Angela E. Moser, M.A.
Mary Ann Campbell, Ph.D.
Centre for Criminal Justice Studies & Psychology Department
University of New Brunswick
Saint John Campus
March, 2012
Intimate partner violence (IPV) is a widespread public health problem, estimated to occur in
12% to 40% of adult romantic relationships in Canada and accounting for nearly one quarter of
violent crimes reported to police (Canadian Center for Justice Statistics, 2011; Statistics
Canada, 2006). Police officers are often the first responders to IPV and have the opportunity to
offer proactive interventions to reduce the risk of repeat episodes. However, decision-making
about which proactive interventions may be most helpful for which type of offender and victim is
often not straightforward. Fortunately, instruments have been developed to assist with risk
assessment and risk management. One such instrument is the Ontario Domestic Assault Risk
Assessment (ODARA), which was developed by Hilton, Harris, and Rice (2004) in Ontario,
Canada. Unlike other risk tools, the ODARA is developed specifically for use by police officers
and is scored based on information commonly available to them in police records.
Originally, the ODARA was designed for assessing the risk of future physical violence by male
perpetrators against their female partners and it has been found to effectively predict whether
an offender will reoffend and how quickly he is likely to do so. Although this is the most
common circumstance of IPV, this profile only represents one type of IPV offender. Thus, the
current research examined a diverse sample of police reports of IPV that also included female
offenders, same sex couples, and perpetrators of non-physical violence.
A random sample of 200 offenders was selected from police reports of IPV generated in 2004.
These cases were then followed for 6 years to record details of any subsequent IPV offending.
Three major research questions were addressed: 1) what is the typical profile of perpetrators of
IPV and the contextual details of this violence in the Saint John community, 2) to validate the
utility of the ODARA risk instrument for predicting subsequent episodes of IPV in both male and
female perpetrators, as well as in cases of non-physical partner abuse, and 3) to evaluate the
actions of responding police officers to the index call for service to determine which responses
maximized reductions in subsequent intimate partner violence episodes.
Results of this research found that the ODARA was able to discriminate between recidivists and
non-recidivists, regardless of perpetrator gender, victim gender, or type of violence committed
(physical vs. non-physical). Male and female perpetrators did not significantly differ in their
ODARA risk profiles, their frequency of IPV offending, or the amount of injury inflicted on their
victims. Using a scale derived from descriptions of offenders’ behaviour found within sampled
police reports, the current study also found evidence for the role of psychopathic personality
traits in IPV offending. When a score based on these traits was added to the ODARA total
score, significant improvement was noted in the prediction of whether, how often, and how
quickly a perpetrator would reoffend.
The current research also found that police decision to arrest was related to situational factors
(e.g., victim injury) rather than to offender risk-level. Although the police officer’s need to arrest
is recognized, it should be noted that offender arrest was not associated with reductions in IPV
recidivism. Rather than relying on “gut instincts” and primarily reactive policing methods to
respond to IPV, it is recommended that police formally assess risk to triage offenders and
victims into suitable intervention channels based on risk level. Ongoing engagement with, and
monitoring of, high risk individuals is likely necessary to reduce future IPV. The delivery of
appropriate, effective interventions will require collaboration between police and various outside
agencies (e.g., Department of Social Development, Domestic Violence Outreach), to form an
integrated community strategy targeting intimate partner violence.
Table of Contents
Executive Summary .................................................................................................................... ii
Table of Contents....................................................................................................................... iii
1.0 Introduction ........................................................................................................................... 1
1.1 Predictors of IPV ............................................................................................................ 2
1.2 Historical Overview of Criminal Justice System’s Response to IPV ................................ 4
1.3 The Role of Risk Assessment ........................................................................................ 5
1.4 Risk Assessment Instruments for IPV ............................................................................ 6
1.5 IPV Risk Assessment by Police Officers ........................................................................ 7
1.6 Current Study................................................................................................................. 8
2.0 Method .................................................................................................................................. 8
2.1 Sample .......................................................................................................................... 8
2.1.1 Inclusionary Criteria ................................................................................................ 9
2.2 Measures ....................................................................................................................... 9
2.2.1 Ontario Domestic Assault Risk Assessment (ODARA) ........................................... 9
2.2.2 Level of Injury Scale (L-Injury) ............................................................................. 10
2.2.3 Linear Violence Scale (L-Violence) ...................................................................... 10
2.3 Procedure .................................................................................................................... 10
2.3.1 Coding of Index IPV Incidents............................................................................... 11
2.3.2 Coding of Police Action......................................................................................... 11
2.3.3 Coding of Recidivism Incidents ............................................................................. 11
2.3.4 Additional Predictor Variables ............................................................................... 11
2.3.5 Recidivism Outcomes ........................................................................................... 12
2.3.6 Interrater Reliability of Coded Measures ............................................................... 12
2.4 Statistical Analysis ....................................................................................................... 13
2.4.1 Original ODARA ................................................................................................... 13
2.4.2 Escalation of Violence .......................................................................................... 14
2.4.3 Predictive and Incremental Validity of Additional Predictor Variables .................... 14
2.4.4 Revised ODARA (R-ODARA) .............................................................................. 14
3.0 Results ................................................................................................................................ 14
3.1 Context of IPV in Saint John ........................................................................................ 14
3.1.1 IPV Offender Characteristics ................................................................................ 14
3.1.2 Victim Characteristics ........................................................................................... 15
3.1.3 IPV Offense Characteristics .................................................................................. 16
3.2 Utility of the ODARA .................................................................................................... 17
3.2.1 ODARA Risk Level Comparisons.......................................................................... 18
3.2.2 Incremental Validity of Psychopathic Traits ........................................................... 19
3.3 L-Injury, L-Violence, and Recidivism over Time ........................................................... 20
3.4 Pre-Index Contact with Police Services ........................................................................ 21
3.5 Police Response to IPV Events.................................................................................... 22
3.5.1 Police Action by ODARA Risk Level ...................................................................... 23
4.0 Discussion........................................................................................................................... 24
4.1 The Role of Perpetrator Gender in IPV Risk Assessment ............................................ 26
4.2 Relevance of the IPV Context to the Assessment of Risk ............................................ 27
4.3 Victims and Revictimization ......................................................................................... 29
4.4 Police Response .......................................................................................................... 30
4.5 IPV Risk Management and Intervention ....................................................................... 34
5.0 Conclusion .......................................................................................................................... 39
6.0 References.......................................................................................................................... 41
Appendix A: Variables Coded in the ODARA ............................................................................ 53
Appendix B: Level of Injury Scale (L-Injury) and Linear Violence Scale (L-Violence) ................. 54
List of Tables and Figures
Table 1. P-Trait Scale for the Measurement of Psychopathic Personality Traits ....................... 55
Table 2. Offender Characteristics ............................................................................................. 55
Table 3. Percentage of Pre- and Post-Index IPV Offending Events Committed by ModeratelyPersistent and Stable-Persistent Offender Subtypes .................................................. 56
Table 4. Victim Characteristics ................................................................................................. 56
Table 5. Index Offense Characteristics for Male and Female Perpetrators ............................... 56
Table 6. Correlations between Contextual Variables of the Index Event and IPV Recidivism ... 57
Table 7. Injury and L-Violence Scores for Male and Female Offenders .................................... 57
Table 8. Predictive Validity of ODARA Total Score for Predicting IPV Recidivism as Measured
by ROC Curve Analysis .............................................................................................. 58
Table 9. Pearson Correlations Between ODARA Total Score and IPV Characteristics............. 58
Table 10. Total ODARA Scores by ODARA Risk Category and Perpetrator Gender ................. 59
Table 11. ODARA Risk Level Comparisons for Pre- and Post-Index IPV Offending .................. 59
Table 12. Mean Survival Time by ODARA Risk Category ......................................................... 59
Table 13. Summary of Multiple Regression Analysis for Suspect, Victim, and IPV Context
Variables in Predicting Arrest..................................................................................... 60
Figure 1. Kaplan-Meier survival curves for ODARA risk categories .......................................... 61
Validation and Expansion of the Ontario Domestic Assault Risk Assessment (ODARA)
Instrument: An Early Warning System
Intimate partner violence (IPV) is the actual or threatened violence that occurs between a victim and
perpetrator who are currently, or were formerly, involved in an intimate relationship (Campbell, 2002;
Whitaker & Lutzker, 2009). The partners may be male or female, homosexual or heterosexual, living
together or separated, and they may or may not be currently involved in a sexual relationship. IPV
includes acts of physical violence (hitting, kicking, choking), sexual violence (sexual assault, sexual
coercion, rape), psychological/emotional abuse (verbal attacks, isolation, controlling behaviour),
stalking (following and harassing of an unwilling target; Matthews, 2004), and/or communicating the
threat of physical or sexual violence (Whitaker & Lutzker, 2009).
IPV is now widely recognized as a serious public health concern with painful consequences for victims,
their families, and society as a whole (Campbell, 2002; Garcia-Moreno, Jansen, Ellsberg, Heise &
Watts, 2006, Howe & Alpert, 2009). Abused women have poorer overall physical and mental health
than non-abused women (Ratner, 1993). Injuries, fear, and stress associated with IPV can lead to
serious health problems, such as chronic pain, gastrointestinal disorders, gynaecological problems
(vaginal bleeding, pelvic pain, urinary tract infections), central nervous system symptoms (fainting,
seizures), and a variety of mental health problems, including depression, anxiety, and post-traumatic
stress disorder (PTSD; Campbell & Lewandowski, 1997; Campbell & Soeken, 1999; Campbell, 2002;
Coker, Smith, Bethea, King, & McKeown, 2000; Golding, 1999; Ratner, 1993, Ruiz-Perez, PlazaolaCastano, & Rio-Lozano, 2007). Furthermore, children who live in a violent home are at increased risk
of experiencing direct physical harm, as well as the indirect negative effects of being exposed to such
stressful environments. Child abuse is estimated to co-occur in 30% to 60% of adult IPV cases
(Kuelbs, 2009).
Roughly one third of women worldwide have experienced some form of abuse by an intimate partner
(Garcia-Moreno et al., 2006; Ruiz-Perez et al., 2007). A multi-country study conducted by the World
Health Organization (WHO) involved interviewing over 24,000 women in ten different countries
confirmed the pervasiveness of IPV across a variety of geographical and cultural contexts. The WHO
study reported the prevalence of IPV among women who had ever been in an intimate relationship to
be between 15% and 71%, with most sites reporting prevalence rates between 30% and 60% (GarciaMoreno at al., 2006). Estimates of the prevalence of violence involving intimate partners in Canada
range from 12% to 29% for current relationships and from 20% to 40% when based on lifetime
estimates (Naumann, Langford, Torres, Campbell, & Glass, 1999; Statistics Canada, 2006). Notably,
intimate partner violence accounts for approximately one quarter of all violent crimes reported to police
in Canada, with spousal homicides accounting for 47% of all family related homicides (Canadian Center
for Justice Statistics, 2011).
Despite these relatively high IPV rates, the fact remains that this type of crime is underreported to the
police (Wolf, Ly, Hobart, & Kernic, 2003). Estimates of the proportion of IPV incidents that are actually
reported range from 2% to 52% of all IPV cases (Dunford, Huizinga, & Elliott, 1990; Johnson, 1990),
meaning that a large number of violent incidents fall below the radar of the criminal justice system. The
reasons for which IPV victims choose not to contact police are poorly understood, but may include
situational and personal characteristics, prior negative experiences with police, and/or fear of
repercussions from the perpetrator (Wolf et al., 2003). Wolf et al. found that victims were reluctant to
call police unless the abuse was severe enough to have caused an observable injury. Other victims
believed that police would trivialize the matter or take the side of the attacker. Victims reported fear of
being blamed and fear of retaliation from the batterer. Still others were embarrassed and ashamed to
contact police, especially when they had been sexually abused or raped. It has been argued that
police officers should be aware of these concerns when responding to IPV calls and as such, Wolf et al.
(2003) advocated for police training that emphasizes the importance of listening to victims and
reassuring them that they are not to blame for the violence and the importance of holding the suspect
As police officers are often the first to respond to incidents of IPV, they are required to quickly assess
the situation and determine a fitting course of action based on the limited information available to them.
These assessments tend to be subjective and are based on the officer’s appraisal of the immediate
danger and their experience with similar situations in the past (Hoyle, 2008). It has been shown,
however, that structured assessment tools provide more accurate estimates of the risk of future
violence in a given situation than do subjective appraisals of violent events (Andrews & Bonta, 2006;
Kropp & Hart, 2004).
The current study assessed the predictive validity of a structured risk assessment tool known as the
Ontario Domestic Assault Risk Assessment (ODARA; Hilton, Harris, Rice, Lang, Cormier, & Lines,
2004). The ODARA is specifically designed for use by police officers to inform their decisions about
appropriate responses to IPV situations with divergent circumstances, but has yet to be validated in
Atlantic Canada. To contextualize the importance of IPV risk assessment and its methods for the
current study, the following introduction will provide a discussion of several theoretical approaches to
the study of IPV and an overview of risk factors associated with this type of behaviour. This discussion
will be followed by an historical overview of the criminal justice system’s response to IPV, as well as a
discussion about the field of violence risk assessment in general and specific to the prediction of IPV.
The limitations of existing risk tools will be discussed and the goals of the current proposed research
will be explained.
Predictors of IPV
Despite the divergence in the literature with regards to the definition and causes of IPV, there are a
number of well-established factors that increase the risk of violence in an intimate relationship. These
factors pertain to the perpetrator, characteristics of the victim, the victim-perpetrator dynamic, and
contextual/situational factors.
Perpetrators of IPV are not a homogenous group; they vary on multiple dimensions, including severity
and frequency of violence, type of aggressive acts committed, historical and demographic factors,
substance use, and mental health. Nonetheless, a number of empirically identified offender
characteristics have been consistently related to intimate partner violence. Some of these include age,
past violent behaviour, general antisocial behaviour, hostility and alcohol/drug abuse (Hilton et al.,
2004). Kingsnorth (2006) examined a sample of IPV arrestees and found that the three perpetrator
characteristics most strongly predictive of rearrest for IPV within an 18-month follow up were use of a
weapon, the offender’s prior arrest for any offence, and the presence of a protective order. Certain
sociocultural and demographic features also have been linked to the broader behaviour of domestic
violence, including unemployment, low education, a lack of social supports, and having experienced or
witnessed domestic violence in childhood (Grann & Wedin, 2002).
Bennett, Goodman, and Dutton (2000) examined the presence of psychological abuse within the
relationship and found it to be an important risk factor for IPV. Psychological abuse consisted of
dominance, isolation, and emotional/verbal abuse in their study. They found that psychological abuse
was predictive of more severe, long term IPV. In particular, high scores on psychological dominance
accurately classified over 75% of IPV recidivists. Graham-Kevan and Archer (2008) found that the use
of certain controlling behaviours was predictive of physical aggression in IPV. These behaviours
consisted economic abuse, the use of coercion and threats, emotional abuse, and isolation. Wallach
and Sella (2008) examined attribution styles of men who had been violent against their partners and
found that evasion of responsibility for having committed the offence was associated with an increased
risk of future violence. Furthermore, these men failed to recognize the need to take positive action to
stop the cycle of violence. Associations have been found between psychopathic personality traits and
the commission of domestic violence (Gondolf & White, 2001; Swogger, Walsh & Kosson, 2007).
There is also evidence that psychopathy is both a descriptive and predictive construct with respect to
IPV perpetrators, with prevalence estimates of psychopathy ranging from 15% to 30% among batterers
(Huss & Langhinrichsen-Rohling, 2000). Dutton (2002) argued that IPV perpetrators display an
“abusive personality,” which is characterized by features reflective of psychopathic personality traits,
such as manipulation, impulsivity, anger, and a proclivity for substance abuse and promiscuity. It is
important for police to be aware of the presence of these character-based risk factors, which may not
necessarily be ascertained through observations of the physical scene alone and require the use of an
information gathering tool that will cue police to ask the right questions to ascertain these traits.
Less attention has been given to characteristics associated with IPV victimization than to IPV
perpetration; however, some victim vulnerability factors have been identified. Although often met with
controversy and criticism, studies have found an increased risk of IPV victimization to be associated
with inadequate access to resources, younger age, unemployment/low income, lack of social support,
depression, drug/alcohol abuse, inconsistent behaviour/attitude towards the violent partner, and
violence in the family of origin (Belfrage & Strand, 2008; Golinelli, Longshore, & Wenzel, 2008; Nixon,
Resick, & Nishith, 2004). It is possible that studies on risk factors for repeated abuse often fail to
include victim-related variables because of the notion that doing so places blame on the victim.
Advocates have worked hard in recent decades to change the institutional culture of blaming the victim.
Identifying IPV victim characteristics could be perceived as flying in the face of this progress by
suggesting that the victim was somehow at fault. On the other hand, it is important to gain a better
understanding of these risk factors if victims are to become empowered and learn how to reduce their
vulnerability (Cattaneo & Goodman, 2005).
In addition to perpetrator- and victim-related factors, there are a number of contextual factors
associated with IPV. These include elements of the situation in which the violent event occurs and
certain features of the victim-perpetrator relationship. The dissolution of a relationship is one of the
more well-known contextual risk factors for IPV (Robinson, 2006). Wilkinson and Hamerschlag (2005)
examined situational determinants of IPV and discovered that the status of the relationship was
significantly associated with violence risk. Specifically, abuse was more frequent and severe after a
couple had separated than while they were still involved in a relationship. Furthermore, women who
had previously resided with their partner were more likely to be victimized than those who had never
cohabitated. Women who sought restraining orders were also more likely to experience subsequent
severe abuse compared with women who did not seek protective orders. Finally, women who have
recently left a violent relationship are at the greatest risk of homicide by their former partner (Robinson,
Evidence suggests that IPV is repetitive by nature. Farrell, Buck, and Pease (1993) found that when
women called police to report an incident of IPV, they were likely to call again. Furthermore, the
probability of a subsequent call to police increased with the number of previous calls. Thirty five
percent of households reported a second IPV incident to police within five weeks of the first incident.
Within the five weeks after the second incident, 45% of households reported a third IPV incident. It
may be possible to predict the likelihood of future IPV incidents based on the number of past incidents
that have been reported to police (Mele, 2009). Therefore, the number of previous calls to the police
should be taken into consideration when assessing the risk of future IPV and may be a useful
benchmark for officers to consider when estimating this risk. In addition, the violence risk assessment
literature may further inform the assessment of risk in IPV situations and highlight response strategies
to manage this risk.
Historical Overview of Criminal Justice System's Response to IPV
Intimate partner violence was largely ignored by the criminal justice system during the 1960’s. The
practice at the time encouraged officers to avoid arrest if possible, opting instead to calm the parties
through mediation in an effort to simply prevent a breach of the peace (Belknap & Hartman, 2000;
Schneider, 2000). As recently as the 1980's, the general practice within policing was still to discourage
arrest and prosecution in cases of IPV (Garner & Maxwell, 2009; Leisenring, 2008). By the late 1980’s,
however, conventional wisdom concerning the criminal justice response to IPV had begun to change,
largely due to the influential Minneapolis experiment conducted by Sherman and Berk (1984). In this
experiment, offenders were randomly assigned to one of three police response groups when officers
were called to an incident of IPV. The three responses were: 1) to arrest the suspect, 2) to provide
counselling to the parties involved, or 3) to bar the suspect from the home for 8 hours. A total of 330
victims were involved in the experiment. Police records were reviewed and victims were interviewed to
determine whether or not subsequent IPV episodes occurred over a six month period after the initial
incident. Findings indicated that arrest was the most effective method of reducing recidivism, as the
number of repeat occurrences was significantly lower among those suspects who were arrested than
among those who were counselled or removed.
Sherman and Berk’s ground breaking study contributed to large scale changes in police response to
IPV, including the implementation of mandatory or pro-arrest policies in several U.S. states and
Canadian provinces (Schneider, 2000). Officers, under mandatory arrest policies, are required to arrest
a perpetrator if there are reasonable grounds to indicate that IPV has occurred, regardless of whether
or not the victim wants to make a formal complaint. Under pro-arrest policies, officers are allowed to
use more discretion in their decision, although arrest is encouraged in most cases.
The adoption and implementation of mandatory arrest laws has been controversial (Leisenring, 2008).
On the positive side, such policies have been viewed favourably because they remove decision making
power from both the victim and the police and send a message that IPV is a serious concern and will
not be tolerated. However, critics have argued that mandatory arrest laws strip victims of their power to
make decisions regarding their own interest. This also may lead victims to be less likely to involve the
police in future IPV incidents because they fear that their partner will be arrested (Leisenring, 2008).
Debate also exists as to whether or not mandatory arrest policies actually contribute to reductions in
IPV recidivism. Indeed, studies examining the effectiveness of such policies, including replications of
the Minneapolis experiment, have produced contradictory results. Some have even demonstrated a
“backfiring” effect of arrest in which violence was escalated (Schmidt & Sherman, 1996). Thus, given
the diverse nature of the problem, critics have argued against a uniform policy of IPV response and
suggest that police response should be tailored to meet the unique requirements of a given situation
(Crenshaw, 1994; Epstein, 1999).
As a result of increased attention to IPV, the number of cases entering the criminal justice system has
increased substantially in recent years (Bennett et al., 2000). Despite the widespread implementation
of mandatory arrest laws, there is evidence that police are still more lenient with perpetrators of IPV
than with perpetrators of other forms of violence. Avakame and Fyfe (2001) found that police were less
likely to arrest men who had assaulted their female partners than men who had committed other types
of assault. Even if an offender was arrested, there is no guarantee that he/she would be prosecuted
and convicted. Garner and Maxwell (2009) examined the extent of prosecution for IPV over five
countries and found that only 16.4% of reported IPV incidents resulted in an arrest. Of those cases in
which an arrest was made, only 30% of these offenders were convicted, which supports previously
reported findings that prosecution rates for IPV are rare and infrequent (Sherman, 2000; Hartman &
Belknap, 2003). The conviction rate is lower still, with only 16.4% of reported IPV incidents and 30.5%
of IPV arrests leading to convictions (Garner & Maxwell, 2009). Thus, many police reported IPV
incidents do not lead to arrest and only a small proportion of those arrested and prosecuted are
ultimately convicted.
Given that the estimated number of IPV incidents reported to police, and the number of subsequent
arrests, prosecutions, and convictions are alarmingly low, these data speak to the need for
improvements in the criminal justice response to IPV. One means of improving criminal justice
response is to adopt a process for identifying those individuals who have the greatest need for
intervention and protection. The development of methods to assess risk factors associated with IPV is
critical if experts intend to: 1) distinguish low, moderate and high risk IPV cases; 2) develop useful
intervention strategies that address identified risk factors at a level that matches the assessed risk
level, and 3) successfully reduce the likelihood of future IPV incidents. To achieve these goals, front
line professionals would benefit from valid, user-friendly risk assessment tools that are tied to risk
management strategies.
The Role of Risk Assessment
Methods developed to reduce IPV should be consistent with the Risk-Need-Responsivity Model (RNR;
Bonta & Andrews, 2007), which is used in correctional settings to guide the assessment and treatment
of offenders. From a RNR perspective, assessing the risk of violence is the first necessary step in
reducing its frequency. Correctional interventions that are designed and implemented according to the
RNR approach tend to be more effective in reducing recidivism than treatments that do not rely on
these principles (Andrews, Bonta & Hoge, 1990; Bonta & Andrews, 2007).
The first principle of the RNR model is the risk principle, which states that the risk of future criminal
behaviour can be reliably predicted and that the level of intervention and supervision should be
matched to this level of risk. More specifically, higher risk offenders should receive more intensive
intervention and supervision, while minimal intervention and monitoring may be appropriate for lower
risk offenders.
The need principle of the RNR model focuses on the dynamic, or criminogenic, needs of the offender.
Criminogenic needs are risk factors that are associated with an increased probability of recidivism, but
are also changeable with time and/or intervention. Examples of criminogenic needs are antisocial
attitudes/values, criminal associates, and substance abuse. In order to successfully reduce risk, case
management plans must be able to facilitate positive change in an offender’s criminogenic needs
(Bonta & Andrews, 2007; Wong, Gordon, & Gu, 2007).
The final RNR principle is the responsivity principle, which is concerned with the readiness and ability of
the offender to receive and benefit from treatment. This principle states that an offender’s individual
characteristics and the context in which the intervention is to occur will influence treatment
effectiveness and must be taken into consideration when designing effective case management plans
for that individual. These responsivity factors may include cognitive and intellectual abilities, mental
health problems, physical disabilities, motivation and readiness for treatment, and cultural or familial
influences. The interventions used also must be those that have been empirically shown to reduce risk
for the behaviour in question.
