESIST IN CRIMINAL OFFENDING.Policy ImplicationsSeveral relevant policy implications have been identified through this research. First and foremost is the recommendation to clearly define what it means for an offender to be deemed high risk in order to understand the validity of the forecasting model. Secondly, enhancements to risk assessments should be explored given the finding that 71 per cent of high-harm offenders were not assessed as high risk. The use of data already held by police in identifying risk factors such as those with suicidal tendencies should be considered in order to identify the “needle in the haystack” (Sherman, 2007).Research conducted by the DVDRC and others suggests that suicidal ideations and previous assaults committed against a victim are predictors of high-harm IPV. In order to improve identification of risk, targeting information held by health care and social services in relation to suicidal tendencies and IPV needs to be more widely shared with police. Using automatic digital tracking, these markers should be utilized to target the offenders at highest risk of lethal IPV.The final and most valuable policy recommendation is to institute an automated digital tracking system of the forecasting model that includes all IPV offenders. Through tracking and monitoring the forecasting model, managers will be able to identify the false negatives and risk factors that can be added to the model to improve the accuracy. By reducing false negatives, the most at-risk victims will be identified for outreach and patrol officers will be better informed and prepared when attending IPV calls that have the potential for violence.ConclusionThis study aimed to provide a fulsome account of both how offenders are being identified for assessment in London, Ontario, as well as the post-identification offending behaviour for those forecasted to be high risk. This research has provided an exploration of both the targeting practice and an assessment of the accuracy of the model by tracking the outcomes. It is clear that the current process, utilizing primarily ODARA, identifies individuals that are likely to engage in repeat offending in general but not IPV specifically. It is hoped that this study has provided compelling reasons to improve targeting efforts aimed to reduce the highest harm IPV offending and to spark consideration for evidence-based change among executive decision makers.Maria Wright serves as an Acting Sergeant with the London Police Service. This article is a summary of a Thesis completed by her in part fulfilment of the requirements for the master’s degree in Applied Criminology and Police Management at Cambridge University (2021). She can be reached at, M., & Ariel, B. (2015). Targeting escalation in reported domestic abuse: evidence from 36,000 callouts. International Criminal Justice Review, 25(1), 30-53.Chambers-McClellan, A., (2002). Evidence for the escalation of domestic violence in 911 call records Doctoral dissertation, Medical College of Georgia.Sherman, L. (2007). The Power Few: Experimental Criminology and the Reduction of Harm. Journal of Experimental Criminology, 3, 299–321.Sherman, L., & Berk, R.A., (1984). The specific deterrent effects of arrest for domestic assault. American Sociological Review 49, 261-272.Walker, L.E., (1984). The battered woman syndrome. New York, NY: Springer.
Stay in Touch
Register with your email address for access to our publications

Implementing Evidence-based Policing

Policing can no longer move forward with initiatives and programs without critically questioning their effective-ness. EBP goes further than simply examining data at the surface level. Instead, it looks to employ empirical methods and past findings to ques-tion why we do what we do, and to evaluate if our strategies are benefi-cial or harmful.


Using Data Differently

Policing professionals and leaders know that meaningful and robust data is needed to identify, understand and address systemic issues that affect community safety and police-community relationships. This includes collecting, using and reporting race- and identity-based data to support strategies and action plans that advance equity in policing, and transparency and accountability to the public.


A Collaborative Approach to Sourcing Video Data for Police Investigations

In July 2019, a 40-year-old male was charged – and eventually convicted – with attempted murder after racing his tractor-trailer through a Belleville city suburb and slamming into his partner’s apartment building. Thankfully, she was uninjured in the spectacular attack. Video evidence demonstrating the suspect’s reckless driving behaviour was a key part of the investigation and an element that required hours of investigator’s time to source.