Within the context of effective criminal intervention in the RNR model, it is not sufficient simply to
predict risk. Rather, the goal is to manage and reduce risk. RNR models of offender rehabilitation
have been applied by correctional services worldwide and have resulted in reduced recidivism rates
with adolescent offenders, female offenders, and male offenders (Ward, Melser, & Yates, 2007). There
is an abundance of empirical evidence to support the conclusion that these reduced recidivism rates
are a direct result of targeting higher risk offenders with more intense treatment that addresses their
criminogenic needs and matching rehabilitation programs to the individual offender's characteristics
(Andrews & Bonta, 2003; Loesel, 1995; Ward et al, 2007).
Risk Assessment Instruments for IPV
Although progress has been made in identifying risk factors and developing correctional interventions to
target IPV (Whitaker, Morrison, Lindquist, Hawkins, O’Neil, Nesius et al., 2006), there is a paucity of
empirically validated risk assessment instruments for the specific prediction of IPV recidivism as
opposed to general violence (Bennett et al., 2000). One of the best known tools is the Spousal Assault
Risk Assessment Guide (SARA) developed by Kropp, Hart, Webster, and Eaves (1999) for predicting
the likelihood of future IPV. The SARA is perhaps the most empirically studied and supported
instrument for measuring IPV risk. It is comprised of 20 risk factors, divided into three categories that
require the assessor to evaluate an individual’s criminal history, social functioning, and mental health.
Summary judgments of risk as low, moderate, or high (i.e., structured professional judgment) are made
by a trained professional after considering the assessment data in its entirety.
Although the SARA has been found to have predictive utility, it requires in-depth criminal history and
clinical data about the perpetrator, which may not be readily available to front line responders. It also
requires the formation of clinical judgments with regards to an offender’s mental health. For these
reasons, the SARA is not easily implemented or scored by front line police workers or those not trained
in the mental health field. With these limitations in mind, Kropp and Hart (2004) modified the SARA to
form the Brief Spousal Assault Form for the Evaluation of Risk (B-SAFER). The B-SAFER includes ten
risk factors, divided into two categories. The first group of variables is related specifically to spousal
assault and the second is related to the psychosocial adjustment of the perpetrator. The B-SAFER
boasts two major advantages over the SARA. First, it takes less time and is less resource intensive to
administer. Second, the items are comprised of less professional jargon and do not rely on clinical
mental health assessments. These modifications permit ease of use by non-mental health
professionals (Kropp & Hart, 2004). Despite the promise of the B-SAFER, it requires more research to
further establish its validity and use constraints. In addition, some police officers are uncomfortable
with the SPJ process and prefer more concrete risk prediction decision aids.
Few advances have been made in the development of risk tools for assessing spousal violence since
the development of the SARA and the B-SAFER. One exception is the Canadian-based Ontario
Domestic Assault Risk Assessment (ODARA; Hilton et al., 2004). The ODARA is an actuarial risk
assessment instrument that was developed using only police report data. Unlike the B-SAFER and the
SARA, the ODARA was designed to be scored by police officers using only the information available to
them when called to an incident of IPV. The expectation is that officers will use the tool to inform
decisions about how best to respond to the situation (Hilton, Harris & Rice, 2010). The ODARA is
comprised of 13 items, grouped into five categories: police and criminal record items (domestic
violence, general violence, sentence, violation of conditional sentence or protective order); relationship
items (children, stepchildren, abused partner during pregnancy, victim fear); assault history items
(threats, confinement, IPV incidents outside of the home); substance abuse items (drug and alcohol
use/abuse); and victim support items (barriers to victim support; Hilton et al., 2004).
ODARA items are scored dichotomously (present or not present) and the sum of these scores is
intended to predict the likelihood reoffending. A score of zero represents the least likelihood of
reoffending and a score of seven or higher places a suspect in the highest risk category. Most
offenders fall somewhere between these two extremes. Higher scores indicate that an offender will
commit more frequent IPV acts, will commit them sooner and will cause greater injury to their victim
than those with lower scores. Research with the ODARA conducted by Hilton and her colleagues has
consistently found the tool to have high predictive validity (Hilton et al., 2004; Hilton, Harris, Rice,
Houghton, & Eke, 2008; Hilton, Harris, Popham, & Lang, 2010a; Hilton, Harris, & Rice, 2010b).
IPV Risk Assessment by Police Officers
Police officers are now becoming increasingly involved in risk assessment and management practices
for IPV as a proactive policing strategy. Their goal is to identify high risk cases that can then be
targeted to receive violence prevention interventions (Hoyle, 2008). The major distinction between
police officers and other risk assessment experts is that police officers have limited information
available to them to assess risk and they are usually not trained in assessment skills. As such, police
officers have traditionally relied on “gut feelings” about the risk of future violence when assessing IPV
incidents. Based on their subjective appraisal of the risk level, they may choose one of a variety of
potential responses, such as arresting the perpetrator, separating the parties, or providing counsel to
diffuse the situation. Rarely, however, do they consider the long term risk of future IPV when making
these decisions. Instead, they tend to focus on the immediate threat of danger should the two parties
remain in close proximity (Hoyle, 2008; Hoyle & Sanders, 2000). There is an abundance of research to
support the notion that structured assessment tools are superior to "gut feelings" about future violence
decision making among criminal justice-related professionals (Andrews & Bonta, 2006; Grove, Zald,
Lebow, Snitz, & Nelson, 2000; Kropp & Hart, 2004; Mossman, 1994).
Various facets of IPV need to be considered when assessing risk, including the severity, imminence,
type, frequency, and duration of the violence within the relationship (Douglas & Ogloff, 2003). The
implementation of an instrument to aid police officers with these assessments removes subjectivity from
the equation and may result in more appropriately delivered responses. Such tools provide a
structured way for officers to gather relevant information and, in turn, have the potential to provide
better service in response to the victim’s specific needs (Robinson, 2006).
Ultimately, risk assessment in IPV aims to identify those victims who are most at risk of experiencing
violence in the future and to target resources towards those who present the highest level of risk
(Hoyle, 2008; Robinson, 2006). Thus, risk assessment tools for IPV used by police should: 1) be based
on the information typically available to police, 2) contain valid risk items based on this information, 3)
produce results that meaningfully inform decision making about risk management strategies and police
response, 4) be structured and standardized to create consistency across cases, and 5) be user
friendly and time efficient. This approach would be consistent with the RNR model of effective
correctional intervention.
There is some evidence to suggest that the information contained in police reports of IPV may be
sufficient, if not instrumental, in predicting the likelihood of future violence (Messing, 2007; Trujillo &
Ross, 2008). Messing (2007) examined police reports of domestic violence to identify factors within that
pool of information that were associated with IPV recidivism. Fifty high violence cases were evaluated.
Police generated narratives (descriptions of the incident) were examined to determine what type of
information was consistently present in police data sources. Identified variables were related to the
occurrence of violence, the type of violence, and the level of victim injury. Messing demonstrated that it
was, in fact, possible to form meaningful judgments of IPV risk based on police report data. She found
a number of independent variables to be significantly associated with increased levels of violence in her
sample, including the use of a weapon, cohabitation without marriage and arguments relating to sex,
jealousy, and issues of control. The major limitation of her study, however, was the use of point in time
data, which eliminated the ability to make judgments about IPV recidivism or escalation over time.
Nevertheless, her research established that it was possible to collect information pertinent to IPV risk
from police records and that the unique data contained in police reports can contribute to more
accurate prediction of IPV than do other sources of information (Messing, 2007; Trujillo & Ross, 2008).
This could represent an improvement in the delivery of intervention and crime reduction strategies as
acute front line assessments may be a richer source of information than retroactive assessments
conducted after the IPV incident.
Current Study
Given the diverse nature of IPV, police responses need to be adapted to meet the unique requirements
of each situation (Crenshaw, 1994; Epstein, 1999). This can be achieved through the use of evidencebased, structured risk assessment protocols that evaluate specific elements of individual cases. The
proposed research aimed to validate an existing decision tool designed to assist police officers with the
assessment of future risk of intimate partner violence. This was achieved by examining prospective
police report data and testing the ability of the ODARA to predict multiple post-index outcomes,
including the likelihood of future IPV, the severity of IPV incidents, the escalation of violence, and the
level of harm caused to the victim. The utility of the ODARA was also assessed on a broader sample
than the one for which it was developed (i.e., males committing physical violence against female
intimate partners). Its utility had not previously been validated for females perpetrators of IPV or for
suspects who commit non-physical acts of IPV, such as threats or harassment.
The following main hypotheses were tested in the current research study:
1) The original ODARA would correctly categorize offenders as low, moderate or high IPV
risk, even when the sample includes females and incidents of non-physical violence
(threats, harassment). It was also expected to accurately predict the likely time interval
before the next IPV incident will occur.
2) The level of police response (low, moderate, high) would at least partially match the level
of risk of future IPV (low, moderate, high) as determined by the total ODARA score.
3) Measures of injury and violence severity change would significantly vary over time in
cases of repeat violence and be predicted by the ODARA total score.
4) Variables associated with psychopathy, the number of previous calls to police and the
willingness of the suspect to take responsibility for his/her actions would add incremental
validity beyond the total score of the ODARA for predicting IPV outcomes given that
these variables also have been found to predict IPV.
A random sample of 200 cases was drawn from the 4076 IPV related reports generated by the Saint
John Police Force between January 1st and December 31st of 2004. Cases were selected according to
a table of random numbers. Each selected case was reviewed by the first author prior to inclusion in
the sample to determine whether it met the operational definition of IPV for the purposes of the current
study (described below). If a case did not contain the necessary inclusionary criteria, then it was
discarded and a new case was randomly selected to replace it. This screening process continued until
200 appropriate cases were identified. These 200 cases comprised the IPV index events (baseline
events for which initial measurements were taken). For each index event, the perpetrator was identified
and it was noted whether the incident was the first reported IPV episode committed by this party.
Subsequent files were then reviewed for approximately the next six years (ending December 31, 2009)
to identify additional IPV committed by the perpetrator.
The sample included 174 males and 26 females. The mean age of the perpetrators was 35.5 years
(SD = 10.8). The sample was predominantly Caucasian (94%), with the remaining 6% being
represented by First Nations (2%), African Canadians (2%), East Indian (1.5%) and Asian (.5%).
Inclusionary Criteria. The operational definition of IPV used in the current study was relatively broad
given the uncertainty at the onset of the research regarding the nature, quality and completeness of the
information that would be contained in the police reports. The incident reports that were included were
required to meet three minimum criteria: 1) the victim and perpetrator were currently, or previously,
involved an intimate relationship, 2) police were called as a result of an issue that directly involved the
two parties in the intimate relationship, 3) officers who responded to the call identified the incident as a
"dispute" between the intimate partners, whereby the perpetrator’s aggression was directed at, or
intended to victimize, the other party. Both a victim and a perpetrator were unambiguously identified. If
a report did not clearly identify which party played which role, that report was not included in the
Permission to access police records for the purpose of this study was obtained from the Chief of Police,
Saint John Police Force, Saint John, New Brunswick, Canada. The first author was a civilian member
of the Saint John Police Force and was, therefore, security cleared and authorized to access and
review all case files required to carry out this research. All information was reviewed on site, only by
the first author and a secondary coder who reviewed 20% of the cases for the purpose of assessing
interrater reliability of the coding scheme used to extract information from police records. This second
coder was also a civilian member of the Saint John Police Force and, therefore, also had the
appropriate security clearance and authorization to review the information contained in the police files.
A synopsis of the proposed research was submitted to the Human Ethics Review Board of the
University of New Brunswick-Saint John Campus, and received the Board’s approval prior to its
Ontario Domestic Assault Risk Assessment (ODARA; Hilton et al., 2010b, see Appendix A). The
ODARA is an observer-rated risk assessment tool developed for police officers to estimate the risk of
domestic assault. It was used in the current study to establish a baseline risk level for each index
event. The ODARA provides an overall risk score based on information gathered by police and
contained in police reports of domestic incidents. According to the ODARA, a domestic incident is
defined as an event in which the man being assessed has assaulted his current or previous female
cohabitating partner and/or her children. It is intended to only apply to heterosexual relationships in
which the male has committed a physical assault. It does not include assaults by children on parents or
by siblings on one another (Hilton et al., 2010b). This definition was modified for the purposes of the
current study to reflect a non-gendered definition and the inclusion of homosexual relationships, as well
as non-physical incidents of violence (e.g., threats, harassment), but parent-child and sibling-sibling
assaults were excluded.
The ODARA consists of 13 dichotomously scored items selected for their ability to predict IPV
recidivism. Each item receives a score of 0 (not present) or 1 (present). ODARA items include the
following: prior domestic incident; prior nondomestic incident; prior custodial sentence of 30 days or
more; failure on prior conditional release; threat to harm or kill at the index assault; confinement of the
partner at the index assault; victim concern; more than one child; victim’s biological child from a
previous partner; violence against others; substance abuse; assault on victim when pregnant; and
barriers to victim support. Scores for each item are tallied to produce a total risk score, representing
the likelihood of reoffense. Higher scores indicate that an offender will likely commit a greater number
of IPV acts, commit them sooner and cause greater injury to their victim than those with lower scores
(Hilton et al., 2004). Based on the score they receive, offenders are categorized as having a low (score
0-2), moderate (score 3-6), or high (score > 7) risk of IPV recidivism. In the event that information
required to score an item was missing or unclear, then that item was not scored and the total score was
prorated using a table of adjusted scores provided by Hilton et al. (2010b). This prorating method can
only be used when there are < 5 missing items.
Level of Injury Scale (L-Injury; Messing, 2007, see Appendix B). The amount of injury sustained by
the victim at the index event and each subsequent IPV recidivism event during the follow up period was
assessed using a 5 level Injury Scale (L-Injury) developed by Messing (2007). Injury scores ranged
from 0 (no injury and no complaints of pain) to 4 (broken bones, loss of consciousness, stitches,
broken/missing teeth). The descriptor of level 4 injuries was modified to include a broader range of
severe injuries than was included in Messing’s original scale (e.g., hospitalization); however, the scale
still represents the 4 levels of injuries. The types of injuries described by levels 1 to 3 included swelling,
scratches, bruising, black eye, bloody nose, etc. Permission was obtained from the original author of
the L-Injury scale to use this measure in the current research (J. Messing, personal communication,
October 11, 2009).
Linear Violence Scale (L-Violence; Messing, 2007, see Appendix C). The form and severity of
violence perpetrated by the accused at the index and subsequent IPV incidents in the follow up period
were measured using Messing's (2007) Linear Violence Scale (L-Violence). Messing conducted a
linear regression with the L-Injury scale as the dependent variable and several violent actions (punch,
kick, push, bite) as independent, predictor variables. The results of this regression were used to design
a linear scale that measured the severity of violence perpetrated by a suspect. The resulting linear
violence scale (L-Violence) classified violent behaviour by the suspect into one of five categories,
ranging from 0 (no violence) to 4 (severe violence). Based on the L-Violence scale, acts of aggression
by the perpetrator that had the potential to cause more severe injury to the victim were considered to be
more violent than actions that had the potential to cause less injury. The L-Violence scale is similar to
the L-Injury scale, but it focuses specifically on the action of the suspect and, therefore, captures the
level of intended aggression regardless of the actual injury sustained by the victim. For example, a
perpetrator may be unsuccessful in making contact when attempting to inflict injury on the victim, yet
the level of violence attempted was high. According to the L-Violence scale, punches, bites, and hits
with an object represent the most severe category of violence. As with the L-Injury scale, the
description of Level 4 violence was modified to include a broader range of violent actions than was
originally included in Messing’s scale. The items shoot/attempt to shoot and stab/attempt to stab were
included in the most severe category of the L-Violence scale. Strangulation, kicking, stomping, and
grabbing were considered moderately violent (level 3). Throwing objects, slapping and punching were
minor acts of violence (level 2), while forcible entry into the victim’s home and acts of vandalism were
considered abuse (level 1). The L-Violence scale corresponds to the L-Injury scale for the severity of
harm endured by the victim. Permission was obtained from the original author of the L-Violence scale
to utilize this measure in the current research (J. Messing, personal communication, October 11, 2009).
Police files recorded during a six-year period (January 1, 2004 to December 31, 2009) were accessed
to obtain information about each IPV event. A detailed coding process was used for each index
incident to capture known victim and suspect characteristics, elements of the situation, and elements of
the relationship. The ODARA also was scored in its original form for each index event to capture index
IPV risk level. ODARA total scores given to the perpetrator at the index event were used to categorize
him/her as low (L), moderate (M), or high (H) risk of IPV recidivism in accordance with scoring
instructions for the instrument. L-Injury and L-Violence scores were also obtained for the index events.
These scores provided a baseline of violence severity from which to assess escalation (or deescalation) of violence in subsequent IPV incidents.
Coding of Index IPV Incidents. A comprehensive coding guide, based loosely on the content of
Messing’s (2007) coding scheme, was used during the review of the current police file data. Coded
victim and perpetrator variables included such items as gender, race, prior criminal record, probation
orders, parole, drug and/or alcohol influence. Contextual elements of the incident were coded, such as
the time/day/month the incident occurred, whether the victim or suspect placed the call to police, the
reason for the altercation, and whether or not there was a weapon involved.
Coding of Police Action. The coding guide also included a variable describing the action of the police
in response to each dispute. These responses were examined to determine how they mapped onto the
level of risk exhibited by the perpetrator, as determined by the ODARA risk classifications of low (L),
moderate (M), or high (H) IPV risk.
Coding of Recidivism Incidents. After the index events were coded, each subsequent IPV incident
that occurred during the follow-up period was coded for each suspect identified at the index period.
The amount of time (in days) before the next IPV incident occurred was recorded, as was the total
number of repeat IPV incidents committed by each suspect after the index event. The L-Injury and the
L-Violence scales were scored for each subsequent occurrence of IPV to assess whether there was a
change in the level of violence and injury that occurred over time.
Additional Predictor Variables. Several variables were examined to determine whether they added
incremental validity to the original ODARA, one of which was a measure of psychopathic personality
traits (P-trait). A link between psychopathy and IPV offending has been repeatedly identified within the
literature (Boyle, O’Leary, Rosenbaum, & Hasslett-Walker, 2008; Gondolf & White, 2001; Swogger, et
al., 2007). Therefore, the P-trait scale was developed by the authors of the current study to capture
features of psychopathic traits and was subsequently coded for index events by the first author. P-trait
items were selected by reviewing well established formal measures of psychopathic personality traits
and choosing characteristics that would be potentially identifiable and measurable from police contact
with an individual. The elements chosen were common to the construct of psychopathy and could be
assessed through behavioral observation noted in the files (e.g., hostility, callousness, lack of remorse,
dominance, etc.; see Table 1). In-depth descriptions of each item, including examples of behaviors that
reflect each characteristic, are provided in the coding guide.
As a result of meeting with members of the Saint John Police Force (Sgt. J. Oliver, personal
communication, May 27, 2009), two additional variables were identified as being potentially relevant in
the assessment of IPV risk. The first variable was the number of previous calls to the police.
According to the Saint John Police Force, in their jurisdiction, the risk of future IPV is elevated for
couples who have made several prior calls for service to the police. There is also empirical evidence to
support this conclusion. Bannerman (2002) found that serious domestic assaults have, on average,
nine previous calls for service. Thus, the number of pre-index calls to police was recorded and
evaluated as a risk factor in the current sample.
A second variable of interest identified by the Saint John Police Force was whether the offender was
willing to take responsibility for his/her actions. Wallach and Sella (2008) identified a common
attribution style among IPV offenders, characterized by the evasion of responsibility for commission of
the violence, and noted that this attribution style increases the risk for future violent behavior. This was
also noted by the Saint John Police Force as being consistent with their experience. As a result, the
current study included a variable that captured the offenders’ willingness to admit wrongdoing and
accept responsibility for their violent actions. This variable was coded as 0 (accepts no responsibility
whatsoever), 1 (accepts some responsibility, but also uses justifications and rationalizations to divert
personal responsibility), or 2 (takes full responsibility for his/her actions).
A number of victim, perpetrator, and contextual variables with potential relevance to the prediction of
future IPV events were also included in the coding guide as potential incremental validity variables to
the ODARA. These items included such variables as age, gender, and the nature of the
victim/perpetrator relationship (married, separated, boyfriend/girlfriend, etc.). When known, variables
such as employment and socioeconomic status were coded, as was the victim’s level of fear, whether
there was violence in the family of origin (both victim and suspect), the victim’s behaviour towards the
suspect, and the presence of psychological abuse and/or control.
Recidivism Outcomes. The ability of the ODARA total score to predict a suspect’s risk of IPV
recidivism was assessed using five dependent measures: 1) whether a subsequent IPV recidivism
incident occurred at least once during the follow up period, 2) the length of time (in days) to the first
post index IPV incident, 3) the number of subsequent post-index recidivism incidents within the follow
up period, 4) the injury severity of each subsequent recidivism incident as measured by the L-Injury
scale, and 5) the degree of violence escalation across subsequent IPV incidents as measured by the LViolence scale. For the purpose of these outcomes, a separate recidivism incident was considered
when there has been at least a 24-hour period of no calls to police and there was no file information to
indicate that the incident was ongoing and still part of the index event. A recidivism event was
operationally defined in an identical fashion to that used to identify the index IPV event from case files
(i.e., a dispute or assault in the context of an intimate relationship, either with the same or a different
victim intimate partner victim). Thus, a sensitive criterion was used in which the event did not require an
arrest, charge or conviction to be labeled as a recidivism event.
Interrater Reliability of Coded Measures. Twenty percent (20%) of cases were randomly selected to
assess the interrater reliability (IRR) of the coding scheme. A civilian employee of the Saint John
Police Force was trained by the primary researcher in the use of the ODARA, the L-Injury scale, the LViolence scale, and the coding of additional items in the coding guide (e.g., psychopathic personality
traits). This training was achieved through verbal instruction followed by practice scoring of four cases
that were not included in the research sample.
IRR was assessed using an intraclass correlation coefficient (ICC) for the ODARA total scores, RODARA total scores, P-trait total scores, L-Injury and L-Violence ratings. The ICC is commonly used to
assess the consistency of measurements made by two independent raters on a continuous variable
(Bartko & Carpenter, 1976). A two-way mixed model was used to calculate ICC’s and is denoted by
ICC(3,k), for which “3” refers to the third model (mixed model) and “k” refers to the number of raters (in
this case, 1 or 2; McGraw & Wong, 1996). For the purpose of the current study, the single-rater index
(ICC(3,1)) was used as it is a more conservative estimate of IRR. Absolute agreement was selected to
account for systematic variation between the raters. A minimum ICC criterion of .70 was used to reflect
acceptable IRR (Barrett, 2001; Gardner, 1995). All variables exceeded this threshold.
Inter-rater agreement for categorical variables was assessed using the Kappa statistic. Kappa provides
a quantitative measure of the “true” agreement between observers. This calculation is based on the
difference between the amount of actual agreement and the level of agreement that would be expected
by chance alone and, therefore, is a more robust measure than a simple percent agreement (Viera &
Garrett, 2005). Based on a commonly cited interpretive guideline, Kappa values of .41 to .60 reflect
moderate agreement, values of .61 to .80 indicate substantial agreement, and values of .81 to .99
denote almost perfect agreement (1.00 being perfect; Landis & Koch, 1977; Sim & Wright, 2005; Viera
& Garrett, 2005). As with ICC values, a minimum Kappa criterion of > .70 was used to determine
acceptable IRR for categorical variables (Barrett, 2001; Gardner, 1995). All relevant variables met or
exceeded this threshold.
Statistical Analysis
Original ODARA. A receiver operating characteristic curve (ROC) analysis was used to assess the
predictive validity of the original ODARA in the current sample of community IPV offenders. The ROC
analysis indicated the degree to which total scores on the ODARA were able to predict dichotomous
IPV recidivism (i.e., repeat IPV incident/no repeat IPV incident). The analysis was conducted
separately for male and female perpetrators.
ROC curves are widely used in the assessment of accuracy of predictions about violent recidivism
(Mossman & Somoza, 1991; Rice & Harris, 2005; Swets, Dawes, & Monahan, 2000). The ROC curve
was derived from signal detection theory, and is a plot of the true positive rate (percentage of offenders
correctly categorized as recidivists) against the false positive rate (percentage of offenders incorrectly
categorized as recidivists) for each possible decision threshold of a given instrument. Stated otherwise,
it is a plot of the sensitivity (i.e., hit rate for accurately identifying a recidivist) versus 1-specificity (the
false alarm rate associated with incorrectly identifying a non-recidivists as a recidivist). The area under
the ROC curve (AUC) provides a succinct evaluation of the performance of a risk instrument based on
the sensitivity and specificity and is estimated by a score range of 0 to 1.00. The AUC can be
interpreted as the probability that a randomly selected recidivist will score higher than a randomly
selected non-recidivist. The larger the AUC, the greater the difference between the sensitivity and the
specificity of the instrument, and the more accurate the test is in its prediction. If the AUC = .50, the
ability of the instrument to make accurate predictions is the same as chance. An AUC of .70 or greater
is considered a large effect size, with 1.00 reflecting perfect prediction. Two major advantages of the
ROC over other measures used to evaluate prediction tools are that a normal distribution need not be
assumed, and the measure is independent of the base rate of the target behaviour in the population.
When a phenomenon of interest occurs infrequently in a population, some traditional measures are
unable to accurately evaluate risk tools designed to predict this phenomenon. Their accuracy is
increased when base rates are high, and decreased when base rates are low (Rice & Harris, 2005).
ROC methods, however, provide consistent evaluations, regardless of the base rate.
A survival analysis was conducted to examine the utility of the ODARA to predict the time (in days) to
recidivism. Survival analyses are used to model the time from an initial event to the next occurrence of
that event. The time between the two events of interest is known as the survival time (Luke & Homan,
1998; Parmar & Machin, 1995). In this case, the survival analysis was used to model the length of time
between the index IPV event and a subsequent act of IPV. The hazard ratio (HR) is used to measure
the relative survival of more than one group and provides an estimate of the difference between
survival curves for each group (Parmar & Machin, 1995). IPV suspects were divided into low (L),
moderate (M), and high (H) risk groups based on the ODARA scores. The hazard ratio was then
calculated to determine whether time to recidivism significantly differed as a function of risk level. High
risk cases were expected to reoffend at a faster rate than moderate and low risk cases. The survival
analysis gives an indication of the riskiest time frame during which offenders at each risk level are likely
to reoffend. This information is critical to the design and implementation of strategies aimed at reducing
IPV recidivism.
Escalation of Violence. Two repeated measures Analyses of Variance (ANOVAs) were conducted to
determine whether there was a change in the level of injury sustained by the victim (L-Injury scale) and
in the level of violence perpetrated by the suspect (L-Violence scale) at subsequent recidivism
incidents. The ODARA risk category (L, M, or H) was included as a between subjects variable to
determine whether high risk suspects displayed significantly greater escalation of violence than
moderate and low risk suspects. A multivariate ANOVA was also conducted to determine whether high
risk offenders had a higher mean score on the L-Injury and L-Violence scales than moderate and low
risk offenders at the index event.
Predictive and Incremental Validity of Additional Predictor Variables. A hierarchical multiple
regression analysis was conducted to determine whether additional variables coded from the police file
information added any predictive validity to the original ODARA. Individual correlations of items with
each recidivism outcome were calculated to determine whether these items were associated with each
outcome. Only those items of interest that were individually correlated with at least one recidivism
outcome were included in the regression analysis.
Revised ODARA (R-ODARA). After finding that the P-trait scale added predictive validity to the
original ODARA, this variable was combined with the original ODARA to form a revised ODARA tool (RODARA). The ability of the R-ODARA to assess the risk of IPV recidivism was then assessed using
similar analyses to those described above to assess the predictive validity of the original ODARA (ROC
and Survival Analysis).
Context of IPV in Saint John
Most perpetrators in the current sample (97.5%; n = 195) were in heterosexual relationships. Only 5
were in same-sex relationships, all of which were female-female couples. Almost half of the
perpetrators (43%; n = 86) were living common-law with their victims at the time of the index incident.
Only 18% (n = 36) were married and slightly more than a third (34.5%; n = 69) were no longer in a
relationship with the victim. A small proportion (4.5%; n = 9) of couples were in a relationship, but living
separately after having cohabitated in the past. The mean relationship length was 6.7 years (SD = 8.5)
and ranged from 6 months to 55 years.
IPV Offender Characteristics. The index offense was not the first act of IPV perpetrated by most
offenders in the sample, regardless of perpetrator gender, χ2(1) = .24, p = .62. Sixty percent (n = 104)
of male offenders and 65% (n = 17) of female offenders had committed at least one prior IPV act with
the current or a previous partner (see Table 2). There was no significant gender difference in history of
alcohol and/or drug misuse by offenders (χ2(1) = .09, p = .76), with 63% (n = 110) of male and 59% (n =
15) of female offenders having abused substances in the past.
Less than one-tenth (8.3%; n = 17) of perpetrators were willing to accept full responsibility for their IPVrelated actions and over half (53%; n = 106) refused to admit any responsibility whatsoever. Males and
females displayed the same pattern of unwillingness to be accountable for the violence they had
committed, χ2(1) = 3.79, p = .29.
Almost half (41%; n = 82) of offenders demonstrated ineffective anger management strategies, as
reported by their partners or directly observed by the responding officer during the index event, and this
finding did not significantly vary by perpetrator gender, χ2(1) = .004, p = .95. After controlling for
perpetrator age and gender in Block 1 of a hierarchical regression (R2 = .006, F(2, 194) = .582, p = .56),
poor anger control was still found to be a statistically significant predictor of IPV recidivism in Block 2,
R2ch = .13 Fch(1, 193) = 29.11, p < .001.
Statistically significant differences were observed in the criminal histories of male and female offenders.
Most male offenders had a criminal record (62%; n = 108); however, less than a third of females had a
record (31%; n = 8), χ2(1) = 8.95, p = .003. Furthermore, 30% (n = 52) of male offenders had served a
prior custodial sentence of at least three months, as opposed to only 4% (n = 1) of female offenders, χ
(1) = 7.76, p = .005.
IPV perpetrators were followed for an average of 3.5 years (SD = 2.5) after the index offense. The
follow-up period ranged from 1 to 2191 days (approximately 6 years). The mean number of post-index
recidivistic incidents committed by the 88 individuals who reoffended was 1.5 (SD = 2.6), and ranged
from 1 to 18. Over half of these recidivists (53.4%; n = 47) committed 1-2 subsequent IPV incidents
during the follow up period, while roughly one-third (35%; n = 31) reoffending 3-5 times. Nearly 7% (n =
6) committed 6-9 recidivistic incidents, and 4.5% (n = 4) reoffended 10 or more times post index.
Differences were noted among subgroups of perpetrators in their frequency of pre- and post-index IPV
offending. Sixty-one percent (n = 122) of the sample had committed at least one pre-index IPV offense
with the current or a different partner, while 56% (n = 112) did not reoffend during the follow up period.
However, for approximately half of the non-recidivists (48%; n = 54), the index offense was not their
first act of IPV. Only 29% of the total sample (n = 58) were categorized as one-timers, having had no
pre- or post-index allegations of IPV made against them. More than one third of the sample (37%; n =
74) was classified as being moderately persistent. These individuals either had committed no pre-index
IPV incidents but reoffend during the follow up period, or had committed pre- but not post-index IPV.
The remaining 34% (n = 68) of perpetrators were categorized as stable persistent, meaning that they
had committed violence towards their partners both prior to and subsequent to the index incident. As
shown in Table 3, no statistically significant gender differences were observed in the distribution of
these three IPV offender types, χ2(2) = .08, p = .96.
Moderately persistent offenders had committed, on average, 2 pre-index IPV offenses (SD = 2.5, range
= 0-14) and .59 post-index IPV offenses (SD = 1.2, range = 0-6). Almost half of the moderately
persistent group (46%; n = 34) had committed 1 or 2 prior acts of violence against an intimate partner,
while 20% (n = 15) had done so 3 to 6 times and 7% (n = 5) had committed between 7 and 14 preindex IPV offenses. Approximately 20% (n = 15) of moderately persistent offenders committed 1-2
post-index acts of IPV and roughly 7% (n = 5) committed between 3 and 6 post-index acts (see Table
Stable persistent offenders committed an average of 4.3 (SD = 5.1) pre-index acts of IPV, ranging from
1 to 32 prior incidents. Forty-six percent (n = 31) had 1 to 2 previous incidents, with 43% (n = 29)
having between 3 and 6, 6% (n = 4) having 7 to 14, and 6% (n = 4) having committed more than 15
prior violent offenses against an intimate partner. They also continued to recidivate a mean of 3.6
times post-index (SD = 3.3, range = 1-18). The post-index offending pattern of the stable persistent
group was similar to their pre-index pattern, with 47% (n = 32) having 1 to 2 post-index incidents, 43%
(n = 29) having 3-6, 7% (n = 5) reoffending 7-14 times and 3% (n = 2) committing more than 15 postindex IPV offenses (see Table 3).
Victim Characteristics. The victims of IPV were predominantly female (89.5%; n = 179) and the vast
majority were Caucasian (95.5%; n = 191). Other ethnicities represented included First Nations (2%; n
= 4), African Canadian (1%; n = 2), East Indian (1%; n = 2) and Asian (.5%; n = 1). The mean victim
age was 33.6 years (SD = 11.2). Five of the 179 female victims (2.8%) were pregnant at the time of the
index offense.
More than 75% of victims were unwilling to press charges against the suspect (77.5%; n = 155). This
reluctance was observed with equal frequency among both male (76%; n = 16) and female (79%; n =
141) victims, χ2(1) = .11, p = .46 (see Table 4). Some victims (18%; n = 36) initially indicated to police
that they wanted the suspect charged, but later changed their minds and refused to proceed. Most
victims had a history of being in violent relationships (68.2%; n = 136), as determined by victim selfreports in case files as well as by police reports of past IPV against the victim. This was equally the
case for male (57%; n = 12) and female (69.5%; n = 124) victims, χ2(1) = 1.32, p = .18. More than
three quarters of male (76%; n = 16) and female victims (76%; n = 136) exhibited a pattern of
inconsistent behaviour towards the suspect (i.e., repeated cycle of leaving the relationship then
Statistically significant gender differences were observed in the frequency of psychological abuse that
victims reported in their relationships. Significantly more females (30.2%; n = 54) than males (4.8%; n
= 1) reported that their partners engaged in coercive or threatening behaviour towards them, χ2(1) =
6.08, p = .014. The same pattern was observed with behaviours intended to demean or belittle, which
were reported by 27.4% (n = 49) of females and only 4.8% (n = 1) of males, χ2(1) = 5.12, p = .02. Forty
percent (n = 72) of female victims reported that they were fearful of the suspect, however none of the
male victims conveyed this fear, χ2(1) = 13.2, p < .001. There were no significant gender differences in
the presence of jealousy and/or possessiveness within the relationship. Almost one quarter (24%; n =
43) of female victims indicated that the suspect was overly jealous and/or possessive, as did 14.3% (n
= 3) of male victims, χ2(1) = 1.0, p = .24.
IPV Offense Characteristics. Contextual details of the index offenses are displayed in Table 5. The
index offense involved physical violence in 76% (n = 152) of cases, with injuries being sustained by the
victim in 46% (n = 92) of these incidents. Index offenses that were not physical in nature consisted of
threats (17%; n = 34) and harassing phone calls/criminal harassment (7%; n = 14). A weapon was
present in 8.5% (n = 17) of cases and used in 5% (n = 10). Weapons of convenience (i.e., beer bottle,
ashtray, cane) were the most common weapons implemented in the attacks. A knife was used in five
cases and a firearm in two. The perpetrator was under the influence of alcohol and/or drugs in
approximately half of the index cases (54.1%; n = 108) and victims were under the influence in 29.2%
(n = 58) of cases. The most common reasons noted for index altercations included alcohol and/or drug
use by one or both partners (33%; n = 66), jealousy (22%; n = 44), termination of the relationship (18%;
n = 36), and children (13.5%; n = 27). Other less common reasons included general stress (5.9%; n =
12), money (5.4%; n = 11), and sex (2.2%; n = 4). In most cases, the call to police was placed by the
victim (68%; n = 136). Approximately one quarter of calls came from neighbours or other uninvolved
parties and only 7% (n = 14) of 911 calls were made by a child in the home. Correlations between
contextual variables of the index event and IPV recidivism are displayed in Table 6.
Victim injury was measured on a 5-point scale of severity (i.e., L-Injury), ranging from 0 (no injury) to 4
(hospitalization, loss of consciousness, broken bones). Over half of the victims (54%; n = 108) reported
no physical injury during the index incident. Eight percent (n = 16) of victims reported suffering pain
despite having no observable physical injuries (L-Injury = 1). Approximately 16% (n = 32) suffered
abrasions and/or minor scratches (L-Injury = 2), while 17% (n = 34) sustained bruising, cuts, a black
eye, and/or bloody nose (L-Injury = 3). Five percent (n = 10) suffered injuries at the highest level (LInjury = 4), reflecting broken bones, unconsciousness, lacerations requiring stitches, and/or
The level of violence attempted/committed by the offender was also measured on a 5-point scale (i.e.,
L-Violence), ranging from 0 (no violence) to 4 (severe violence). Less than one-tenth (7%; n = 14) of
the index offenses were non-violent, consisting of non-threatening harassment. Seventeen percent (n
= 34) of offenders were verbally aggressive, made threats, or damaged property during the index
incident, corresponding to an L-Violence score of 1. Almost one-third (28%: n = 56) scored a 2,
reflecting behaviours such as pushing, slapping or throwing objects at the victim. Approximately one
quarter (24%; n = 48) of perpetrators scored a 3 on the L-Violence scale during the index incident. This
is the second highest level of violence, corresponding to behaviours such as grabbing, stomping,
choking and slamming. Another 24% (n = 48) achieved the highest L-Violence score of 4, reflecting
behaviours such as hitting the victim with an object, punching, stabbing, shooting, or attempting to
stab/shoot the victim.
The mean L-Injury and L-Violence scores for all cases were 1.11 (SD = 1.35) and 2.41 (SD = 1.22),
respectively. There was no statistically significant effect of suspect gender on L-Injury (F(1, 198), = .01,
p = .91, η2 = .00) or L-Violence, F(1, 198), = .16, p = .69, η2 = .001. These means included those cases
in which the victim sustained no injury and/or the perpetrator used no physical violence. When cases of
non-physical violence were removed to examine only those cases in which physical violence was
committed and resulted in injury to the victim (i.e., L-Injury > 1, n = 92), the mean L-Injury score
increased to 2.40 (SD = .91) and the L-Violence score increased to 3.27, SD = .77 (see Table 7).
These mean scores were reflective of injuries consisting of scratches, swelling, bloody nose, and/or
bruising, inflicted by moderate violence such as grabbing, slamming, choking, stomping, or hitting the
victim with an object. Suspect gender did not significantly impact the amount of injury inflicted on the
victim (F(1, 90) = 1.12, p = .29, η2 = .01); however, the severity of physical violence perpetrated by
females was significantly greater than that perpetrated by males, F(1, 92) = 4.68, p = .03, η2 = .05 (see
Table 8).
Utility of the ODARA
The mean ODARA total score for the sample of 200 index cases was 5.17 (SD = 2.49), reflecting a
moderate level of risk. There was no significant difference between the mean score for male (M = 5.28,
SD = 2.52) and female (M = 4.46, SD = 2.25) perpetrators of IPV, F (1, 198) = 2.4, p = .12, η2 = .01.
Higher ODARA total scores were positively associated with IPV recidivism (r = .35, p < .001). Using the
full sample, ROC analysis indicated that the ODARA total score was strongly predictive of IPV
recidivism, AUC = .70, 95% CI [.63, .77]. When cases of physical violence were removed to examine
only cases in which the abuse consisted of threats or harassment (n = 48), again a large effect size
was found for predicting recidivism, AUC = .70, 95% CI [.54, .85]. When separated by gender, a large
effect size was found for the prediction of recidivism among males (AUC = .70, 95% CI [.63, .78]),
whereas the predictive accuracy for recidivism among females corresponded to a moderate effect size,
AUC of .67, 95% CI [.46, .88]. However, the CI’s overlapped for males and females, suggesting that
they were sampling the same population parameter. Nevertheless, the CI for females was substantially
wider than it was for males and it crossed into chance prediction, reflecting less precision in predictive
accuracy for females (see Table 8). This dispersion may be a function of the low number of female
perpetrators in the sample.
A statistically significant correlation was found between higher ODARA total scores and fewer days
passed prior to the first new post-index IPV incident (r = -.36, p < .001). Higher ODARA total scores
were also significantly correlated with a higher number of pre-index IPV offenses committed by the
perpetrator (r = .39, p < .001), a higher number of post-index IPV offenses (r = .39, p < .001), the
victim’s unwillingness to press charges (r = .15, p = .03), and the presence of psychological abuse in
the relationship (r = .40, p < .001). Upon examining the entire sample, there was no statistically
significant correlation found between ODARA total scores and L-Injury scores (r = .12, p = .11) or L-
Violence scores, r = .11, p = .13. However, when examining only those cases in which injury and/or
physical violence was present, the correlation between ODARA total score and L-Injury increased to
.24 (p = .04) and increased to .36 (p = .002) between ODARA total score and L-Violence (see Table 9).
Thus, when physical aggression does occur during the IPV incident, higher ODARA total scores as
associated with greater intended harm and actual harm towards the victim.
ODARA Risk Level Comparisons. ODARA scores were used to categorize offenders as low (0-2 =
low risk), moderate (3-6 = moderate risk), or high risk for recidivism (7-13 = high risk) based on the
guidelines of Hilton et al. (2010b). Of the 200 offenders in the current sample, 16% (n = 32) fell in the
low risk category, with a mean ODARA total score of 1.41 (SD = .62, range 0-2). The majority of the
sample (48%, n = 96) fell in the moderate range with a mean score of 4.38 (SD = .99, range 3-6), and
36% (n = 72) fell in the high risk category with a mean score of 7.94 (SD = 1.01, range 7-11).
The number of female offenders in each risk category was normally distributed, with 23% (n = 6) in the
low risk category, 50% (n = 13) in the moderate risk category and 27% (n = 7) falling in the high risk
category. Male offenders were also normally distributed across the three risk categories, with 15% (n =
26) being low risk, 48% (n = 84) being moderate risk, and 37% (n = 64) being high risk. Risk category
was not significantly dependent on suspect gender, χ2(2) = 1.57, p = .46 (see Table 10).
Post-index IPV recidivism differed significantly among the three ODARA risk categories, χ2 (2) = 20.33,
p < .001. The low-risk category had significantly fewer recidivists (15.6%; n = 5 out of 32) than the
moderate risk category (40.2%; n = 39 out of 96), which in turn had significantly fewer than the high-risk
category (62.5%; n = 45 out of 72; see Table 11).
A one-way ANOVA revealed that the number of IPV acts committed by offenders prior to the index
offense significantly differed by risk level, F (2, 197) = 14.8, p <.001 η2 = .13. Tukey post-hoc
comparisons indicated that the high risk group committed significantly more pre-index IPV offenses (M
= 3.9, SD = 4.1) than the moderate (M = 1.7, SD = 3.6, p < .001) or low risk groups, (M = .19, SD = .74,
p < .001); however, low and moderate groups did not significantly differ from one another (p = .10; see
Table 11). As with pre-index events, a statistically significant main effect of risk category was found on
total number of post-index IPV events, F (2, 197) = 15.5, p < .001, η2 = .14. Tukey post hoc
comparisons indicated that this difference reflected a significantly higher number of events in the high
risk category versus the low and moderate risk categories (p < .001), whereas low and moderate did
not significantly differ from each other, p = .31.
A log-rank (LR) test was used to compare Kaplan-Meier survival curves (time to reoffence) for each risk
category. The log-rank test is used to test the null hypothesis that there is no difference between the
groups in the probability of an event occurring at any point in time. The analysis revealed that the low,
moderate, and high risk offenders significantly differed in the time it took them to commit another IPV
incident after the index event, LR(2) = 22.2, p < .001 (see Figure 1). Over the approximately six-year
follow up, high risk offenders were the quickest to reoffend (M = 2.4 yrs, SD = 2.3), followed by the
moderate risk (M = 3.6 yrs, SD = 2.4) and then low risk (M = 4.9 yrs, SD = 1.8) cases (see Table 12).
Only 5 out of 32 low-risk offenders (15.6%) recidivated during the 6 year follow up and all 5 had done
so within the first 3 years. The first low-risk suspect to reoffend did so 20 days post index, whereas the
second case did not reoffend until approximately 4.5 months after the index offense. There was one
individual in both the moderate and one in the high risk categories who reoffended the day after the
index incident. By 3 months post index, 12% (n = 11) of moderate and 23% (n = 17) of high risk
suspects had reoffended. After 6 months, these rates increased to 23% (n = 22) and 37% (n = 27),
respectively. The rate of reoffending within the moderate risk category was 40% (n = 38 out of 96) and
all moderate risk recidivists had reoffended in less than 3 years (32 months) after the index offense.
The recidivism rate among high-risk offenders was 62.5% (n = 45 out of 72) and all these cases had
reoffended within 2.2 years (26 months) of the index incident.
L-Injury scores did not initially appear to differ by ODARA risk category, F (2, 197) = 2.1, p = .12, η2 =
.02; however, finding may have been due to a floor effect on the L-Injury scale given that there was a
significant number of cases in which no injury was reported or observed (n = 108). Thus, these cases
were removed from analysis and a one-way ANOVA was rerun on the remaining 92 cases in which
there was injury to the victim (L-Injury > 1). This subsequent analysis revealed that when an injury was
inflicted, the severity did vary by risk category, F (2, 89) = 3.9, p <.05, η2 = .08. Tukey post hoc
comparisons indicated that the mean L-Injury score for high risk offenders (M = 2.7, SD = .82) was
significantly elevated compared to low risk offenders (M = 2.0, SD = .82; p = .03). There was no
significant difference between L-Injury scores of moderate risk offenders (M = 2.3, SD = .97) from the
low (p = .49) or the high (p = .14) risk offenders. L-Violence scores did not significantly vary as a
function of ODARA risk category (F (2, 197) = .90, p = .41, η2 = .01). This remained so even after
removing cases in which the violence was non-physical (i.e., verbal aggression, threats; F (2, 149) =
1.3, p = .27, η2 = .02). Thus, high risk IPV offenders may cause greater injury against their victims than
do low risk cases, but offenders in all three risk groups attempt to inflict similar levels of violence in a
given IPV incident.
Incremental Validity of Psychopathic Traits. A 17-item scale was coded for each offender to capture
the presence of psychopathic traits (i.e., P-Trait). The intraclass correlation coefficient for the P-Trait
scale was .71, which exceeded the criterion for acceptable IRR. Internal consistency analysis revealed
a Cronbach’s α of .88 for the original scale, suggesting that its items were meaningfully associated with
each other. However, tests of multicolinearity identified a strong inter-item correlation (r = .79) between
two items (absence of anxiety to police presence, and procriminal attitudes/antiauthority). As such, the
item absence of anxiety to police presence was removed as it had a lower overall item total correlation
than the procriminal attitudes/antiauthority item. Removal of this item did not substantially decrease the
internal consistency of the scale, which became α = .86. Three additional items were removed due to
low item-total correlations (superficial charm, r = .19, shallow emotions, r = .09, and parasitic
orientation, r = .19). A final item, prior breach of conditional sentence, was removed due to content
overlap with an ODARA item that already captures the context of this variable. The internal
consistency of the final 12-item psychopathic trait scale remained high, α = .86. Individual items on this
scale were coded on a 3 point scale indicating the degree to which the suspect displayed the trait (0 =
no, 1 = somewhat, 2 = yes), resulting in a possible total score range of 0-24 (see Table 1). The mean
P-Trait total score was M = 6.2 (SD = 5.1) and ranged from 0 to 21. There was no significant difference
between males and females on this variable, t (198) = -.19, p = .85. There was a significant positive
correlation between P-Trait total score and ODARA total score, r = .62, p <.001. The P-Trait scores
were divided into three equal groups, representing low (0-3), moderate (4-8) and high (9+)
psychopathic trait groups. Chi square analysis revealed that P-Trait level varied significantly by
ODARA risk category, χ2(4) = 70.2, p < .001. Seventy percent (n = 40) of the 58 offenders scoring 9 or
higher on the P-Trait were also categorized by the ODARA as being high risk. Twenty-nine percent (n
= 17) of these offenders with a high level P-Trait score were classified as moderate risk by the ODARA
and only 1.7% (n = 1) was classified as low risk. There were 66 moderate P-Trait cases, the majority of
which (63.6%; n = 42) fell in the moderate-risk category on the ODARA and most of the low-risk
ODARA cases (87.5%; n = 28) scored less than 3 on the P-trait scale.
ROC analysis revealed that the P-Trait scale was strongly predictive of recidivism (AUC = .80, 95% CI
[.74, .86]. Two hierarchical multiple regressions were conducted to test the incremental validity of PTrait over ODARA total score for predicting a) dichotomous post-index IPV reoffending, and b) the
number of post-index IPV incidents. Alpha was set at .025 for these analyses to minimize Type I Error.
Block 1 of the first analysis contained perpetrator age and gender, which were non-significant as a
block of variables that explained .8% of the variance, F (2, 197) = .81, p = .45. In Block 2, the ODARA
total score significantly explained an additional 12% of the variance in IPV recidivism, Fch (1, 196) =
26.7, p < .001. P-Trait total was added to the model in Block 3 and had incremental validity over
ODARA total by significantly accounting for an additional 16% of the variance, Fch (1, 195) = 43.3, p <
.001. In contrast, ODARA total failed to significantly explain any additional variance in post-index
reoffending beyond P-Trait total, R2ch = .00, Fch (1, 195) =.09, p = .77.
A second hierarchical multiple regression was performed to test the incremental validity of the P-Trait
total score over ODARA total score in predicting the number of recidivistic IPV incidents. Age and
gender were included in Block 1 and did not significantly contribute to the prediction of recidivism, R2ch
= .01, F (2, 197) = 1.2, p =.31. ODARA total score, added in Block 2, significantly explained an
additional 14% of the variance in number of post-index IPV incidents committed, Fch (1, 196) = 33.8, p <
.001. When added in Block 3, P-Trait total score had incremental validity over ODARA total score by
significantly accounting for an additional 23% of the variance in the number of post-index IPV incidents,
R2ch = .23, Fch (1, 195) = 72.9, p < .001. ODARA total score had no significant incremental validity over
P-Trait total score in predicting this outcome, R2ch < .001, Fch (1, 195) = .01, p = .92.
A Cox regression was conducted to investigate the effect of several variables on the time passed, in
days, until the first post-index IPV offense or end of the follow-up period in the cases that did not reoffend. Cox regression analyses model the effect of unit increases in a variable on the time it takes for
an event to occur (in this case, recidivism). Since this was one of two Cox regressions to be conducted
during the data analysis, alpha was set at .025. Age and gender, in Block 1, did not significantly
contribute to the prediction of this outcome, χ2( 2) = 1.4, p =.49. In Block 2, ODARA total score was a
significant predictor of time to first post-index IPV offense (β = .23, p < .001, exp(B1) = 1.26), χ2ch (1) =
26.1, p <.001. P-Trait total was a significant predictor of time to recidivism in Block 3 (β = .16, p < .001,
exp(B1) = 1.17), and further improved the predictive ability of the model, χ2ch (1) = 37.9, p <.001. The
ODARA total score did not contribute significant incremental validity over P-Trait total in predicting time
to first post-index IPV offense, χ2ch (1) = 1.2, p = .27.
Each perpetrator’s P-Trait score was added to his/her ODARA total score to create a new R-ODARA
total score. ROC analysis indicated that this combination created a large effect size for the prediction
of recidivism (AUC = .79, 95% CI [.72, .85]), with only minimal overlap in CI’s for the original ODARA
total score at the lower bounds, 95% CI [.63, .77], (see Table 8). When separated by gender, the RODARA was found to be strongly predictive of recidivism for both males (AUC = .78, 95% CI [.72, .85])
and females, AUC = .82, 95% CI [.66, .85]. The R-ODARA also demonstrated high interrater reliability
with an ICC of .80.
L-Injury, L-Violence, and Recidivism over Time
There was a significant strong positive correlation between L-Injury and L-Violence scores, r = .63, p <
.001, in that the more severe the violence attempted, the greater the victim injury. The correlation
between L-Injury and the number of pre-index IPV incidents that a suspect committed was weaker, r =
.14, p = .04. No significant correlation was found between number of pre-index IPV incidents and LViolence scores, r = .10, p = .16. Thus, the level of injury and violence at index was not associated with
the frequency of past IPV incidents, which suggests that the severity of these factors are driven more
by the situational context of the IPV incident itself than by past behaviour.
A mixed ANOVA was conducted to determine if L-Injury and L-Violence scores changed across
successive IPV recidivism incidents and as a function of risk level. Measures of L-Injury and L-Violence
were compared at three points in time (index, first IPV recidivism event, and second IPV recidivism
event). Adding a fourth point in time to the analysis resulted in too few cases for analyses, as only 27
offenders recidivated more than three times in the follow-up period. For this reason, measurements
taken at three points in time (index, first post-index event, and second post-index event) were included
in the analysis. Moreover, low-risk offenders were excluded from this analysis, as only 2 offenders in
that group recidivated more than once. Therefore, the final analysis included only moderate (n = 20)
and high risk (n = 30) offenders. The mixed ANOVA revealed no significant change in L-Injury scores
as a function of time (F(2, 96) = .33, p = .72, η2 = .01), nor was there a significant interaction between
time and ODARA risk category, F(2, 96) = 1.0, p = .36, η2 = .02. Risk category itself, however, did have
a significant main effect (F(1, 48) = 4.9, p < .05, η2 = .09), whereby L-Injury scores were higher for highrisk offenders than for moderate-risk offenders. There was no statistically significant main effect of time
on L-Violence scores (F(2, 96) = .19, p = .83, η2 = .004), nor was there a significant interaction between
time and ODARA risk level, F(4, 96) = 1.8, p = .17, η2 = .04. Moreover, in contrast to L-Injury, there was
no significant between subjects effect of risk category on L-Violence scores, F(1, 48) = 1.7, p = .19, η2
= .03. Thus, as noted above, the level of harm intended and caused to the victim did not significantly
escalate or decrease over time at least across these three incidents of IPV. The intended violence did
not vary as a function of risk level, but higher risk offenders tended to be more successful in causing
their intended harm towards the victim than low risk offenders.
Pre-Index Contact with Police Services
ROC analysis revealed that the number of prior calls to police was moderately predictive of post-index
IPV recidivism, AUC = .69, 95% CI [.62 to .77]. Upon examination of the cut-points of the ROC curve to
determine the number of pre-index calls for service that provided optimal prediction, it was noted that
both sensitivity (.77) and specificity (.52) were maximized at just 1 prior call.
Two hierarchical regression analyses were conducted to test the incremental validity of number of preindex calls over ODARA total score in predicting a) recidivism (Yes/No), and b) number of recidivistic
IPV incidents over time. Again, alpha was set at .025 for these two analyses to control for Type I Error.
After controlling for age and gender in Block 1 of the first model (F (2, 197) = .73, p = .49) and entering
ODARA total score in Block 2 (R2ch = .12, Fch (1, 196) = 26.8, p < .001), number of pre-index calls was
added in Block 3 and explained an additional 2% of the variance, which was a small but significant
portion, R2ch = .02, Fch (1, 195) = 5.2, p = .02. However, pre-index calls for service had no significant
incremental validity over ODARA total score, as the ODARA total score significantly accounted for an
additional 7% of the variance when included in Block 3 (F (1, 195) = 15.3, p < .001) beyond pre-index
calls in Block 2, R2ch = .08, F (1, 196) = 16.1, p < .001.
The second hierarchical regression was conducted with number of recidivistic IPV incidents as the
dependent variable. Again, age and gender did not significantly contribute to the post-index predictive
ability of the model in Block 1, F (2, 197) = 1.17, p = .31. ODARA total score explained 14% of the
variance in Block 2 (Fch (1, 196) = 33.77, p < .001) and number of prior calls, included in Block 3,
significantly explained an additional 13% of the variability in the number of post-index IPV incidents
committed by the suspect, Fch (1, 195) = 36.79, p < .001. When Blocks 2 and 3 were reversed so that
number of pre-index calls was included in Block 2, this variable significantly explained an additional
24% of the variance in number of post-index IPV incidents beyond suspect age and gender in Block 1,
Fch (1, 196) = 61.5, p < .001. When ODARA total score was added in Block 3, it significantly explained
another 4.3% of the variance in the model (Fch (1, 195) = 11.8, p = .001). Thus, number of pre-index
calls did not have true incremental validity over ODARA total score, as it did not explain any variance
beyond that accounted for by ODARA total.
A Cox regression analysis was conducted to assess the utility of number of prior calls in predicting time
to first post-index IPV offense. Given that this was the second of two Cox regressions being
conducted, alpha was set at .025. Suspect age and gender were entered in Block 1 and did not
significantly predict time to next IPV offense, χ2(2) = 1.6, p =.44. In Block 2, ODARA total score was a
significant predictor of time to first post-index IPV offense (β = .23, p < .001, exp(β) = 1.3) and
significantly improved the predictive ability of the model, χ2ch (1) = 27.7, p <.001. Number of prior calls
was included in Block 3 and was not a significant predictor of time to next IPV offense at the stringent p
< .025 level, although was significant at the standard alpha level of .05 (β = .05, p < .05, exp(β) = 1.1,
χ2ch (1) = 4.7, p = .03). When Blocks 2 and 3 were reversed so that number of prior calls (β = .07, p <
.001, exp(β) = 1.1) was included in Block 2 of the model (χ2ch (1) = 19.7, p < .001), and ODARA total
score (β = .21, p < .001, exp(β) = 1.2) was added in Block 3, ODARA total score significantly improved
the predictive ability of the model (χ2ch (1) = 19.9, p < .001). Therefore, number of pre-index calls did not
add true incremental validity beyond the ODARA total score in predicting time to next IPV offense.
Two hierarchical regressions were conducted to examine the prediction of L-Injury scores and LViolence scores by number of prior calls for service. Alpha was set at .025 for these two regression
analyses. After controlling for age and gender in Block 1 of the first model (Fch (2, 197) = .34, p = .71),
number of prior calls was included in Block 2 and failed to significantly predict higher L-Injury scores,
R2ch = .02, Fch (1, 196) = 4.1, p = .04. The second hierarchical regression was conducted with LViolence as the dependent variable. Variance associated with age and gender was statistically
controlled in Block 1, F (2, 197) = .18, p = .83. Number of prior calls was added in Block 2 but was not
found to significantly predict L-Violence scores, R2ch = .01, Fch (1, 196) = 2.0, p = .16. Thus, number of
prior calls for service did not predict the intended or actual harm caused to the victim at the index IPV
Police Response to IPV Events
For 63% (n = 126) of cases in the sample, the police file was forwarded by the responding officer(s) to
Records with no further action being taken beyond documentation of the incident. In 30% (n = 60) of
cases, the suspect was never spoken to by police. This was a result of the suspect having fled the
scene prior to police arrival and the police failing to subsequently make contact with him/her. This
response was the case even in the high risk group, in which approximately one third (32%; n = 23) of
high risk offenders had no contact with police in relation to the index incident. There was no significant
effect of risk category on officers' decisions to proceed with charges or send the file to records, χ2(2) =
1.98, p = .37. However, the decision to forward the file to records varied as a function of L-Injury. The
mean L-Injury score for files sent to records was significantly lower (M = .86, SD = 1.3) than for files not
sent to records (M = 1.6 SD = 1.4, F (1, 198) = 15.2, p < .001. The same was true for L-Violence
scores, whereby files sent to records had a significantly lower mean L-Violence score (M = 2.2, SD =
1.2) than files not sent to records, M = 2.9, SD = 1.0, F (1, 198) = 14.1, p < .001. Thus, the decision to
send a file to records with no further action depended on the severity of violence and injury at the
An arrest was made in 37% (n = 74) of index cases; however, 34% (n = 25) of those arrested never
went to court. The charge most often laid by police was common assault (81.1%; n = 60). Assault with
a weapon, assault causing bodily harm, threats and harassment accounted for another 15% (n = 11),
with the remaining 3.9% (n = 3) consisting of harassing phone calls, breach of peace and probation
violation. Of the 49 individuals who did go to court, only 16 (32.6%) were convicted. The overall
conviction rate for reported IPV index incidents in the sample was 8%. When examining those cases in
which an arrest did occur, the conviction rate was 21.6%. Peace Bonds were issued in almost half of
the cases (49%; n = 24) that appeared before the court. The most frequently imposed sentence was
probation (37.5%; n = 6), followed by a combination of probation and jail (31.2%; n = 5).
Police Action by ODARA Risk Level. Level of police action was coded on a three point ordinal scale
of increasing intensity of response (1 = spoke to parties/took report only, 2 = removed one party/no
charges laid, 3 = arrested/charged suspect). The hypothesis that police action would match risk level,
with more intensive responses corresponding to higher risk offenders, was not supported given that
there was no significant difference in the intensity of police action score across the three risk
categories, F(2, 197) = .64, p = .53, η2 = .006. This was the case for both male (F(2, 171) = .60, p =
.55, η2 = .007) and female offenders, F(2, 23) = 1.6, p = .21, η2 = .12. There also was no difference in
police response intensity as a function of recidivism subtype (one-timer, moderately persistent, stable
persistent; F(2, 197) = .84, p = .44, η2 = .008) and this was true for both male (F(2, 171) = 1.4, p = .24,
η2 = .02) and female offenders, F(2, 23) = .50, p = .62, η2 = .04.
For more than half of the cases in each ODARA risk category (low - 53%, moderate - 56%, high - 54%),
police action was limited to speaking with the involved parties and taking a report. Arrest rates also did
not significantly vary between low (38%), moderate (33%), or high (44%) risk categories, χ2 (4) = 2.6, p
= .62. Notably, the intensity of police response also did not significantly affect post-index recidivism.
Specifically, whether the suspect was arrested or not had no statistically significant effect on whether
he/she reoffended (χ2 (1) = .09, p = .77), or on the number of times he/she reoffended during the follow
up period, F (2, 197) = .57, p = .57, η2 = .006.
Overall, there was no significant difference in the police response to male or female suspects, F (1,
198) = 1.6, p = .20, η2 = .008. The most common police response for both male (52%; n = 90) and
female (65%; n = 17) suspects was to speak with the involved parties. Police arrested 39% (n = 68) of
male suspects and 27% (n = 7) of female suspects. In 9% (n = 16) of cases involving male suspects
and 7% (n = 2) of cases involving female suspects, police removed one party from the scene of the
altercation with no charges pending.
A signification correlation was found between the intensity of police intervention and L-Injury scores (r =
.29, p < .001), indicating that the police were more likely to arrest the suspect when the victim suffered
more serious injuries. There was also a significant positive correlation between intensity of police
action and L-Violence scores (r = .32, p < .001), indicating that police were more likely to arrest the
suspect when the intended violence observed at the time of the call was high, rather than low.
A hierarchical multiple regression analysis examined suspect, victim, and IPV context variables in
predicting arrest (see Table 13). The overall model explained 43% of the variance in police decision to
arrest the suspect. Block 1, containing suspect gender, age, impairment by drugs and/or alcohol,
willingness to accept responsibility, and P-Trait score, was not significantly predictive of arrest, R2ch =
.09, Fch (5, 74) = 1.4, p =.23. Victim variables were included in Block 2, and consisted of victim age,
victim gender, impairment by drugs and/or alcohol, and willingness to press charges. This Block
significantly explained an additional 29% of the variance in the decision to arrest (Fch (4, 70) = 8.2, p <
.001), with the only significant unique predictive contribution being made by the victim's willingness to
press charges, β = .56, t = 5.6, p < .001. Variables pertaining to the nature of the IPV incident were
contained in Block 3, including L-Injury scores, L-Violence scores, and the presence of a weapon. LInjury was the only significant variable in this block, β = .39, t = 2.9, p = .006. This Block significantly
explained an additional 14.4% of the variance in the decision to arrest, Fch (3, 67) = 6.7, p < .001.
ODARA total score was included in Block 4 and did not significantly contribute to the model's prediction
of arrest beyond these other variables, R2ch = .002, Fch (1, 66) = .30, p = .58.
Collectively, the above results suggest that police response was driven by the circumstances of the
specific IPV incident rather than by a perpetrator’s risk status of engaging in IPV. However, significant
differences were found in the frequency with which police referred victims to alternate services (i.e.,
Department of Social Development, Victim Services) as a function of offender risk category, χ2 (2) =
7.52, p = 02. Significantly more victims in relationships with high risk offenders (63%; n = 45) received
referrals than victims whose partners were low risk (34%; n = 11). The rate of community referrals
made at the moderate risk level (53%; n = 51) did not significantly differ from either the low (p = .17) or
high risk groups, p = .34. Thus, there appears to be recognition of need for additional intervention for
high risk cases among police officers.
IPV is a grave social problem, the enormity of which is exemplified by the 44% recidivism rate among
perpetrators in the current sample. Physical violence was used in more than three quarters of the IPV
episodes examined in the current study and the victim was injured in approximately half of these cases.
Accurately predicting IPV risk is the first step in determining suitable responses that effectively reduce
the occurrence of partner abuse. Responses should be evidence-based and grounded in the Risk,
Need, Responsivity model of case management (Andrews & Bonta, 2010), such that intervention
intensity matches offender risk level, intervention focuses on risk factors associated with IPV, and
intervention uses research-informed strategies to promote risk reduction. It is critical that police
officers, who are often the first responders to IPV incidents, appropriately assess the perpetrator's risk
of reoffending in order to best identify the cases requiring more intensive and proactive policing
strategies to reduce this risk.
The current research set out to examine a means of facilitating risk-management decision making by
police officers responding to cases of IPV in Saint John, New Brunswick. In particular, the current
research examined the utility of a newly developed actuarial risk-assessment instrument, the ODARA,
as a means of facilitating this type of decision-making. A second goal of the current research was to
identify risk factors from archival police reports that were predictive of IPV recidivism to determine
whether they added incremental validity to the ODARA. Finally, police responses to IPV incidents were
examined to determine whether response intensity matched suspect risk and to identify which
strategies, if any, were effective in reducing IPV recidivism over a 6 year follow-up period.
Several Canadian studies have provided empirical support for the utility of the ODARA in making
accurate risk classifications for male IPV offenders who have committed physical violence against their
female partners (Hilton et al., 2004; Hilton & Harris, 2009; Hilton, Harris, Popham, & Lang, 2010a;
Hilton et al., 2010b). The current research is the first study to have demonstrated the ODARA’s
effectiveness in predicting IPV recidivism in a sample that includes female offenders, same sex
couples, and perpetrators of non-physical intimate partner violence. In the current diverse sample, the
ODARA was able to discriminate between an offender who would commit another act of violence and
one who was less likely to do so, the frequency with which they could be expected to reoffend, and how
long it would take before the next incident would occur. Offenders categorized as low risk were least
likely to reoffend, committed fewer repeat IPV acts, and took considerably longer to do so than
moderate risk offenders, who in turn were less likely and slower to reoffend than high risk offenders.
Additionally, the results of the current study provided support for the utility of the ODARA in
distinguishing IPV recidivists from non-recidivists regardless of perpetrator gender, victim gender, or
type of violence committed (physical vs. non-physical).
In addition to being predictive of recidivism, ODARA total scores were also related to the amount of
injury the victim suffered during physical assaults when there was physical aggression present. In
cases in which the victim was injured as a result of the violence, more severe injuries were caused by
offenders with higher ODARA scores. This is consistent with Hilton et al. (2004; 2010b), who reported
an association between elevated ODARA scores and increased assault severity. Assaults, threats, and
harassment were attempted with equal frequency at all three ODARA risk levels in the current sample,
but high-risk cases were much more successful with their intended goal of causing the victim physical
harm. Victims of low- and moderate- risk offenders were less likely to suffer the intended
consequences of the assault. Notably, there was little to no relationship between the amount of
violence a perpetrator attempted against his or her partner at any one time and this person’s risk of
being abusive in the future. Indeed, violence severity was not found to be related to ODARA risk level
in the current sample. For several of the repeat offenders, post-index offenses consisted of threats and
harassment interspersed with physical assaults. Similarly, there were non-recidivists whose index
event involved a high level of violence. Thus, severity of attempted harm is not likely to be a reliable
indicator of whether a subsequent IPV incident will occur, and relying on such a factor may lead to
over- or under-estimations of risk when responding to calls for service in the future.
The lack of an association between attempted violence and recidivism in the current sample contradicts
the popular notion that violence severity escalates over time (Frye, Manganello, Campbell, WaltonMoss, & Wilt, 2006; Zara, Ponsoda, & Carrillo, 2009; Zeeve, 2008). It is possible that the lack of
escalation (or de-escalation) was due to the limited number of repeat incidents for many of the
offenders in the sample, which restricted the comparison of IPV severity to only three points in time.
Following cases over additional incidents may have showed the expected escalation or de-escalation in
violence. It also may be that the time between incidents washed out any change in L-Injury and LViolence over time. Escalation or de-escalation may occur when there is a short timeframe between
IPV episodes (i.e., a few days or weeks), but not when the incidents are spaced out (months or years).
Alternatively, the current data could reflect the fact that the degree of injury and violence do not change
in a linear and predictable fashion. Bennett-Cattaneo, Cho and Botuck (2011) conducted a longitudinal
examination of intimate partner stalking and found that severity did not follow a predictable trajectory.
Rather, case-specific variables were implicated in the escalation and de-escalation of stalking for
individual clients. This may be true in the current sample as well, in that the severity of an individual IPV
episode is related to acute, dynamic factors inherent in the situation at the time. Thus, degree of
violence in a single incident may not be a useful risk marker for determining whether the behaviour will
persist or for gauging the anticipated severity of the next incident. If this is the case, then a more useful
and proactive risk management strategy would be to consider the perpetrator and victim’s full case
history when attempting to predict future IPV rather than base this decision on a single episode.
In addition to validating the ODARA for use in Atlantic Canada, the current research examined a
number of factors that were hypothesized to enhance the validity to the ODARA. One of the
hypotheses was that the number of pre-index IPV-related calls to police would be predictive of the
number of future calls. This hypothesis was partially supported, as a greater number of pre-index IPV
calls did predict recidivism. Adding the number of prior calls to the ODARA total score also added to
the prediction of whether or not a perpetrator reoffended, but the number of prior calls did not contribute
to the prediction of how rapidly the next offense would occur beyond what was already estimated by the
ODARA total score. Furthermore, review of sensitivity and specificity data within the ROC analysis
indicated that there only had to be one pre-index IPV call to police to maximally increase the likelihood
that there would be future calls. This result highlights the importance of intervening with perpetrators
and victims of IPV after the first incident in an effort to reduce the risk of future violence, especially in
moderate to high risk cases.
Repeat IPV episodes occurred much more quickly for moderate and high-risk offenders in the current
sample. Three months after the index offense, more than 10% of moderate risk and 20% of high risk
offenders had committed another abusive act. After six months, the recidivism rates increased to more
than 20% of moderate-risk and almost 40% of high-risk offenders. It took only 26 months for all of the
high-risk recidivists to reoffend. Knowing this trend, front-line professionals may achieve greater
effectiveness in responding by adopting a proactive rather than a primarily reactive approach to cases
involving IPV. If an offender is assessed as high-risk, repeated follow-ups/check-ins with both parties
involved in the IPV episode should be conducted within the first ninety days to circumvent future
violence. These check-ins create opportunities for police to monitor IPV risk, but also opportunities to
engage the perpetrator and/or the victim in a way that might facilitate their willingness to seek
assistance and participate in community services that address risk factors associated with IPV (e.g.,
shelters, social services, family/couples therapy, domestic violence-focused interventions, conflict
resolution skill building, interpersonal effectiveness training). Partnerships with the community will be
key to ensure that there are adequate community services for which perpetrators and victims can be
`referred and case managed in meaningful ways. Given that only 5 out of 32 offenders in the low-risk
category had an instance of repeat violence in the entire 6 year follow up, the need to focus resources
and intervention on this group is reduced relative to moderate and high risk cases.
The Role of Perpetrator Gender in IPV Risk Assessment
Perpetrators in the current sample were classified, on average, as having a moderate risk of
reoffending, and this was true for both male and female perpetrators. There were also no gender
differences found in the mean overall ODARA score, or in the ability of the tool to predict IPV
recidivism. This means that male and female perpetrators fell into low, moderate and high risk
categories in equal proportions. Further highlighting the gender similarity in IPV, both genders had an
equal likelihood of reoffending violently against their partner. The fact that females were represented in
all three risk classifications, not just the low-risk category, contradicts the long-standing view that
female perpetrated IPV is less severe than that perpetrated by males (Dobash & Dobash, 1979;
Dobash, Dobash, Cavanagh, & Lewis, 1998; DeKeseredy, 2006).
The actual injury caused to the victim was similar for male and female perpetrators in the current
sample, but females attempted to commit more severe violence than males. Specifically, female
perpetrators were more likely to use severe acts of violence, such as punching the victim or hitting the
victim with an object, whereas males were more likely to attempt less serious violent actions, such as
grabbing or slapping the victim. This finding is consistent with reports that when women do commit
IPV, their actions are more severe than those of abusive men. Cho and Wilke (2010) found that,
although men were victims of IPV less often and received fewer injuries than females, they were more
likely to experience more severe violence by their female abusers. Dutton, Nicholls, and Spidel (2006)
conducted a literature review on domestic violence and concluded that females are as abusive as
males in intimate relationships. They further concluded that women who commit violence against their
partners possess many of the same characteristics and risk factors as male abusers. This perspective
contradicts previous views, which state that women are rarely the aggressors in abusive relationships
and that they use violence for different reasons than men (i.e., self-defense; Dobash et al., 1998;
Henning, Martinsson & Holdford, 2009; Miller, 2001). In fact, a number of studies suggest that
women’s IPV is often not defensive and that men sometimes use violence in self-defense. A study by
Sommer (1994) in Winnipeg, Canada, found self-defense to be the motive for only 10% of female
perpetrators and 15% of male perpetrators of IPV.
Examinations of clinical populations, including domestic violence intervention programs, have found
that women report similar reasons for their aggression to those reported by men, including anger,
jealousy, control, a lack of commitment from their partner, and a means of “getting through” to their
partner (Carrado, George, Loxam, Jones & Templar, 1996; Graham-Kevan, 2009; Harned, 2001;
Henning, Jones, & Holdford, 2005). In the current study, various reasons for the violence were
reported to police, including jealously, money, children and alcohol. There were no gender differences
in the reported reasons for the abuse, with the exception of “breaking up,” which was reported
significantly more often by female perpetrators than male perpetrators as the reason for their violence.
Thus, feelings of “losing control” over a partner, which often has been cited as an explanation for male
perpetrated post-separation violence (Brownridge, Chan, Hiebert-Murphy, Ristock, Tiwari, Leung &
Santos, 2008; Sev’er, 1997), also contributes to female perpetrated IPV. A review of several studies
that examined gender variations in the relationship of various risk factors to IPV was conducted by
Medeiros and Straus (2006). They reported that, in 72% of cases, there were no gender differences in
the relationship between a risk factor and IPV between male and female offenders. Specifically, risk
factors such as youthfulness, conflict in the relationship, dominance, and having an angry personality
were noted with equal frequency in female and male abusers.
In general, the current research challenges the traditional feminist perspective (Dobash & Dobash,
1979) which views female IPV as a less serious phenomenon than male-perpetrated partner violence.
Notably, the current findings suggest that a female abuser is as likely as a male to reoffend against a
partner, is as likely to be a high-risk offender, and is as likely to inflict as much harm as a male
perpetrator. Furthermore, feminist theory is less able to account for IPV within the context of
homosexual relationships, especially in the case of female-female pairings. Examination of the current
sample reveals that the only instances of same-sex violence occurred in lesbian relationships. Lie and
Gentlewarrier (1991) conducted a large scale survey of over 1000 lesbians and reported that over half
(52%) had been abused by a female partner and more than half reported having committed violence
against their partner. Similarly, Lie, Schilit, Bush, Montagne, and Reyes (1991) found that women who
had been in both lesbian and heterosexual relationships reported higher rates of verbal, physical, and
sexual abuse by lesbian partners than by heterosexual partners, which emphasizes the importance of
developing a non-gendered theory of IPV rather than adhering to the traditional feminist perspective
that explains domestic violence in terms of patriarchal aggression.
Although the ODARA was developed with males, the current study shows that it also can be used to
inform the prediction of female IPV recidivism. A moderate effect size was found for the prediction of
female recidivism (AUC = .67) and this did not substantially differ from the predictive ability the ODARA
displayed with male perpetrators (AUC = .70). It should be noted that the confidence interval was
slightly broader for females (95% CI [.46, .88]) than for males (95% CI [.63, .78]), indicating less
precision in predicting female recidivism. This variation may be due to the lower number of females in
the sample. It also could be that the ODARA is not capturing risk factors for female perpetrators of IPV
as well as it should, given that it was developed using male data. Studies have found that there are
certain risk factors that may be predictive of IPV recidivism in females, but not in males. Ménard,
Anderson, and Godboldt (2009) reported that the severity of the original assault and having previously
terminated the relationship with the victim were predictive of repeat IPV by females, but not by males.
In contrast, history of probation or parole and unemployment were predictive of IPV recidivism by
males. The relationship between gender and risk factors for IPV recidivism needs to be further
explored so that risk tools designed for use with male and female offenders can offer greater utility in
the prediction of repeated IPV episodes. The identification and addition of gender sensitive IPV risk
items should enhance the ability of the ODARA to predict female perpetrated IPV acts.
Relevance of the IPV Context to Assessment of Risk
One of the original goals of the current research was to identify additional suspect, victim, and
contextual variables beyond those scored by the ODARA that would enhance the validity of the tool as
a means of predicting IPV recidivism. Several variables, including suspect/victim age and gender, level
of suspect/victim impairment by alcohol/drug, presence of a weapon, length of relationship, type of
relationship (i.e., common-law, married, etc.), and suspect willingness to take responsibility for his/her
actions, were examined for their potential influence on IPV recidivism, but few of these variables were
found to explain any additional variance beyond that which was already explained by the ODARA. The
lack of incremental predictive or discriminative power of these factors may have stemmed from the fact
that many of these characteristics were present in the majority of IPV episodes. For example, whether
or not a perpetrator was willing to take responsibility for the violence he or she had committed was not
found to be predictive of recidivism. This lack of predictive power likely stemmed from the fact that over
90% of perpetrators in the sample refused to take full responsibility for their actions. Despite the
inability of this variable to predict IPV recidivism over and above the ODARA, it appears to be a key
element of the IPV offender profile relative to individuals who do not engage in violence. Henning et al.
(2005) identified denial and/or minimization of the violence and blaming the victim as significant risk
factors for both male and female partner abuse reoffending and argued that these factors should be
important targets for treatment.
Despite not adding incremental validity to the ODARA, several contextual variables were significantly
correlated with IPV recidivism. Suspect characteristics, such as poor anger control, jealousy,
possessiveness, and having a criminal record, were all significantly correlated with IPV recidivism in the
current study. Certain victim characteristics, such as a history of abusive relationships and inconsistent
behaviour towards the suspect were also associated with recidivism. Although the inclusion of these
offender and victim-oriented variables did not enhance risk prediction above information captured by
the ODARA, they are relevant to risk management and should be targeted for intervention to reduce
Interestingly, the suspect’s level of substance-induced impairment was not correlated with recidivism,
despite the fact that approximately 60% of the offenders in the sample had a history of substance
abuse and more than half were under the influence of drugs or alcohol at the time of the index offense.
Although alcohol and/or drug use was not predictive of IPV recidivism or of the amount of injury and
violence that occurred, it was a common element of most IPV cases in the current sample. This was
true whether the suspect was male or female and regardless of risk level. This finding adds to a large
body of contradictory findings with respect to the role of substance misuse in domestic violence. The
inhibiting effects of substances and their influence on one’s problem solving and emotional functioning
likely contribute to the occurrence of IPV. Although addictions do not justify the use of violence,
substance abuse may precipitate conflict and impair an individual’s ability to use appropriate coping
skills (Florsheim & Moore, 2008; Orford, Velleman, Copello, Templeton & Ibanga, 2010; Smyth &
Wiechelt, 2005). Given this fact, it is likely that IPV offenders with comorbid addictions would benefit
from substance abuse treatment programs in addition to IPV-focused interventions. Indeed, studies
have found that offenders who receive treatment for their substance abuse show greater reductions in
IPV following IPV-focused interventions than those whose addictions are left untreated (O’Farrell, FalsStewart, Murphy & Murphy, 2003; Stuart, Ramsey, Moore, Kahler, Farrell, Recuperco, & Brown 2003).
Fals-Stewart, Leonard, and Birchler (2005) discovered that drinking alcohol was more strongly
associated with the likelihood of severe IPV among men with antisocial personality disorder (ASPD)
compared with those without ASPD who drank. It could be that personality organization rather than
alcohol consumption is the more relevant risk factor for intimate partner abuse. Indeed, antisocial
personality characteristics may be a strong predictor of IPV recidivism (Harris, Hilton, & Rice, 2011).
Recent research suggests that IPV offending is more influenced by stable, long-term traits than by
situational factors (Nicholls, 2011; Norlander & Eckhardt, 2005; Storey et al., 2009) and that enduring
antisociality may in fact play a causal role. Harris et al. (2011) were able to significantly account for
long term IPV recidivism using measures of psychopathy (PCL-R score), antisociality (symptom count
for DSM-IV Antisocial Personality Disorder) and criminal history (e.g., prior correctional sentences, past
non-domestic violence). This domain of enduring antisocial traits explained a significantly greater
portion of the variance (R = .417) in long-term IPV offending than domains reflecting neighbourhood
characteristics (R = .205) or relationship variables (R = .303). Moreover, in a meta-analysis, Norlander
and Eckhardt (2005) reported higher levels of anger and hostility in IPV perpetrators than in non-violent
individuals. They further noted that IPV perpetrators scoring in moderate- to high-risk categories
displayed more anger and hostility than did low- to moderate-risk IPV subtypes. These findings are
consistent with the current study in which poor anger control by perpetrators, as reported by the victims
or directly observed by the responding officer, was predictive of IPV recidivism.
Although there is considerable evidence to support the role of psychopathy in interpersonal violence
(Boyle et al., 2008; Felson & Lane, 2010; Hilton et al., 2008; Harris et al., 2011; Walsh, Swogger,
O’Connor, Schonbrun, Shea, & Stuart, 2010), most formal assessment measures of psychopathy
require a high level of clinical training to administer. As such, the utility of this construct for firstresponders to IPV incidents (e.g., police officers) has been of limited value. Hilton et al. (2010b)
identified the measurement of psychopathy in front-line context as a challenge for future research. In
response to this challenge, the current study identified a group of observable and measurable traits that
could be gleaned from police and victim descriptions of the offender’s behaviour. The resulting P-trait
scale included items such as the expression of procriminal attitudes, poor anger control, impulsivity,
callousness, and deceitfulness. These traits were found to be strongly predictive of IPV recidivism in
the current sample. The addition of P-trait total score to the ODARA total score created the R-ODARA,
and significantly improved the prediction of whether, how often, and how quickly an offender would
reabuse his or her partner relative to use of the ODARA in its original form. Findings indicated that the
more psychopathic personality traits an offender displayed, the greater his or her likelihood of
committing future IPV. Furthermore, high-risk offenders displayed significantly more psychopathic traits
than did low- or moderate-risk offenders. Stable persistent offenders also displayed more psychopathic
personality characteristics than did less prolific IPV offenders. Taken together, this evidence suggests
a role for stable, enduring personality characteristics in IPV, which may account for the fact that many
of the situational variables examined in the current study did not contribute to the prediction of
recidivism. These traits can be informally assessed from police records, and should inform risk
assessment in cases of IPV.
Victims and Revictimization
There has been much research attention devoted to perpetrator characteristics that influence risk of
IPV reoffending, yet there is a relative paucity of data on victim vulnerability factors despite indications
that certain factors can increase one’s risk of being victimized by an intimate partner (Belfrage &
Strand, 2008; Golinelli, et al., 2008; Nixon et al., 2004). This deficit is partially due to society’s
reluctance to “blame the victim”; however, some research suggests that characteristics of both parties
and their relationship dynamic influence IPV risk (Moffitt, Robins, & Caspi, 2001). To maximize the
reduction of future violence, comprehensive IPV intervention and prevention strategies need to address
both victimization risk factors and perpetrator risk factors (Nicholls, 2011). Thus, a greater
understanding of victim and victim-perpetrator dynamic risk factors is required to inform such
Almost 75% of victims in the current research had a history of being in violent relationships. Hoyle
(2008) argued that many victims tend not to behave in rational, harm-reducing ways after being
victimized. Through intervention, victims can become empowered and inform themselves about ways
to make better choices about and within their intimate relationships and, thereby, minimize their risk of
further victimization. It is important for victims to be aware of risk factors that are in their control so that
they can make informed changes in their life and escape the cycle of violence (Cattaneo & Goodman,
2005). Based on the current research, a history of being in violent relationships among victims was
predictive of violent recidivism by their current partner. In other words, victims who were abused by
previous partners were likely to be repeatedly abused by their current partner. This was especially true
for victims in relationships with high-risk, stable-persistent offenders. These victims also displayed
significantly more inconsistent behaviour towards their partners than did victims in relationships with
less prolific IPV perpetrators (i.e., one-timers). This inconsistent behaviour by a victim, defined as
repeated cycles of breaking up and reconciling with his or her abuser, also was found to predict
repeated violence. It is critical for victims to recognize this pattern when they are trying to escape from
an abusive relationship so that the pattern can be broken.
Victim age, gender, and criminal record were not predictive of revictimization. The predictive ability of
several other contextual variables such as the victim’s employment, education, and the availability of
supportive peers, could not be assessed in the current sample because these factors were rarely
mentioned in the context of police reports. Future research on IPV risk factors should explore the
influence of these variables. Dynamic victim characteristics, such as alcohol/drug use and pregnancy,
were not linked to repeat victimization either. Revictimization, like IPV recidivism, may be influenced
more by internal traits rather than by dynamic, context-specific factors. This argument is in line with the
model proposed by Foa, Cascardi, Zoellner, and Feeny (2000), in which psychological difficulties, such
as PTSD, depression and anxiety put women at risk of revictimization, whereas resilience (defined by
Foa et al. as an ability to adjust to and recover from adverse circumstances) serves to reduce the risk
of revictimization. Further research is needed to clarify the interrelationship among IPV, psychological
difficulties, and resilience.
Although few gender differences were found relating to the perpetration of IPV in the current sample,
there was a gender difference noted among victims. Several variables were identified that reflected the
presence of psychological abuse within the relationship. Male and female victims reported equal levels
of jealousy and possessive behaviour by their partners, but there were significant gender differences in
the reporting of other types of emotional abuse. Significantly more females than males reported that
their partners engaged in coercive or threatening behaviour towards them. Females also reported
higher rates of belittling and demeaning conduct by their partners. Almost half of the female victims
reported to police officers at the scene that they feared their partner, whereas none of the male victims
admitted to being fearful of their female abuser. This latter finding is inconsistent with other data, given
that female perpetrators caused as much injury to their victims as did male perpetrators and committed
more serious violent actions. It is possible that male victims underreported their fear of an abusive
female partner due to stereotypical gender expectations of masculinity (Cercone, Beach & Arias, 2005;
Stanko & Hobdell, 1993).
Police Response
The hypothesis that the level of police response would match the level of perpetrator risk was not
supported by the current findings. Offender risk category had no relation to the action of the police in
response to the index IPV event. In more than 60% of cases, police response consisted simply of
talking with the involved parties and generating a police report. Arrests were made in just over a third
of cases, regardless of the offender's risk level. Police were just as likely to arrest a suspect who was
at a low risk of reoffending as they were to arrest one who was high-risk. Furthermore, perpetrator
gender did not appear to influence the decision to arrest. Although slightly fewer women (27%) than
men (34%) were arrested, this difference was not found to be significant. This lack of an arrest-bias on
the part of the Saint John Police Force is noteworthy as it challenges the frequently reported tendency
for police to arrest a disproportionate number of males for domestic violence offences relative to
females (Capaldi, Shortt, Kim, Wilson, Crosby, &Tucci, 2009; Hamel, 2011).
It was concerning that approximately one third of suspects were never spoken to by police, as they had
left the scene prior to police arrival and were not subsequently contacted. This occurred with equal
frequency among high- and low-risk offenders and is consistent with research that has found a key
factor in police decision to arrest is whether the suspect has left the scene (Buzawa & Hotaling, 2000;
Robinson, 2000). Obviously, it is difficult to make an arrest if the suspect is not present, but in cases of
IPV, it is rare for police to be unaware of the suspect’s identity or unable to ascertain his or her
whereabouts. Of further concern is the fact that over 60% of cases received no police intervention
beyond the officer attending the scene and taking a report. This could be due in part to the fact that in
Canada, there are no nationally legislated procedures governing police response to IPV. Notably, this
response is not unique to the Saint John Police Force. Recent research has found similar rates of nonintervention by officers attending domestic violence incidents throughout the 10 Canadian provinces.
Barrett, St. Pierre, and Vaillancourt (2011) examined a subset of survey data from Canada’s 1999
General Social Survey. The selected respondents were females who reported experiencing physical or
sexual IPV by a male perpetrator. Fewer than half of these women stated that police attended the
scene, took a report, and/or initiated an investigation and in roughly 75% of cases, the police left the
perpetrator in the home.
Although approximately one third of offenders in the current sample were arrested by police, only one
third of those arrested subsequently appeared in court and even fewer were convicted of an offense.
The overall conviction rate for reported IPV index incidents in the current sample was 8%, which is
consistent with other reports of IPV convictions being rare and infrequent (Garner & Maxwell, 2009;
Hartman & Belknap, 2003; Sherman, 2000). Thus, arresting IPV perpetrators is not likely to have a
long-term impact on subsequent IPV episodes if most cases are never prosecuted and no other
interventions are provided. Proactive police-based responses and community interventions may prove
more impactful. Officers were significantly more likely to refer victims of high-risk perpetrators to
alternate community resources, such as Victim Services or the Department of Social Development.
This finding indicates that, although police arrest decisions were not influenced by risk level, these
officers may have intuitively viewed high-risk offenders as more dangerous and provided victims with
more options and resources in those cases. Seith (2005) reviewed 126 case files and interviews with
police officers and found that even when arrest did not take place, officers would often provide
information about legal rights and procedures, refer victims to relevant institutions, and suggest safety
Perhaps the police feel that their hands are tied in terms of what they can do in situations in which their
“gut feelings” tell them that there is an elevated level of threat in the absence of observable evidence.
They may not feel justified enacting a formal arrest if they cannot observe evidence of a physical
assault or cannot prove that a crime has been committed. Even if the police do exercise their right to
make an arrest against the victim’s will, without a reasonable probability of conviction, the crown
prosecutor may not proceed with the charge and the case is likely to be dropped before reaching the
level of the courts. Seasoned officers who have had the frustrating experience of making an arrest and
conducting an investigation only to have their case not approved by the crown are less likely to make
the same effort in the future. This frustration leads to disillusionment and even detachment for some
police officers who have experienced the revolving door of domestic violence, whereby they repeatedly
deal with the same victims and suspects with little support from the courts (Horwitz, Mitchell, LaRussaTrott, Santiago, Pearson, Skiff, et al., 2011). Despite this frustration, when officers sense that the victim
is in a dangerous situation, they provide him or her with options for resources or safety planning
measures in hopes of improving the victim’s circumstances.
Another reason that police may choose to arm the victim with information rather than arresting the
suspect is their uncertainty regarding sustained victim cooperation. An officer may be hesitant to invest
a lot of time and effort into an investigation when the victim is wavering in his or her decision to support
the charges. Even though officers have the authority to make an arrest without being requested to do
so by the victim, without a formal complaint and statement from the victim indicating his or her desire to
press charges against the suspect, the probability of conviction is very low (Garner & Maxwell, 2009;
Hartman & Belknap, 2003; Sherman, 2000). Conceivably, police officers may feel that empowering the
victim to make a change in the situation is more likely to reduce the chances of future violence than
would trying to target the suspect's behaviour. Police may feel that arresting the suspect is not going to
do anything to improve the situation. This notion was indeed supported and reinforced by the current
findings, given that whether or not police made an arrest had no bearing on whether a suspect
committed a repeat episode of IPV during the follow up period. Although this seems to contradict the
reasoning behind the highly popular pro-arrest policies currently practiced by many police forces
(Schneider, 2000; Sherman & Berk, 1984), it is actually not surprising. If police are encouraged to
make an arrest when an assault occurs, regardless of a suspect’s risk, the action would not be
expected to have a noticeable effect given that everyone is being treated the same way. The fact of the
matter is that low-risk offenders probably will not reoffend, whether they are arrested or not, and highrisk offenders probably will.
Despite Sherman and Berk (1984)’s initial conclusion that arrest was effective in reducing recidivism,
subsequent replications have not found the same results and reviews of their study cite numerous
methodological problems, such as officers did not always make the responses that they were supposed
to as part of the study protocol and victim reports lacked credibility (Binder & Meeker, 1988; Lempert,
1989; Mederer & Gelles, 1989). Mixed and conflicting findings since have been reported in the
literature regarding the impact of arrest, with some research reporting no effect or even a negative
effect in which arrest has been linked to increases in recidivism (Berk, Campbell, Klap & Western,
1992a; 1992b; Sherman, 1992; Sherman & Smith, 1992). Despite these contradictory findings, and the
fact that the original research was fundamentally flawed, the results of Sherman and Berk's (1984)
study have become the most widely cited in the field and provided a basis for recommendations that
law enforcement agencies develop policies requiring arrest as the preferred response for domestic
violence (Buzawa & Buzawa, 2003). The implementation of pro-arrest and mandatory arrest policies
throughout much of the United States and Canada followed.
Perhaps the most reasonable interpretation of the mixed findings surrounding the impact of arrest is
that arrest alone will not reduce recidivism because additional intervention is required to address the
factors contributing to the violence. Although the idea that arrest is effective on its own has intuitive
appeal, it rests on the assumption that the offender considers arrest to be a deterrent to the criminal act
(Buzawa & Buzawa, 2003). Deterrence depends on the offender’s ability to weigh the costs and
benefits of his or her behaviour and to determine that the costs outweigh the benefits. This would
require that, in the moment before committing a violent act, the perpetrator has considered that: a) the
police may be called, b) the police may attend and arrest him or her, and c) that the long term negative
consequences of being arrested would outweigh the short term “benefits” of committing the violence,
despite a low probability of a conviction. If the threat of arrest and conviction is not salient to the
perpetrator, then these consequences will not deter the violent behaviour, especially in the case of
repeat IPV offenders who have abused their partners in the past without being “caught”, or have been
treated leniently by the criminal justice system. This being said, arrest should not be abandoned as an
option by police. Police obviously need to maintain a means of regaining control of dangerous
situations and individuals, protecting victims, and responding to the commission of a crime, but making
an arrest in the absence of intervention to address the broader context of the perpetrator and victim’s
risk of subsequent IPV is unlikely to have the desired effect of reducing future violence.
Risk assessments have long been used in the criminal justice system to inform bail, sentencing and
parole release decision-making. Given the heavy reliance on such methods by other areas of the
criminal justice system, it makes sense for police to incorporate risk assessments into their practices to
promote proactive policing responses intended to prevent future crime before it occurs. When victims
of violence are seeking help, police represent the first point of contact with the criminal justice system.
Rather than relying on “gut feelings” about the dangerousness of a situation and subsequent risk of
future violence, police could benefit from a tool that provides an unbiased, objective assessment of the
risk to inform their decision-making. Their resulting response should involve proactive police measures
in keeping with the principles of contemporary evidence-based methods of community policing, which
support community partnerships and problem-solving techniques (Parent & Whitelaw, 2008).
Police have considerable discretionary power in the decision to make a formal report, to make arrests,
and to lay charges. Research shows that when police attend calls of domestic violence, they respond
in such a way that minimizes the danger present at the time, often by separating the parties and,
sometimes, arresting the aggressor. Rarely, however, are these decisions based on considerations of
long-term risk (Hoyle, 2008; Hoyle & Sanders, 2000; Trujillo & Ross, 2008). Police tend to use
situational decision-making, focusing on the immediate threat of danger should the two parties remain
in close proximity. This approach is consistent with traditional reactionary policing, whereby officers
respond to situations after they have happened, with evidence and facts playing a prominent role in the
investigation process.
This reactive tendency was observed in the current sample, whereby police action relied heavily on
what they saw before them at the time of the incident. In fact, almost half of the variance in the
decision to arrest the suspect came from two situational variables. The strongest predictor of arrest
was whether the victim requested that the suspect be arrested, in which case the police would almost
always oblige. Similarly, there were many cases in which the suspect was not arrested at the victim’s
request, in spite of the fact that the police would have been justified in pressing charges. The next best
predictor of arrest, consistent with past research (Hilton et al., 2004; 2010b), was observable injury to
the victim. The more serious the victim’s injury, the more likely the police were to arrest the suspect.
This practice makes sense and is what would be expected, but the broader risk context must still be
considered in an officer’s response. Sometimes injury was not inflicted in high risk cases, yet the threat
of harm to the victim remained high. By intervening more formally and intensely situations involving
high risk perpetrators (even in incidents in which injury has not occurred), police could potentially
prevent a serious or lethal injury from taking place in the future.
Exclusive use of reaction-based policing contradicts the newly emerging model of intelligence-led
policing, which emphasizes future behaviours and responding with proactive interventions (Carter &
Carter, 2009). By incorporating a risk assessment instrument, such as the ODARA or R-ODARA, into
their response, police officers enhance their ability to identify high-risk cases in the absence of
observable evidence, such as injury. From a practical view, prior to attending the scene of a suspected
domestic violence call for service officers should obtain background information on the parties involved
either via Canadian Police Information Centre (CPIC) or in-house records. This can be done quickly
while en route to the call via the dispatch centre or mobile data terminal (MDT) in the police car. This
information would be then be used to formulate a preliminary risk assessment and help to inform and
structure potential responses prior to police arrival at the scene. Once on scene, the officer must then
respond to the situational demands at hand, being cognizant of all the contextual variables contributing
to the IPV episode that may modify the initial risk assessment and associated intensity of response.
When subsequently preparing the official police report, the officer would take a few moments to formally
score the risk assessment instrument. Scoring of the ODARA and R-ODARA does not require any
information outside of what police routinely collect when responding to IPV incidents, therefore the time
commitment to complete the assessment is minimal. Once the total score is obtained, the offender’s
risk level should be noted and included in the file so that the next officer to receive a call involving that
party can use the risk information to inform his or her response. It may also be time efficient to create a
mechanism in which in-vehicle computers include software that calculates total risk scores from an
officers’ ratings and includes risk management response options based on the identified IPV risk level.
To ensure its valid use and to maximize its role in decision-making, police officers would be required to
receive training on the administration and scoring of the selected risk tool, its appropriate uses and its
IPV Risk Management and Intervention
The challenge is to determine which police responses, other than arrest, would work to effectively
reduce IPV recidivism. Identifying interventions for intimate partner abusers needs to be done with
consideration of evidence-based practices. Interventions must target factors that have been empirically
linked to the perpetration of partner abuse if they are to have long-term impact (Andrews & Bonta,
2010). Rather than focusing solely on restoring the peace in the moment, a long-term strategy would
be for police to gather and make use of intelligence gained from witnesses, victim, suspect, and prior
dispute calls for service to inform responses that will decrease the likelihood of police having to respond
to subsequent IPV events involving one or both parties. The evidence suggests that rather than
targeting situational factors, which are not linked to IPV offending, it would be more beneficial to target
enduring, long-term personality traits that contribute to IPV behaviour.
From a policing intervention perspective, innovative strategies are required. Davis, Weisburd and
Hamilton (2010) evaluated the effect of a second responder program offered by police for the
prevention of repeat domestic violence. In their study, victims who called the Redlands, California
Police Department to report domestic violence were randomly assigned to receive: a) a second
response within 24 hours, b) a second response within one week, or c) no second response. Reviews
of police records, victim interviews and surveys conducted six months after the initial complaint
revealed no reduction in domestic violence in any of the three conditions. In fact, second-response
conditions were associated with increased domestic violence but the recidivism data is difficult to
interpret because increased reporting could be a function of victims feeling more comfortable reporting
subsequent abuse as a consequence of their increased contact with police.
Stover, Berkman, Desai, and Marans (2010) found that women who received a Domestic Violence
Home Visit Intervention (DVHI) from police following a domestic dispute reported greater satisfaction
with police and were more likely to report future domestic disputes the 12-month follow up period.
Unfortunately, despite having a more positive view of the police than women in the control group, there
were no reductions in severity or frequency of subsequent domestic violence. If the goal is to increase
victim’s willingness to reach out to police for assistance, or to help victims engage with relevant
community resources, then police home visit programs seem to be a viable means of achieving that
outcome. However, if the goal is to facilitate genuine reductions in intimate partner violence, then
interventions need to go much further and address underlying antecedents of the behaviour via greater
community resources and partnerships to provide the necessary interventions.
Cavanaugh, Solomon, and Gelles (2011) conducted a pilot study of a theoretically-grounded
intervention for IPV that uses a psychoeducational and behavioural approach. Dialectical Behaviour
Therapy (DBT) is a treatment approach for individuals who react in a dysfunctional way to intense
emotional experiences that negatively influence their interpersonal relationships and are associated
with problematic personality dispositions (Linehan, 1993; Linehan, Armstrong, Suawrez, Allmon, &
Heard, 1991). This emotional dysregulation leads to a feeling of being out of control, resulting in drastic
coping measures that could include violent and controlling behaviour, substance abuse, and/or selfharm. DBT has demonstrated effectiveness in the treatment of individuals with borderline personality
disorder (BPD; Linehan, 1993; Linehan et al., 1991). As previously discussed, perpetrators of intimate
partner violence share many characteristics with individuals with BPD and BPO (Albertson, 2009;
Dutton, 2005; Ehrensaft et al., 2006; Holtzworth-Munroe, 2000; Marshall & Holtzworth-Munroe, 2010;
Stuart et al, 2006). Conceivably then, DBT might be an effective method of intervention for reducing
risk of recidivism in perpetrators of IPV.
In their evaluation, Cavanaugh et al. (2011) compared a Dialectical Psychoeducational Workshop
(DPEW) to a standard anger management workshop (AMW) delivered in the context of IPV offender
treatment. Self-report questionnaires were given to participants in both programs at baseline and at
program termination. There was a significant difference in all post-treatment scores, whereby the
DPEW group showed significant improvements in adaptive coping skills, anger management skills,
empathy skills, and decreased potential risk for expressions of physical violence relative to the AMW
group. Given that their study was limited by a small sample size and the absence of data on actual
recidivism, it is unclear whether this type of program would lead to a veritable reduction in partner
violence. However, it is certainly promising given that improvements were noted in those personality
features that have been empirically linked to IPV offending behaviour.
The delivery of clinical IPV intervention program requires training and staffing resources that go beyond
those of most policing agencies, which is where partnerships with Public Safety, Social Services,
Mental Health, and non-profit community agencies become crucial. Police can become a voice in
advocating for better evidence-based IPV interventions in their communities so that these communities
have resources to which they can refer both perpetrators and victims. Integrated community responses
to IPV may be more effective in reducing perpetrator violence and increasing victim safety than
traditional reactive policing strategies because these strategies address multiple aspects of the problem
(Alpert & Moore, 1993; Pennington-Zoellner, 2009; Reuland, Schaefer Morabito, Preston & Cheney,
No single agency can provide all of the services required to meet the diverse needs of victims and
perpetrators. Research by Horwitz et al. (2011) provided police officers with an opportunity to
participate in a focus group to discuss their experiences with, opinions of, and responses to, domestic
violence. Officers viewed themselves as “one thread in a complex interwoven fabric, limited by scope
of practice to make long-term changes without an effective link to other professionals” (p. 623). Officers
expressed the need for greater collaboration between the criminal justice system and community
service providers. Different agencies that focus on individual aspects of the IPV issue should come
together to form multidisciplinary teams to case manage victims and perpetrators and deliver
appropriate support services. An example of such a team could include representatives from the
police, public safety, emergency shelters, mental health services, domestic violence outreach services,
and intervention programs for perpetrators.
There are a number of jurisdictions throughout the United States that have developed coordinated
responses between police and women’s shelters or other family violence outreach agencies (Reuland
et al., 2006). For example, police in Arlington, TX, Huntsville, AL, and Broward County, FL, engage in
on-scene collaboration with IPV support workers who are available 24-hours a day to attend domestic
violence calls with officers. The support workers are dispatched at the request of officers and provide
crisis intervention, information to victims about services available to them, assistance with safety
planning, helping with witness statements, transportation of victims and their children or pets to
shelters, and follow-up support and referrals for legal, financial or counseling agencies. The presence
of victim support workers allow police to concentrate on investigation of the crime that took place and
dealing with the perpetrator. Case studies of these programs have found them to be successful in
achieving the goals of enhanced victim services and safety, while less success has been noted for
reductions in the frequency of domestic violence incidents (Reuland et al., 2006).
Another innovative approach to IPV responding was implemented in 2000 in the Logan River Valley,
Queensland, Australia (Foelz, 2002). The Queensland Police Service, in conjunction with community
agencies, developed an integrated community response to domestic violence. One phase of this
response strategy is known as the Fax-Back project. When police attend a call of domestic violence,
the officer asks the victim if he/she wants to be contacted by a support worker at an outreach agency.
If the victim says yes, then they must sign the “fax-back” referral form to authorize police to fax the form
to the outreach service. Police also leave the victim with a brochure about the service, general
information on IPV, and useful phone numbers. When the outreach worker receives the fax from
police, they contact the victim by phone as soon as possible after receiving the referral (usually within
24 hours). The outreach worker talks with the victim to learn his or her specific needs and concerns
with the goal of linking the victim with the services relevant to their needs. The fax-back worker
provides support and information on legal services, housing/emergency accommodation, income
support, child services, counseling, and assistance with the development of a safety plan. The
rationale is that victim feelings of fear and isolation will diminish through being connected with a
supportive network of resource providers. Police are not given the option of taking a discretionary
approach in delivering this response. The fax-back service is offered in all cases of IPV, regardless of
the details of the situation at hand. One unpublished external evaluation of the program (Elliott, 2001)
reported that the service was effective in improving safety and security of victims, assisting victims with
addressing their needs for support, and enhancing victim’s access to resources.
Although attending to the needs of the victim is a critical aspect of the IPV problem, the initiatives
described above focus on victim support and are, therefore, unlikely to directly achieve reductions in
IPV perpetration by offenders. Hagemann-White (2006) argued that although a necessary part of the
response to IPV, providing support and protection for victims is not sufficient to reduce future violence.
In order to achieve reductions in IPV, responses must also contain a rehabilitative component through
interventions such as perpetrator education programs (Hoyle & Sanders, 2000). Depending upon
whether a perpetrator poses an immediate risk to the victim, court mandated programs can be
implemented as a sentencing option. Other sentencing options (e.g., custodial sentences) do not
reduce recidivism and victims often want the suspect to “get help” rather than be punished (Barrett et
al., 2011).
IPV interventions are generally conducted in a group format and are derived from one of two models:
the feminist perspective or the cognitive-behavioural therapy (CBT) model (Healey, Smith, & O’Sullivan,
1998). Despite the overwhelming amount of evidence that domestic violence is not solely a maleperpetrated phenomenon, this awareness has not yet translated into the clinical practice of delivering
gender-inclusive interventions for IPV offenders. The majority of contemporary psychoeducational IPVrelated programs remain grounded in feminist theory. Feminist-based IPV treatment programs, such as
the popular Duluth model, posit that men abuse women as a means of exerting their power and control.
Men are encouraged to critically examine their sexist assumptions and the methods they use to control
their partners and work towards changing these values and beliefs.
Despite the fact that the Duluth model has received virtually no empirical support, it remains the most
commonly used, court sanctioned intervention for male domestic batterers in Canada and the United
States (Corvo, Dutton & Chen, 2009). This model targets stereotypes and social beliefs, while leaving
the underlying antecedents of partner abuse, namely dysfunctional emotions and cognitions,
unexamined (Dutton, 2002). Additionally, this model fails to account for female partner violence against
males, or abuse within same-sex relationships.
In contrast to the Duluth model, interventions based on the CBT model are gender neutral and rest on
assumptions that IPV perpetrators have deficits in anger control, relationship skills, and communication.
These programs teach conflict-reducing communication, anger management strategies, and
relationship skills. There have been few empirical evaluations of the effectiveness of various types of
IPV offender programs. The few studies that have been conducted generally report small, nonsignificant effects in reducing IPV. Babcock, Green, and Robie (2004) conducted a meta-analysis of 22
batterer intervention studies that included comparison control groups. They concluded that men
referred to intervention programs as part of their criminal justice sanction were 5% less likely to
reoffend than men who received traditional sanctions (probation or community service). There was no
difference found in recidivism rate as a function of the type of therapy (feminist vs. CBT). Feder and
Wilson (2005) conducted a subsequent meta-analysis on only the 10 most rigorous studies, which
included control groups, randomization and official reports to measure recidivism. They found a 7%
reduction in recidivism for offenders in batterer treatment programs; however, when partner reports of
recidivism were used as the outcome measures rather than official police reports, there was no
difference between the treatment group and controls. Thus, more research is required to develop
evidence-based IPV interventions.
One possible contribution to the low effectiveness of IPV interventions may relate to the ongoing
controversy over whether participation in these programs should be mandated or voluntary
(Hagemann-White, 2006; Stuart, Temple & Moore, 2007). Some argue that perpetrators who are
forced into treatment may be unwilling or unmotivated to change. Indeed, the current research found
that most offenders denied responsibility for the abuse and, therefore, may be less willing to take steps
to change their behaviour. However, treatment engagement and therapeutic gains may be enhanced
by incorporating motivational interviewing and non-confrontational strategies into these programs in an
effort to meet clients at their current state of readiness and to assist them in realizing their own reasons
for change.
Methods of increasing motivation are based on the transtheoretical model of behaviour change
developed by Prochaska and DiClemente (1983), which proposes that change is a process that
involves at least five stages starting with pre-contemplation (no consideration of change), through
contemplation, preparation, action, and maintenance (finding ways to prevent relapse). Miller and
Rollnick (1991) used the transtheoretical model as a foundation from which they developed motivational
interviewing. The stages of change model (SCM) was originally implemented as a method of treating
addictions; however, Murphy and Baxter (1997) took these concepts and principles and applied them to
the treatment of IPV. Program facilitators trained in motivational interviewing techniques foster a
supportive relationship with abusers, which has the effect of decreased defensiveness and increased
willingness to explore the need for change. Cismaru and Lavack (2011) reviewed 16 IPV interventions
in Canada, the United States, the UK, and Australia and concluded that the most successful programs
were those that emphasized the beneficial reasons for perpetrators to improve their domestic
relationships and focused on enhancing their confidence in their ability to change. Research has
shown that this approach can be integrated fairly easily into current IPV treatment programs and has
the effect of increasing session attendance and reducing post-treatment recidivism (Taft, Murphy,
Musser, & Remington, 2004).
As IPV is often addressed only after serious injury occurs to the victim and/or police become involved
with the perpetrator, the majority of individuals in treatment for abusive behaviour are court mandated
to attend (Gondolf, 2002). Roffman, Edleson, Neighbors, Mbilinyi, and Walker (2010) proposed a
protocol for encouraging self-referrals to treatment by abusers who have not yet become involved with
the criminal justice system. Roffman et al. outlined how the SCM has been applied to addictions
treatment programs and has been successful with individuals who are contemplating, but not yet
committed to, behaviour change. They argued that these strategies may be promising for improving
outcomes with perpetrators of IPV. Follow up research examined a community sample of nontreatment-seeking IPV offenders and found motivational enhancement therapy (MET) to be effective in
prompting non-court-mandated perpetrators to self-refer into treatment (Mbiliny, Neighbors, Walker,
Roffman, Zegree, Edleson et al., 2011).
Mbiliny et al (2011) recruited men through various forms of media advertisement aimed at capturing the
attention of IPV perpetrators who were concerned about their behaviour, but who were not currently
involved in criminal justice proceedings. Men in the MET group received a telephone-delivered,
individual feedback counselling session using motivational interviewing techniques. Personalized
feedback was provided, focusing on specific behaviours that the client had reported and the definite
consequences of those behaviours for the individual’s life and family. In contrast, the control group
received educational material by mail that discussed general consequences of domestic violence, but
provided no personal feedback for individual circumstances. Participants receiving MET showed
superior motivation to change, increased willingness to attend treatment, and a greater reduction in
self-reported IPV. MET has the innovative potential to reach an underserved population of IPV
perpetrators and circumvent the continuation or escalation of their violent behaviour prior to it reaching
the level of the criminal justice system.
Given that IPV occurs within a dyad and often in response to relationship discord, there is a growing
body of evidence that IPV treatment approaches should involve both the victim and the perpetrator
(Hamel & Nicholls, 2007; O’Farrell & Fals-Stewart, 2006; O’Leary & Cohen, 2007; Stith, Rosen, &
McCollum, 2004). Treating the perpetrator without involving the victim may have less of an impact than
treating the pair as a unit to address the relationship issues (Stuart, Temple & Moore, 2007). Stith et al.
(2004) reported significant improvements in marital satisfaction and attitudes about partner abuse and
significant reductions in aggression among participants in a multi-couple IPV treatment group compared
to a no-treatment control group. The same attitude improvements were not reported in the individual
couples-therapy group (i.e., one couple per therapist as opposed to a group of couples); however,
when partner reports of reabuse was used as the dependent measure, the researchers found that men
who participated in either of the two couples treatment programs were less likely to recidivate than men
in the no-treatment group at both the 6-month and 2-year follow up.
O’Leary and Cohen (2007) argued that the best candidates for couples IPV intervention are those in
relationships characterized by psychological aggression and mild physical aggression. For these
individuals, physical aggression is often confined to the relationship and their relationships tend to be
characterized by problematic communication methods and poor anger control. Couple therapy may be
most appropriate in situations in which the couple has a history of low or moderate levels of violence,
and the victim independently agrees to participate and does not express fear of consequences for
openly discussing the problems in the relationship (Stuart et al., 2007). Indeed, the principles of SCM
outlined above may be applicable to both victims and perpetrators. Brown (1997) argued that victims of
IPV go through various stages of contemplation regarding their decision to remain in the violent
relationship and to seek different levels of support. Programs aimed at empowering victims to regain
control and/or leave a violent relationship may achieve greater success by incorporating principles of
SCM and MET into their agendas.
In summary, the evidence presented in the current study, along with results from previous research
(Holtzworth-Munroe & Stuart, 1994; Stuart, 2005) make it clear that IPV offenders are a heterogeneous
group for which no one-size-fits-all method of intervention is appropriate or effective. The motives for
which people commit IPV are abundant and varied and perpetrators differ in their risk of being violent in
the future. For some perpetrators, committing an abusive act against a partner is an isolated episode
for which they feel great remorse. For others, violent behaviour is a deep-rooted response to negative
emotions and occurs with regularity in their interpersonal relationships. The types of interventions best
suited to reducing violent behaviour in these varied circumstances must be tailored to the needs of the
respective perpetrators and victims involved.
Risk assessment tools, such as the ODARA, are able to differentiate between sub-groups of IPV
offenders based on their level of need for intervention. Those categorized as low risk may require
minimal, or no intervention, as they are unlikely to reoffend even in the absence of treatment. Those
categorized as moderate risk may be best suited to cojoint therapy or psychoeducational workshops
that incorporate motivational interviewing techniques to increase the client’s recognition of the problems
and their desire to change. Interventions using an intensive DBT approach targeting behavioural
change through the reparation of long-standing, dysfunctional patterns of emotion regulation and
problematic personality characteristics may be best suited to high-risk offenders.
All of these interventions must be delivered within the context of appropriate criminal justice sanctions
in order to hold the offender accountable for the crime he or she committed and to ensure the safety of
the victim. In some cases, probation or community service may be an appropriate penalty but
protective orders or custodial sentences should be imposed on higher risk offenders. In cases of lowrisk individuals with no prior criminal history, community service or monetary fines may suffice. It is
important to emphasize that any act of violence against another person is a crime and perpetrators
must be held accountable for this behaviour; however, the criminalization of IPV can be upheld within
the context of a rehabilitative model. Penalties should be imposed for the behaviour in conjunction with
the delivery of risk-reducing interventions.
Police and probation officers must engage in strict monitoring of moderate- and high-risk offenders to
ensure their compliance with conditions, including protective orders and court-ordered batterer
programs. Those who do not comply should be subject to additional, more severe consequences. The
more information sharing that occurs among agencies, such as probation, police and victim services,
the more effective this monitoring will be in the end. Offenders must be held accountable for their
actions and victims deserve to feel validated and protected by the criminal justice system. Proactive
police engagement strategies that build relationships with victims and perpetrators may facilitate their
access to, and participation in, appropriate community interventions to reduce IPV risk for both parties.
The current study demonstrated that the ODARA was effective in distinguishing between low,
moderate, and high risk male and female IPV perpetrators, regardless of whether the violence was
physical or non-physical. Although ODARA identified high risk perpetrators were similar to low and
moderate risk offenders in the level of violence that they attempted to inflict on their partners in a given
incident, it was the high risk offenders who caused the greatest level of injury to their victim. The
disconnect between attempted violence and risk level may be due to the fact that the nature of the
violence committed is driven by situational factors relevant during a given IPV incident, while the
perpetrator’s propensity to cause harm is driven more by stable, personality-based characteristics. The
current research also found that female perpetrators of IPV were similar to male perpetrators in their
risk profile and IPV offending patterns, which contradicts feminist theories of IPV. Indeed, female
perpetrators in the current sample engaged in more severe violence than did males. These findings
challenge the utility of traditional feminist-based models of IPV offender intervention (i.e., the Duluth
model), as these treatments are unable to address the issue of female-perpetrated abuse and do not
target factors for treatment that have been empirically linked to IPV offending.
It is noteworthy that police officers’ decision to arrest the perpetrator in the current sample showed no
relationship with the offender’s risk of IPV. This is problematic given evidence in the literature that highrisk offenders are more likely to repeat the violence and to cause harm. It is also of note that arrest had
no effect on recidivism, and would have had minimal impact in changing the outcomes of high risk
offenders anyway. This is not to suggest that arrest should be abandoned as a response to IPV, but
rather it should be used as a first step in a course of action intended to reduce offender risk. From a
police perspective, it is not the immediate response of the officer (e.g., making an arrest) that will lead
to a reduction in IPV offending. Rather, it is the process that the officer initiates with his or her response
that will ultimately lead to change.
The problem of intimate partner violence cannot be eradicated simply by arresting and convicting a
perpetrator, or by empowering a victim through the provision of information, options and resources.
Nevertheless, both of these steps can serve the short-term goal of reducing or eliminating the pressing
danger in a situation and facilitating greater victim engagement. If, however, the perpetrator possesses
certain stable, enduring personality traits, then he or she is a high risk to become violent again
regardless of the situational circumstances at the time. By using a standardized risk-assessment tool
to identify those offenders who are at a high risk of future violence, police can triage cases and facilitate
access for IPV perpetrators who require interventions that will address the core elements contributing to
their violent behaviour. The delivery of these interventions requires collaboration with agencies such as
Mental Health or Public Safety, which employ professionals trained in the delivery of psychoeducational
and DBT focused therapies. In this sense, police become part of an integrated community strategy
grounded in strong partnerships with multiple agencies.
IPV is a multifaceted issue that must be dealt with through the collaborative efforts of a variety of
professional and community services. Law enforcement is an important piece of the solution; however,
the actions of police are of little long-term value if not part of an overall response to IPV that addresses
the underlying causes of the problem.
Ægisdottir, S., White, M. J., Spengler, P. M., Maugherman, A. S., Anderson, L. A., Cook, R. S.,
Nichols, C. N., Lampropoulos, G. K., Walker, B. S., Cohen, G., & Rush, J. D. (2006).
The meta-analysis of clinical judgment project: Fifty-six years of accumulated research
on clinical versus statistical prediction. The Counselling Psychologist, 34(3), 341-382.
doi: 10.1177/0011000005285875
Albertson, E. K. (2009). Females arrested for intimate partner violence: Insecure attachment
and borderline personality organization. (Doctoral Dissertation). Retrieved from
Dissertations & Theses. (Publication No. AAT NR01745).
Alpert, G., & Moore, M. (1993). Measuring police performance in the new paradigm of policing.
In J. J. Dilulio (ed.). Performance Measures for the Criminal Justice System. Bureau of
Justice Statistics – Princeton Project. Washington, D.C: U.S Dept. of Justice, Office of
Justice Programs, Bureau of Justice Statistics.
Andrews, D. A., & Bonta, J. (2010). Rehabilitating criminal justice policy and practice.
Psychology, Public Policy, and Law, 16(1), 39-55. doi: 10.1037/a0018362
Andrews, D. A., Bonta, J. & Hoge, R. D. (1990). Classification for effective rehabilitation:
Rediscovering psychology. Criminal Justice and Behaviour, 17(1), 19-52. doi:
Au, A., Cheung, G., Kropp, R., Yuk-chung, C., Lam, G., & Sung, P. (2008). A preliminary
validation of the brief spousal assault form for the evaluation of risk (B-SAFER) in
Hong Kong. Journal of Family Violence, 23(8), 727-735. doi: 10.1007/s10896-0089198-z
Avakame, E. F., & Fyfe, J. J. (2001). Differential police treatment of male-on-female spousal
violence: Additional evidence on the leniency thesis. Violence Against Women, 7(1), 2245. doi: 10.1177/10778010122182280
Babcock, J. C., Green, C. E., & Robie, C. (2004). Does batterer’s treatment work? A metaanalytic review of domestic violence treatment. Clinical Psychology Review, 23(8), 10231053.
Bannerman, C. W. (2002). Baker One Domestic Violence Intervention Project: Improving
Response to Chronic Domestic Violence Incidents. Charlotte-Mecklenburg (North
Carolina) Police Department.
Barrett, B.J., St. Pierre, M., & Vaillancourt, N. (2011). Police response to intimate partner
violence in Canada: Do victim characteristics matter? Women & Criminal Justice, 21(1),
38-62. DOI: 10.1080/08974454.2011.536057
Barrett, P. (2001). Assessing the reliability of rating data. Retrieved from
Bartko, J. J., & Carpenter, W. T. (1976). On the methods and theory of reliability. Journal of
Nervous & Mental Disease, 163(5), 307-317.
Belfrage, H., & Strand, S. (2008). Structured spousal violence risk assessment: Combining risk
factors and victim vulnerability factors. International Journal of Forensic Mental Health,
7(1), 39-46. doi: 10.1080/095851800362445
Belknap, J. & Hartman, J. L. (2000). Police responses to woman battering: Victim advocates’
reports. International Review of Victimology, 1-3, 159-77.
Bennett, L., Goodman, L. & Dutton, M. A. (2000). Risk assessment among batterers arrested for
domestic assault. Violence Against Women, 6(11), 1190-1203. doi:
Bennett-Cattaneo, L., Cho, S., & Botuck, S. (2011). Describing intimate partner stalking over
time: An effort to inform victim-centered service provision. Journal of Interpersonal
Violence, 26(17), 3428-3454. doi: 10.1177/0886260511403745
Berk, R. A., Campbell, A., Klap, R., & Western, B. (1992a). Bayesian analysis of the Colorado
Springs spouse abuse experiment. Journal of Criminal Law and Criminology, 83(1),
Berk, R. A., Campbell, A., Klap, R., & Western, B. (1992b). The deterrent effect of arrest in
incidents of domestic violence: A Bayesian analysis of four field experiments. American
Sociological Review, 57(5), 698-708.
Binder, A., & Meeker, J. W. (1988). Experiments as reforms. Journal of Criminal Justice, 16(4),
347-358. doi:10.1016/0047-2352(88)90021-9
Bonta, J. (1996). Risk-needs assessment and treatment. In A. T.Harland (Ed.), Choosing
correctional options that work (pp. 18–32). Thousand Oaks, CA: Sage.
Bonta, J. (2002). Offender risk assessment: Guidelines for selection and use. Criminal Justice
and Behaviour, 29(4), 355-379.doi: 10.1177/0093854802029004002
Bonta, J., & Andrews, D.A. (2007). Risk-need-responsivity model for offender assessment and
rehabilitation (Corrections Research User Report No. 2007-06). Ottawa, Ontario: Public
Safety Canada.
Borum, R. (1996). Improving the clinical practice of violence risk assessment: Technology,
guidelines and training. American Psychologist, 51, 945-956. doi: 10.1037/0003066X.51.9.945
Boyle, D. J., O’Leary, K. D., Rosenbaum, A., & Hassett-Walker, C. (2008). Differentiating
between general and partner-only violent subgroups: Lifetime antisocial behaviour,
family of origin violence, and impulsivity. Journal of Family Violence, 23(1), 47-55. doi:
Brown, J. (1997). Working toward freedom from violence: The process of change in battered
women. Violence Against Women, 3(1), 5-26. doi: 10.1177/1077801297003001002
Brownridge, D. A., Chan, K. L., Hiebert-Murphy, D., Ristock, J., Tiwari, I., Leung, W. C., &
Santos, S. C. (2008). The elevated risk for non-lethal post-separation violence in
Canada: A comparison of separated, divorced, and married women. Journal of
Interpersonal Violence, 23(1), 117-135. doi: 10.1177/0886260507307914
Buzawa, E. S., & Buzawa, C. G. (2003). Domestic Violence: The Criminal Justice Response
(3rd ed.). Thousand Oaks, California: Sage Publications.
Buzawa, E., & Hotaling, G. (2000). The police response to domestic violence calls for
assistance in three Massachusetts towns: Final report. Washington, DC: National
Institute of Justice.
Caetano, R., Schafer, J., & Cunradi, C. B. (2001). Alcohol-related intimate partner violence
among White, Black, and Hispanic couples in the United States. Alcohol Research &
Health¸25, 58-65.
Campbell, J. C. (2002). Health consequences of intimate partner violence. Lancet, 359(9314),
Campbell J, C., & Lewandowski, L. A. (1997). Mental and physical health effects of intimate
partner violence on women and children. Psychiatric Clinics of North America, 20(2),
Campbell, J. C., & Soeken, K. L. (1999). Forced sex and intimate partner violence. Violence
Against Women, 5(9), 1017-1035. doi: 10.1177/1077801299005009003
Campbell, M. A., French, S., & Gendreau, P. (2009). The prediction of violence in adult
offenders: A meta-analytic comparison of instruments and methods of assessment.
Criminal Justice and Behaviour, 36(6), 567-590. doi: 10.1177/0093854809333610
Canadian Center for Justice Statistics (2011). Family Violence in Canada: A statistical profile,
2011. Ottawa: Industry Canada. Retrieved from
Capaldi, D. M., Shortt, J. W., Kim, H. K., Wilson, J., Crosby, L., & Tucci, S. (2009). Official
incidents of domestic violence: Types, injury, and associations with nonofficial couple
aggression. Violence and Victims, 24(4), 502–519.
Carrado, M., George, M. J., Loxam, E., Jones, L., & Templar, D. (1996). Aggression in British
heterosexual relationships: A descriptive analysis. Aggressive Behaviour, 22(6), 401–
415.doi: 10.1002/(SICI)1098-2337(1996)22:6<401::AID-AB1>3.0.CO;2-K
Carter, D. L., & Carter, J. G. (2009). Intelligence-led policing: Conceptual and functional
considerations for public policy. Criminal Justice Policy Review, 20(3), 310-325. doi:
Cattaneo, L. B., Cho, S., & Botuck, S. (2011). Describing intimate partner stalking over time: An
effort to inform victim-centered service provision. Journal of Interpersonal Violence,
26(17), 3428-3454. doi: 10.1177/0886260511403745
Cattaneo, L. B., & Goodman, L. A. (2005). Risk factors for reabuse in intimate partner violence:
A cross-disciplinary critical review. Trauma, Violence, & Abuse, 6(2), 141-175. doi:
Cavanaugh, M., Solomon, P., & Gelles, R. (2011). The dialectical psychoeducational workshop
(DPEW) for males at risk for intimate partner violence: A pilot randomized controlled
trial. Journal of Experimental Criminology, 7(3), 275-291. doi: 10.1007/s11292-0119126-8
Cercone, J. J., Beach, S. R., & Arias, I. (2005). Gender symmetry in dating intimate partner
violence: Does similar behaviour imply similar constructs? Violence and Victims, 20(2),
Cho, H. & Wilke, D. J. (2010). Gender differences in the nature of the intimate partner violence
and effects of perpetrator arrest on revictimization. Journal of Family Violence, 25(4),
393-400. doi: 10.1007/s10896-009-9300-1
Cismaru, M., & Lavack, A. (2011). Campaigns targeting perpetrators of intimate partner
violence. Trauma violence & abuse, 12(4), 183-197. doi: 10.1177/1524838011416376
Coker, A. L., Smith, P. H., Bethea, L., King, M. R., & McKeown, R. E. (2000). Physical health
consequences of physical and psychological intimate partner violence. Archives of
Family Medicine, 9(5), 451-457.
Corvo, K., Dutton, D., & Chen, W.Y. (2009). Do Duluth Model Interventions With Perpetrators
of Domestic Violence Violate Mental Health Professional Ethics? Ethics & Behaviour,
19(4), 323-340. doi: 10.1080/10508420903035323
Crenshaw, K. W. (1994). Mapping the margins: Intersectionality, identity politics and violence,
against women of color. In Fineman, M. & Mykitiuk, R. (Eds.), The Public Nature of
Private Violence. New York, NY: Routledge.
Davis, R., Weisburd, D., & Hamilton, E. (2010). Preventing repeat incidents of family violence:
A randomized field test of a second responder program. Journal of Experimental
Criminology, 6(4), 397-418. doi: 10.1007/s11292-010-9107-3
Department of Justice Canada (2006). Final Report of the Ad Hoc Federal-ProvincialTerritorial Working Group Reviewing Spousal Abuse Policies and Legislation. Retrieved
DeKeseredy, W. S. (2006). Future directions. Violence Against Women, 12, 1078-1085.
Dobash, R. E. & Dobash, R. (1979). Violence against wives. New York: Free Press.
Dobash, R. P., Dobash, R. E., Cavanagh, K., & Lewis, R. (1998). Separate and intersecting
realities: A comparison of men’s and women’s accounts of violence against women.
Violence Against Women, 4(4), 382-414. doi: 10.1177/1077801298004004002
Dobash, R. E., Dobash, R. P., Cavanagh, K., & Medina-Ariza, J. (2007). Lethal and nonlethal
violence against an intimate female partner: Comparing male murderers to nonlethal
abusers. Violence Against Women, 13(4), 329-353. doi: 10.1177/1077801207299204
Domestic Violence and Incest Resource Center (2004). Developing an integrated response to
family violence in Victoria – Issues and Directions. Retrieved from
Douglas, K. S. & Ogloff, J. R. (2003). Multiple facets of risk for violence: The impact of
judgmental specificity on structured decisions about violence risk. International Journal
of Forensic Mental Health, 2(1), 19-34.
Dunford, F. W., Huizinga, D., & Elliott, D. S. (1990). The role of arrest in domestic assault: The
Omaha police experiment. Criminology, 28(2), 183-206.
Dutton, D. G. (1995). The batterer: A psychological profile. New York, NY: Basic Books.
Dutton, D. G. (2002). The abusive personality: Violence and control in abusive relationships (2nd
ed.). New York: Guilford.
Dutton, D. G. (2002). Treatment of assaultiveness. Journal of Aggression, Maltreatment, &
Trauma, 7(1/2), 7-28.
Dutton, D., & Nicholls, T., (2005). The gender paradigm in domestic violence research and
theory: Part 1 - The conflict of theory and data. Aggression and Violent Behaviour,
10(6), 680-714. doi: 10.1016/j.avb.2005.02.001
Dutton, D. G., Nicholls, T. L., & Spidel, A. (2006). Female perpetrators of intimate abuse.
Canada Journal of Offender Rehabilitation,41(4), 1-31. doi: 10.1300/J076v41n04_01
Elliott, A. (2001). Evaluation of the fax-back project – Logan River Valley integrated
community response to domestic violence group. Unpublished manuscript.
Epstein, D. (1999). Redefining the state’s response to domestic violence: Past victories and
future challenges. Georgetown Journal of Gender and the Law, 1, 127-43.
Ehrensaft, M. K., Cohen, P., & Johnson, J. G. (2006). Development of personality disorder
symptoms and the risk for partner violence. Journal of Abnormal Psychology, 115(3),
474-483. doi: 10.1037/0021-843X.115.3.474
Fals-Stewart, W., Golden, J., & Schumacher, J. A. (2003). Intimate partner violence and
substance use: A longitudinal day-to-day examination. Addictive Behaviours, 28(9),
555-1574. doi:10.1016/j.addbeh.2003.08.035
Fals-Stewart W., Leonard, K. E., & Birchler, G. R. (2005). The occurrence of male-to-female
intimate partner violence on days of men’s drinking: The moderating effects of antisocial
personality disorder. Journal of Consulting and Clinical Psychology, 73(2), 239–248.
doi: 10.1037/0022-006X.73.2
Farrell, G., Buck, W., & Pease, K. (1993). The Merseyside domestic violence prevention project:
some costs and benefits. Studies of Crime and Prevention, 2, 21-33.
Feder, L. & Wilson, D. (2005). A meta-analytic review of court-mandated batterer intervention
programs: Can courts affect abusers’ behaviour? Journal of Experimental Criminology,
1(2), 239-262. doi: 10.1007/s11292-005-1179-0
Felson, R. B., & Lane, K.J. (2010). Does violence involving women and intimate partners have a
special etiology? Criminology, 48(1), 321-338. doi: 10.1111/j.1745-9125.2010.00186.x
Florsheim, P., & Moore, D. R. (2008). Observing differences between healthy and unhealthy
adolescent romantic relationships: Substance abuse and interpersonal process. Journal
of Adolescence, 31(6), 795-814. doi: 10.1016/j.adolescence.2007.09.005
Foa, E. B., Cascardi, M., Zoellner, L.A, & Feeny, N.C (2000). Psychological and environmental
factors associated with partner violence. Trauma, Violence, and Abuse, 1(1), 67-91. doi:
Foelz, M. (2002). Getting connected in Lovan River Valley: Police and community services
coordinated response to domestic violence. Retrieved from
Frye, V. Manganello, J., Campbell, J. C., Walton-Moss, B., & Wilt, S. (2006). The distribution
of and factors associated with intimate terrorism and situational couple violence among a
population-based sample of urban women in the United States. Journal of Interpersonal
Violence, 21(10), 1286-1313. doi: 10.1177/0886260506291658
Garcia-Moreno, C., Jansen, H., Ellsberg, M., Heise, L., & Watts, C. H. (2006). Prevalence of
intimate partner violence: findings from the WHO multi-country study on women’s
health and domestic violence. Lancet ,368(9543), 1260-1269. doi: 10.1016/S01406736(06)69523-8
Gardner, W. (1995). Statistical Methods for Rates and Proportions. New York: John Wiley &
Sons. (Second Edition).
Garner, J. H., & Maxwell, C. D. (2009). Prosecution and conviction rates for intimate partner
violence. Criminal Justice Review, 34(1), 44-79. doi: 10.1177/0734016808324231
Gendreau, P., Little, T., & Goggin, C. (1996). A meta-analysis of the predictors of adult offender
recidivism: What works! Criminology, 34(4), 575-607.
Golding, J. M. (1999). Intimate Partner Violence as a Risk Factor for Mental Disorders: A MetaAnalysis. Journal of Family Violence, 14(2), 99-132. doi: 10.1023/A:1022079418229
Golinelli, D., Longshore, D., & Wenzel, S. L. (2008). Substance use and intimate partner
violence: Clarifying the relevance of women's use and partner's use. Journal of
Behavioural Health Services and Research, 36(2), 199-211. doi: 10.1007/s11414-0089114-6
Gondolf, E. W., & White, R. J. (2001). Batterer program participants who repeatedly reassault:
Psychopathic tendencies and other disorders. Journal of Interpersonal Violence, 16(4),
361-380. doi: 10.1177/088626001016004006
Gondolf, E. W. (2002). Batterer intervention systems: Issues, outcomes and recommendations.
Thousand Oaks, CA: Sage.
Graham-Kevan, N. (2009). The psychology of women’s partner violence: Characteristics and
cautions. Journal of Aggression, Maltreatment and Trauma, 18(6), 587-603. doi:
Graham-Kevan, N. & Archer, J. (2008). Does controlling behaviour predict physical aggression
and violence to partners? Journal of Family Violence, 23(7), 539-548. doi:
Grann, M. & Wedin, I. (2002). Risk factors for recidivism among spousal assault and spousal
homicide offenders. Psychology, Crime & Law, 8(1), 5-23. doi:
Grove, W. M., & Meehl, P. E. (1996). Comparative efficiency of informal (subjective,
impressionistic) and formal (mechanical, algorithmic) prediction procedures: The
clinical–statistical controversy. Psychology, Public Policy, and Law, 2(2), 293-323. doi:
Grove, W. M., Zald, D. H., Lebow, B. S., Snitz, B. E., & Nelson, C. (2000). Clinical versus
mechanical prediction: A meta-analysis. Psychological Assessment, 12(1), 19-30. doi:
Gunderson, J. G. (1984). Borderline personality disorder. Washington, DC: American
Psychiatric Press.
Hagemann-White, C. (2006). Combating Violence Against Women. Stocktaking Study on the
Measures and Actions Taken in Council of Europe Member States. Strasbourg
Directorate General of Human Rights, Council of Europe. Retrieved from
Hamel, J. (2011). In dubious battle: The politics of mandatory arrest and dominant aggressor
laws. Partner Abuse, 2(2), 224-245. doi: 10.1891/1946&U8211;6560.2.2.224
Hamel, J., & Nicholls, T. (2007). Family Interventions in Domestic Violence: A Handbook of
Gender-Inclusive Theory and Treatment. New York, NY, USA, Springer Publishing
Hare, R. D. (2003). The Hare Psychopathy Checklist-Revised (PCL-R) (2nd ed.). Toronto,
Canada: Multi-Health Systems.
Harned, M. S. (2001). Abused women or abused men? An examination of the context and
outcomes of dating violence. Violence and Victims, 16(3), 269-285.
Harris, G. T., Hilton, N, Z., & Rice, M. E. (2011). Explaining the frequency of intimate partner
violence by male perpetrators: Do attitude, relationship, and neighborhood variables add
to antisociality? Criminal Justice and Behavior, 38(4), 309-331. doi:
Harris, G. T., Rice, M. E., & Quinsey, V. L. (1993). Violent recidivism of mentally disordered
offenders: The development of a statistical prediction instrument. Criminal Justice and
Behaviour, 20(4), 315-335. doi: 10.1177/0093854893020004001
Hartman, J. L., & Belknap, J. (2003). Beyond the gatekeepers: Court professionals' selfreported attitudes about and experiences with misdemeanour domestic violence cases.
Criminal Justice and Behaviour, 30(3), 349-373. doi: 10.1177/0093854803030003005
Healey, K., Smith, C. & O’Sullivan, C. (1998). Batterer Intervention: Program Approaches and
Criminal Justice Strategies. Washington, D.C.: U.S. Department of Justice.
Heilbrun, K. (1997). Prediction versus management models relevant to risk assessment: The
importance of legal decision-making context. Law and Human Behaviour, 21(4), 347359. doi: 10.1023/A:1024851017947
Henning, K. Jones, A., & Holdford, R. (2003). Treatment needs of women arrested for domestic
violence: A comparison with male offenders. Journal of Interpersonal violence, 18(8),
839-856. doi: 10.1177/0886260503253876
Henning, K., Jones, A., & Holdford, R. (2005). “I didn’t do it, but if I did I had a good reason”:
Minimization, denial, and attributions of blame among male and female domestic
violence offenders. Journal of Family Violence, 20(3), 131-139. doi: 10.1007/s10896005-3647-8
Henning, K., Martinsson, R., & Holdford, R. (2009). Gender differences in risk factors for
intimate partner violence recidivism. Journal of Aggression, Maltreatment & Trauma,
18(6), 623-645. doi: 10.1080/10926770903103248
Hilton, N. Z., Harris, G. T., Popham, S., & Lang, C. (2010a). Risk assessment among
incarcerated male domestic violence offenders. Criminal Justice and Behaviour, 37(8),
815-832. doi: 10.1177/0093854810368937
Hilton, N. Z., Harris, G. T., & Rice, M. E. (2010b). Risk Assessment for Domestically Violent
Men: Tools for Criminal Justice, Offender Intervention, and Victim Services.
Washington, D.C., American Psychological Association.
Hilton, N. Z., Harris, G. T., Rice, M. E., Houghton, R. E., & Eke, A. W. (2008). An indepth
actuarial assessment for wife assault recidivism: The domestic violence risk appraisal
guide. Law and Human Behaviour, 32(2), 150-163. doi: 10.1007/s10979-007-9088-6
Hilton, N. Z., Harris, G. T., Rice, M. E., Lang, C., Cormier, C. A., & Lines, K. J. (2004). A brief
actuarial assessment for the prediction of wife assault recidivism: The Ontario domestic
assault risk assessment. Psychological Assessment, 16(3), 267-275. doi:
10.1037/1040- 3590.16.3.267
Holtzworth-Munroe, A., (2000). A typology of men who are violent toward their female
partners: making sense of the heterogeneity in husband violence. Current Directions in
Psychological Science, 9(4), 140-143. doi: 10.1111/1467-8721.00079
Holtzworth-Munroe, A. & Stuart, G. L. (1994). Typologies of male batterers: Three subtypes and
the differences among them. Psychological Bulletin, 116(3), 476-497.
Horwitz, S. H., Mitchell, D., La Russa-Trott, M., Santiago, L., Pearson, J., Skiff, D. M., & Cerulli,
C. (2011). An inside view of police officers’ experience with domestic violence.
Journal of Family Violence, 26(8 ), 617-625. doi: 10.1007/s10896-011-9396-y
Howe, M. E., & Alpert, E. J. (2009). A public health approach to intimate partner violence. In C.
Mitchell, & D. Anglin (Eds.), Intimate partner violence: A health based perspective (pp.
275-287). New York: Oxford University Press, Inc.
Hoyle, C. (2008). Will she be safe? A critical analysis of risk assessment in domestic violence
cases. Children and Youth Services Review, 30(3), 323-337.
Hoyle, C. & Sanders, A. (2000). Police response to domestic violence: From victim choice to
victim empowerment? British Journal of Criminology, 40(1), 14-36. doi:
Hughes, F. M., Stuart, G. L., Gordon, K. C., & Moore, T. M. (2007). Predicting the use of
aggressive conflict tactics in a sample of women arrested for domestic violence. Journal
of Social and Personal Relationships,24(2), 155-176. doi: 10.1177/0265407507075406
Huss, M. T., & Langhinrichsen-Rohling, J. (2000). Identification of the psychopathic batterer:
The clinical, legal and policy implications. Aggression and Violent Behaviour, 5(4), 403422. doi: 10.1016/S1359-1789(98)00038-X
Jackson, R. L., & Richards, H. J. (2007). Psychopathy and the five factor model: Self and
therapist perceptions of psychopathic personality. Personality and Individual Differences,
43(7), 1711-1721. doi: 10.1016/j.paid.2007.05.009
Johnson, I. M. (1990). A loglinear analysis of abused wives’ decisions to call the police in
domestic-violence disputes. Journal of Criminal Justice, 18(2), 147-159.
Jolin, A. & Moose, C. A. (1997). Evaluating a domestic violence program in a community
policing environment: Research implementation issues. Crime & Delinquency, Vol 43(3),
Kingsnorth, R. (2006). Intimate Partner Violence: Predictors of recidivism in a sample of
arrestees. Violence Against Women, 12(10), 917-935. doi: 10.1177/1077801206293081
Kropp, P. R., Hart, S. D., Webster, C. D., & Eaves, D. (1999). Manual for the Spousal Assault
Risk Assessment Guide (3rd ed.). Toronto, Canada: Multi-Health Systems.
Kropp, P. R. & Hart, S. D. (2000). The spousal assault risk assessment (SARA) guide:
Reliability and validity in adult male offenders. Law and Human Behaviour, 24(1), 101118. doi: 10.1023/A:1005430904495
Kropp, P. R., & Hart, S. D. (2004). The development of the Brief Spousal Assault Form for the
Evaluation of Risk (B-SAFER): A tool for criminal justice professionals. Ottawa, ON,
Canada: Department of Justice (Government of Canada).
Landis, J. R., & Koch, G. G. (1977). The measurement of observer agreement for categorical
data. Biometrics, 33, 159-74.
Lempert, R. O. (1989). Humility is a virtue: On the publicization of policy-relevant research.
Law and Society Review, 23(1), 145-161.
Leisenring, A. (2008). Controversies surrounding mandatory arrest policies and the police
response to intimate partner violence. Sociology Compass, 2(2), 451-466. doi:
Lie, G. Y., & Gentlewarrier, S. (1991). Intimate violence in lesbian relationships: discussion of
survey findings and practice implications. Journal of social service research, 15, 41-59.
Lie, G. Y., Schilit, R., Bush, J., Montagne, M., & Reyes, L. (1991). Lesbians in currently
aggressive relationships: How frequently do they report aggressive past relationships?
Violence and Victims, 6, 121-135.
Linehan, M. M. (1993). Skills training manual for treating borderline disorder. New York:
Guilford Press.
Linehan, M. M., Armstrong, H. E., Suarez, A., Allmon, D., & Heard, H. L. (1991). Cognitivebehavioural treatment of chronically parasuicidal borderline patients. Archives of General
Psychiatry, 48(12), 1060-1064.
Loesel, F. (1995). What recent meta-evaluations tell us about the effectiveness of correctional
treatment. In G. Davies & S. Lloyd-Bostock (Eds.), Psychology, law, and criminal justice:
International developments in research and practice (pp. 537-554). Oxford, UK: Walter
De Gruyter.
Luke, D. A., & Homan, S. M. (1998). Time and change: Using survival analysis in clinical
assessment and treatment evaluation. Psychological Assessment, 10(4), 360-378. doi:
Marshall, A. D., & Holtzworth-Munroe, A. (2010). Recognition of wives’ emotional xpressions:
A mechanism in the relationship between psychopathology and intimate partner violence
perpetration. Journal of Family Psychology, 24(1), 21–30. doi: 10.1037/a0017952
Matthews, D. D. (Ed.). (2004). Domestic Violence Sourcebook (2nd ed.). Detroit: Omnigraphics.
Mbilinyi, L. F., Neighbors, C., Walker, D. D., Roffman, R. A., Zegree, J., Edleson, J., &
O’Rourke, A. (2011). A telephone intervention for substance-using adult male
perpetrators of intimate partner violence. Research on Social Work Practice, 21(1), 4356. doi: 10.1177/1049731509359008
McCrae, R. R., & John, O. P. (1992). An introduction to the five-factor model and its
applications. Journal of Personality, 60, 175-215.
McGraw, K. O., & Wong, S. P. (1996). Forming inferences about some intraclass
correlation coefficients. Psychological Methods, 1(1), 30-46. doi: 10.1037/1082989X.1.1.30
Medeiros, R. A., & Straus, M. A. (2006). Risk factors for physical violence between dating
partners: Implications for gender inclusive prevention and treatment of family violence.
In J. Hamel & T. Nicholls (Eds.), Family Interventions in Domestic Violence: A
Handbook of Gender-Inclusive Theory and Treatment (pp. 59-86). Springer Publishing
Company, New York, NY.
Mederer, H. J., & Gelles, R. J. (1989). Compassion or control: Intervention in cases of wife
abuse. Journal of Interpersonal Violence, 4(1), 25-43. doi:
Mele, M. (2009). The time course of repeat intimate partner violence. Journal of Family
Violence, 24(8), 619-624. doi: 10.1007/s10896-009-9260-5
Ménard, K. S., Anderson, A. L., & Godboldt, S. M. (2009). Gender differences in intimate
partner recidivism: A 5-year follow-up. Criminal Justice and Behaviour, 36(1), 61-76.
Messing, J. T. (2007). Assessing the risk: What police reports reveal about domestic violence
escalation (Doctoral Dissertation). Available from ProQuest Dissertations and Theses
database. (UMI No. 3275518).
Miller, S. (2001). The paradox of women arrested for domestic violence: Criminal justice
processionals and services providers respond. Violence against women, 7(12), 13391379.
Miller, J. D., Lyman, D. R., Widiger, T. A., & Leukefeld, C. (2001). Personality disorders as
extreme variants of common personality dimensions. Can the five factor model
adequately represent psychopathy? Journal of Personality, 69(2), 253-277. doi:
Miller, W. R., & Rollnick, S. (1991). Motivational Interviewing: Preparing People to Change
Addictive Behaviour. New York, NY, Guilford Press.
Moffitt, T. E., Robins, R. W., & Caspi, A. (2001). A couples analysis of partner abuse with
implications for abuse-prevention policy. Criminology and Public Policy, 1(1), 5-37. doi:
Monahan, J. (1981). Predicting Violent Behaviour: An Assessment of Clinical Techniques.
Beverly Hills, CA: Sage.
Mossman, D. (1994). Assessing predictions of violence: Being accurate about accuracy. Journal
of Consulting and Clinical Psychology, 62(4), 783-792. doi: 10.1037/0022- 006X.62.4.783
Mossman, D., & Somoza, E. (1991). ROC curves, test accuracy, and the description of
diagnostic tests. The Journal of Neuropsychiatry and Clinical Neurosciences, 3(5), 330333.
Murphy, C. M., & Baxter, V. A. (1997). Motivating batterers to change in the treatment context.
Journal of Interpersonal Violence, 12(4), 607-619.
Naumann, P., Langford, D., Torres, S., Campbell, J., & Glass, N. (1999). Women battering in
primary care practice. Family Practice, 16(4), 343-352.
Nicholls, T. L. (2011, June). The emperor has no clothes: The glaring gap between what we
know and what we deliver in the domestic violence field. Keynote speech presented at
the 11th Annual International Association of Forensic Mental Health Services
Conference, Barcelona, Spain.
Nixon, R. D., Resick, P. A., & Nishith, P. (2004). An exploration of comorbid depression among
female victims of intimate partner violence with posttraumatic stress disorder. Journal of
Affective Disorders, 82(2), 315-320. doi: 10.1016/j.jad.2004.01.008
Norlander, B., & Eckhardt, C. (2005). Anger, hostility, and male perpetration of intimate partner
violence: A meta-analytic review. Clinnical Psychology Review, 25(2), 119-152.
O’Farrell, T. J. & Fals-Stewart, W. (2006). Behavioural Couples Therapy for Alcoholism and
Drug Abuse. New York, NY: Guilford.
O’Farrell, T. J., Fals-Stewart, W., Murphy, M., & Murphy, C. M. (2003). Partner violence before
and after individually based alcoholism treatment for male alcoholic patients. Journal of
Consulting and Clinical Psychology, 71(1), 91-102. doi: 10.1037/0022-006X.71.1.92
O’Leary, K.D., & Cohen, S. (2006). Treatment of psychological and physical aggression in a
couple context. In J. Hamel & T. Nicholls (Eds.), Family Interventions in Domestic
Violence: A Handbook of Gender-Inclusive Theory and Treatment (pp. 363-380).
Springer Publishing Company, New York, NY.
O’Leary, K. D. & Schumacher, J. A. (2003). The association between alcohol use and intimate
partner violence: Linear effect, threshold effect, or both? Addictive Behaviours, 28(9),
1575- 1585. doi:10.1016/j.addbeh.2003.08.034
Orford, J., Velleman, R., Copello, A., Templeton, L., & Ibanga, A. (2010). The experiences of
affected family members: A summary of two decades of qualitative research. Drugs:
Education, Prevention & Policy, 17(1), 44-62. doi: 10.3109/09687637.2010.514192
Parent, R. B. & Whitelaw, B. (2008). Community-Based Strategic Policing in Canada (3rd ed.).
Nelson College Indigenous.
Parmar, M. K., & Machin, D. (1995). Survival Analysis: A Practical Approach. New York, NY:
John Wiley & Sons.
Pence, E., & Paymar, M. (1983). Education groups for men who batter: The Duluth Model. New
York: Springer.
Pennington -Zoellner, K. (2009). Expanding "community" in the community response to intimate
partner violence. Journal of Family Violence, 24(8), 539-545. doi: 10.1007/s10896-0099252-5
Prochaska, J. Q., & DiClemente, C. C. (1983). Stages and Processes of Self-Change of
Smoking: Toward An Integrative Model of Change. Journal of Consulting and Clinical
Psychology, 51(3), 390-395.
Quann, N. (2006). Offender profile and recidivism among domestic violence offenders in
Ontario, Canada. Ottawa, Canada: Department of Justice, Research and Statistics
Ratner, P. A. (1993). The incidence of wife abuse and mental health status in abused wives in
Edmonton, Alberta. Canadian Journal of Public Health, 84(4), 246-249.
Reuland, M, Schaefer Morabito, M., Preston, C., & Cheney, J. (2006). Police-Community
Partnerships to Address Domestic Violence. US dept of justice, Office of Community
Oriented Policing Strategies. Retrieved from
Rice, M. E., & Harris, G. T. (2005). Comparing effect sizes in follow-up studies: ROC area,
Cohen’s d, and r. Law and Human Behaviour, 29, 615-620. doi: 10.1007/s10979-0056832-7
Robinson, A. L. (2000). The Effect of a Domestic Violence Policy Change on Police Officers'
Schemata. Criminal Justice and Behaviour, 27(5), 600-624. doi:
Robinson, A. L. (2006). Reducing repeat victimization among high-risk victims of domestic
violence: The benefits of a coordinated community response in Cardiff, Wales. Violence
Against Women, 12(8), 761-788. doi: 10.1177/1077801206291477
Roffman, R. A., Edleson, J. L., Neighbors, C., Mbilinyi, L., & Walker, D. (2008). The men’s
domestic abuse check-up: A protocol for reaching the nonadjudicated and untreated
man who batters and who abuses substances. Violence Against Women, 14(5), 589605. doi:10.1177/1077801208315526
Ruiz-Perez I., Plazaola-Castano, J., & Rio-Lozano, M. (2007). Physical health consequences of
intimate partner violence in Spanish women. European Journal of Public Health, 17(5),
437-443. doi:10.1093/eurpub/ckl280
Schmidt, J. D. & Sherman, L. W. (1996). Does arrest deter domestic violence? In Buzawa, E. S.
& Buzawa, C. G. (Eds.), Do Arrests and Restraining Orders Work? Thousand Oaks, CA:
Sage Publications.
Schneider, E. M. (2000). Battered Women and Feminist Lawmaking. New Haven, CT: Yale
University Press.
Seith, C. (2005). (Un)organized responses to domestic violence: Challenges and changes in
Switzerland. In W. Smeenk, & M. Malsch (Eds.), Family violence and police response:
Learning from research, policy and practice in European countries (pp. 165-189).
Ashgate Publishing, Burlington, VT.
Sev’er, A. (1997). Recent or imminent separation and intimate violence against women: A
conceptual overview and some Canadian examples. Violence Against Women, 3(6),
566-589. doi: 10.1177/1077801297003006002
Sherman, L. W. (1992). Policing Domestic Violence: Experiments and Dilemmas. New York:
Free Press.
Sherman, L. W. (2000). Domestic violence and restorative justice: Answering key questions
Virginia Journal of Social Policy and Law, 8, 263-312.
Sherman, L. W., & Berk, R. A. (1984). Minneapolis domestic violence experiment. National
Institute of Justice US.
Sherman, L. W., & Smith, D. A. (1992). Crime, punishment and stake in conformity: Legal and
informal control of domestic violence. American Sociological Review, 57(5), 680-690.
Sim, J., & Wright, C. C. (2005). The kappa statistic in reliability studies: Use, interpretation, and
sample size requirements. Physical Therapy, 85(3), 257-268.
Smyth, N. J., & Weichelt, S. A. (2005). Drug use, self-efficacy, and coping skills among people
with concurrent substance abuse and personality disorders: Implications for relapse
prevention. Journal of Social Work Practice in the Addictions, 5(4), 63-79. doi:
Sommer, R. (1994). Male- and female-perpetrated partner abuse: Testing a diathesis-stress
model. (Doctoral Dissertation). Retrieved from Dissertations & Theses. (Publication No.
Stanko, E. A. & Hobdell, K. (1993). Assault on men: masculinity and male victimization. British
Journal of Criminology, 33(3), 400-415.
Statistics Canada (2006). Measuring violence Against Women: Statistical Trends 2006.
Steadman, H. J., & Cocozza, J. J. (1974). Careers of the criminally insane: Excessive social
control of deviance. Lexington, MA: Lexington Books.
Stith, S. M., Rosen, K. H., & McCollum, E. E. (2004). Treating intimate partner violence within
intact couple relationships: Outcomes of multi-couple versus individual couple therapy.
Journal of Marital and Family Therapy, 30(3), 305-318. doi: 10.1111/j.17520606.2004.tb01242.x
Stover, C. S., Berkman, M., Desai, R., & Marans, S. (2010). The efficacy of a police-advocacy
intervention for victims of domestic violence: 12 month follow-up data. Violence
Against Women, 16(4), 410-425. doi: 10.1177/1077801210364046
Stuart, R. B. (2005). Treatment for partner abuse: Time for a paradigm shift. Professional
Psychology Research and Practice, 36(3), 254-263. doi: 10.1037/0735-7028.36.3.254
Stuart, G. L., Moore, T. M., Gordon, K. C., Ramsey, S. E., & Kahler, C. W. (2006).
Psychopathology in women arrested for domestic violence. Journal of Interpersonal
Violence, 21(3), 376-389. doi: 10.1177/0886260505282888
Stuart, G. L., Ramsey, S. E., Moore, T. M., Kahler, C. W., Farrell, L. E., Recuperco, P. R., &
Brown, R. A. (2003). Reductions in marital violence following treatment for alcohol
dependence. Journal of Interpersonal Violence, 18(10), 1113-1131. doi:
Stuart, G. L., Temple, J. R., & Moore, T. M. (2007). Improving batterer intervention programs
through theory-based research. Journal of the American Medical Association, 298(5),
560-562. doi: 10.1001/jama.298.5.560
Swets, J. A., Dawes, R. M., & Monahan, J. (2000). Psychological science can improve
diagnostic decisions. Psychological Science in the Public Interest, 1(1), 1-26.
Swogger, M. T., Walsh, Z., & Kosson, D. S. (2007). Domestic violence and psychopathic traits:
Distinguishing the antisocial batterer from other antisocial offenders. Aggressive
Behaviour, 33, 253-260. doi: 10.1002/ab.20185
Taft, C. T., Murphy, C.M., Musser, P. H. & Remington, N. A. (2004). Personality, interpersonal,
and motivational predictors of the working alliance in group cognitive-behavioural
therapy for partner violent men. Journal of Consulting and Clinical Psychology, 72(2),
349-354. doi: 10.1037/0022-006X.72.2.349
Thornberry, T. P., & Jacoby, J. E. (1979). The criminally insane: A community follow-up of
mentally ill offenders. Chicago, IL: University of Chicago Press.
Trujillo, M. P. & Ross, S. (2008). Police response to domestic violence: Making decisions about
risk and risk management. Journal of Interpersonal Violence, 23(4), 454-473.
Viera, A., J., & Garrett, J. M. (2005). Understanding interobserver agreement : The kappa
statistic. Family Medicine, 37(5), 360-363.
Wallach, H. S. & Sella, T. (2008). The importance of male batters’ attributions in understanding
and preventing domestic violence. Journal of Family Violence, 23(7), 655-660. doi:
Walsh Z., Swogger, M. T., O’Connor B. P., Schonbrun, Y.C., Shea, M.T., & Stuart, G. L. (2010).
Subtypes of partner violence perpetrators among male and female psychiatric
patients. Journal of Abnormal Psychology, 119(3), 563-574. doi: 10.1037/a0019858
Ward, T., Melser, J., & Yates, P. M. (2007). Reconstructing the risk-need-responsivity model: A
theoretical elaboration and evaluation. Aggression and Violent Behaviour, 12(2), 208228. doi: 10.1016/j.avb.2006.07.001
Webster, C. D., Douglas, K. S., Eaves, D., & Hart, S. D. (1997) HCR-20: Assessing Risk for
Violence, Version 2. Burnaby, B.C.: Mental Health, Law and Policy Institute, Simon
Fraser University.
Whitaker, D. J., & Lutzker, J. R. (Eds.). (2009). Prevention of partner violence. Washington,
DC: American Psychological Association.
Whittemore, K. E., & Kropp, P. R. (2001, August). The victim’s impact in spousal assault risk
assessment. Paper presented at the annual meeting of the American Psychological
Association, San Francisco, CA.
Wilkinson, D. L. & Hamerschlag, S. J. (2005). Situational determinants in intimate partner
violence. Aggression and Violent Behaviour, 10(3), 333-361.
Wolf, M. E., Ly, U., Hobart, M. A., & Kernic, M. A. (2003). Barriers to seeking police help for
intimate partner violence. Journal of Family Violence, 18(2), 121-129. doi:
Wong, S., Gordon, A., & Gu, D. (2007). Assessment and treatment of violence-prone forensic
clients: an integrated approach. British Journal of Psychiatry, 190(49), 66-74. doi:
Zara, M. J., Ponsoda, V., & Carrillo, R. (2009). Predictors of violence and lethality among Latina
immigrants: Implications for assessment and treatment. Journal of Aggression,
Maltreatment & Trauma, 18(1), 1-16. doi: 10.1080/10926770802616423
Zeeve, W. (2008). Conflict escalation to violence and escalation of violent conflicts. Children
and Youth Services Review 30(3), 297-310. doi: 10.1016/j.childyouth.2007.10.007
Appendix A
Ontario Domestic Assault Risk Assessment (ODARA)
1. Has a prior domestic incident
2. Has a prior nondomestic incident
3. Has a prior sentence of 30 days of more
4. Has failed on prior conditional release (bail,
parole, probation, no-contact order)
5. Made threats to harm or kill during the index
6. Confinement of the partner at the index event
7. Victim fears repetition of violence
8. Victim and/or offender have more than one
9. Victim has biological child from previous
10. Offender is violent outside of this relationship
11. Indicator of substance abuse problem
12. Offender has assaulted victim when she was
13. Victim hast at least one barrier to support
Score (0 = absent; 1 = present)
0 or 1
0 or 1
0 or 1
0 or 1
0 or 1
0 or 1
0 or 1
0 or 1
0 or 1
0 or 1
0 or 1
0 or 1
0 or 1
Raw Score (sum of items)
Note. From Hilton et al. (2004)
Appendix B
Level of Injury Scale (L-Injury)
Level of Injury
No injury or complaints of pain
No visible injury, but complaints of pain
Mark, swelling, scratches
Bruising, black eye, cut (no stitches), bloody nose
Broken bones, unconscious, stitches, broken teeth, internal injuries,
hospitalization, death
Note. As described in Messing (2007)
Linear Violence Scale (L-Violence)
No violence
Entry, held down, vandalism, ripped clothing
Push, slap, throwing objects
Slam, choke, kick, stomp, grab
Shoot, stab, punch, bite, hit with object
Note. As described in Messing (2007)
Table 1
P-Trait Scale for the Measurement of Psychopathic Personality Traits
Mean (SD) Score
(0 = no; 1 = somewhat; 2 = yes)
Callous/lacks empathy
.35 (.57)
Lacks remorse/guilt
.43 (.67)
.43 (.64)
Antiauthority/procriminal attitude
.54 (.76)
Poor anger control
1.2 (.74)
Blames others for problems
.57 (.73)
Narcissism/sees self as superior
.22 (.52)
Commits many types of crimes
.87 (.96)
Thrill seeking behaviour
.38 (.50)
.42 (.51)
Promiscuous sexual behaviour
.15 (.46)
Unstable intimate relationships
.65 (.79)
Total (0 to 24)
6.2 (5.1)
Note. P-Trait = Psychopathic Traits Scale. Chronbach’s α = .86. No gender difference in P-Trait
total score, t(198) = -.19, p = .85.
Table 2
Offender Characteristics
% Males (n = 174)
Committed pre-index IPV offense(s)
60 (104)
Committed post-index IPV offense(s)
45 (78)
History of drug/alcohol misuse
63 (110)
Unwillingness to accept responsibility
54 (94)
Has a criminal record
62 (108)
Prior jail sentence of at least 3 months
30 (52)
IPV Recidivism Subtype:
29 (50)
Moderately Persistent
37 (65)
Stable Persistent
34 (59)
Note. IPV = Intimate partner violence. *p < .05. **p < .01. ***p < .001.
% Females (n = 26)
65 (17)
36 (9)
59 (15)
47 (12)
31** (8)
4** (1)
31 (8)
35 (9)
35 (9)
Table 3
Percentage of Pre- and Post-Index IPV Offending Events Committed by ModeratelyPersistent and Stable-Persistent Offender Subtypes
Offender Type
% committing given number of IPV Events (n)
1-2 Events
3-6 Events
7-14 Events
15+ Events
Pre-Index Events
Moderately Persistent
45.9 (34)
20.4 (15)
6.8 (5)
Stable Persistent
45.6 (31)
5.9 (4)
6.0 (4)
Post-Index Events
Moderately Persistent
20.3 (15)
6.9 (5)
Stable Persistent
47.1 (32)
42.6 (29)
7.4 (5)
3.0 (2)
Note. IPV = Intimate partner violence. Moderately persistent: n = 74. Stable persistent: n = 68.
Table 4
Victim Characteristics
Unwilling to press charges against suspect
History of violent relationships
Pattern of inconsistent behavior towards
Coerced/threatened by partner
Demeaned/belittled by partner
Fears suspect
Partner is jealous/possessive
Note. *p < .05. **p < .01. ***p < .001.
% Males (n = 21)
76 (16)
57 (12)
76 (16)
4.8 (1)
4.8 (1)
14.3 (3)
% Females (n = 179)
79 (141)
69.5 (124)
76 (136)
30.2 (54)**
27.4 (49)*
40 (72) ***
24 (43)
Table 5
Index Offense Characteristics for Male and Female Perpetrators
% Overall (N = 200) % Male (n = 174)
% Female (n = 26)
Physical violence
76 (152)
76 (132)
73 (19)
17 (34)
18 (31)
12 (3)
Non-threatening harassment
7 (14)
6 (10)
15 (4)
Victim suffered injury
46 (92)
45 (78)
50 (13)
Weapon present
8.5 (17)
8 (14)
12 (3)
Weapon used
5.0 (10)
4.0 (7)
12 (3)
Offender was impaired
54.1 (108)
54.7 (95)
50 (13)
Victim was impaired
29.2 (58)
29.8 (52)
24 (6)
Note: There were no statistically significant gender differences on any variables. p > .05
Table 6
Correlations between contextual variables of the index event and IPV recidivism
Correlation coefficient
Presence of a weapon
Victim is pregnant
Victim displays inconsistent behaviour towards suspect
Victim has been in abusive relationships in the past
Victim wants suspect arrested
Victim fears suspect
Victim is impaired
Suspect is impaired
Suspect has a criminal record
Suspect is on probation
Suspect displays jealous/possessive behaviour
Suspect is demeaning/belittling towards victim
Suspect uses coercion/threats
Suspect displays poor anger control†
Suspect is unwilling to take responsibility for behaviour†
Note. *p < .05. **p < .01. ***p < .001. Phi coefficients were used because variables are
dichotomous. †Point biserial correlations were calculated for scale variables.
Table 7
L-Injury and L-Violence Scores for Male and Female Offenders
Measurement Instrument
L-Injury - all index cases
L-Injury – cases with L-Injury > 1†
L-Violence - all index cases
L-Violence – cases with L-Injury > 1†
Full Sample
1.11 (1.35)
2.40 (.91)
M (SD)
1.11 (1.37)
2.44 (.92)
1.08 (1.26)
2.15 (.90)
2.41 (1.22)
3.27 (.77)
2.40 (1.17)
3.20 (.77)
2.50 (1.53)
3.69 (.63)*
Note: L-Injury = Level of Injury Scale; L-Violence = Level of Violence Scale. † Only includes
cases with L-Injury score > 1 (n = 92; n = 79 males, n = 13 females). There were no significant
gender differences in L-Injury scores. *p < .05.
Table 8
Predictive Validity of ODARA Total Score for Predicting IPV Recidivism as Measured by ROC
Curve Analyses
AUC (95% CI)
Measurement Instrument
Full Sample (N = 200) Males (n = 174)
Females (n = 26)
ODARA total score
.70 (.63, .77)
.70 (.63, .78)
.67 (.46, .88)
P-Trait total score
.80 (.74, .86)
.80 (.73, .86)
.85 (.70, 1.0)
R-ODARA total
.79 (.72, .85)
.78 (.72, .85)
.82 (.66, .98)
Note. ODARA = Ontario Domestic Assault Risk Assessment (Hilton et al., 2004); IPV = intimate
partner violence; ROC = Receiver Operator Characteristic; AUC = Area Under the Curve; PTrait = Psychopathic Trait Scale; R-ODARA = Revised Ontario Domestic Assault Risk
Table 9
Pearson Correlations Between ODARA Total Score and IPV Characteristics
IPV Characteristic
Number of pre-index IPV offenses
Recidivism (yes = 1, no = 0)
Number of post-index IPV offenses
Days to first post-index IPV offense
Victim unwillingness to press charges at index (yes = 1, no = 0)
Presence of psychological abuse in relationship at index (yes = 1, no = 0)
Index L-Injury total score
Index L-Violence total score
Index L-Injury total score†
Index L-Violence total score†
ODARA total (r)
Note. *p < .05. **p < .01. ***p < .001. † Only includes cases with L-Injury score > 1 (n = 92).
ODARA = Ontario Domestic Assault Risk Assessment (Hilton et al., 2004); IPV = intimate
partner violence; L-Injury = Level of Injury Scale; L-Violence = Level of Violence Scale.
Table 10
Total ODARA Scores by ODARA Risk Category and Perpetrator Gender
M (SD)
ODARA Ratings
ODARA Total score
5.17 (2.49)
5.28 (2.52)
Low Risk
1.41 (.62)a
1.38 (.64)a
Moderate Risk
4.38 (.99)
4.40 (1.02)b
High Risk
7.94 (1.01)
8.00 (1.04)c
4.46 (2.25)
1.50 (.55)a
4.23 (.83)b
7.43 (.54)c
Note. ODARA = Ontario Domestic Assault Risk Assessment. Low risk: n = 32, moderate risk: n
= 96, high risk: n = 72. Within columns, significant differences at the p < .001 level are denoted
by the use of different letter superscripts. There was also no significant difference between
male and females on ODARA total score, F (1, 198) = 2.4, p = .12.
Table 11
ODARA Risk Level Comparisons for Pre- and Post-Index IPV Offending
Mean (SD) # of pre% committing postRisk category
index IPV offenses
index IPV (n)
Low (n = 32)
.19 (.74)a
15.6a (5)
Moderate (n = 96)
1.7 (3.6)
40.2b (39)
High (n = 72)
3.9 (4.1)
62.5c (45)
Mean (SD) # of postindex IPV offenses
Note. Within each column, significant differences between risk categories at the p < .001 level
denoted by use of "a", "b", "c". ODARA = Ontario Domestic Assault Risk Assessment (Hilton et
al., 2004); IPV = intimate partner violence.
Table 12
Mean Survival Time by ODARA Risk Category
Mean (SD) time (years)
4.9 (1.8)a
ODARA Risk Category
3.6 (2.4)b
2.4 (2.3)c
Note. Significant differences between risk categories at the p < .001 level denoted by use of
"a", "b", "c". ODARA = Ontario Domestic Assault Risk Assessment (Hilton et al., 2004).
Table 13
Summary of Multiple Regression Analysis for Suspect, Victim, and IPV Context Variables in
Predicting Arrest
Block 1 - Suspect characteristics
Impairment by alcohol/drugs
Willingness to accept responsibility
P-Trait total score
Block 2 - Victim characteristics
Impairment by alcohol/drugs
Willingness to press charges
Block 3 - IPV context characteristics
Weapon involved
L-Injury score
L-Violence score
Block 4 - ODARA total score
Note. *p < .05. ***p < .001. R = .03 for Block 1. ∆R = .29*** for Block 2. ∆R = .14*** for Block
3. ∆R2 = .002 for Block 4. IPV = intimate partner violence; P-Trait = Psychopathic Traits Scale;
L-Injury = Level of Injury Scale; L-Violence = Level of Violence scale.
Cumulative proportion not committing post-index
Time since index IPV offense (months)
Figure 1.
Kaplan-Meier survival curves for ODARA risk categories. Low, moderate and high risk
offenders significantly differed in the time it took them to commit post-index IPV, LR(2) = 22.2, p
< .001. IPV = intimate partner violence.
Was this manual useful for you? yes no
Thank you for your participation!

* Your assessment is very important for improving the work of artificial intelligence, which forms the content of this project

Download PDF