Evaluating the Ontario Domestic Assault Risk Assessment in an Australian Frontline Police Setting

Abstract

This article describes a prospective validation of the Ontario Domestic Assault Risk Assessment (ODARA) in an Australian sample of 854 family violence cases presenting to police over a 5-month period in 2015. Two hundred cases met inclusion criteria for administration of the ODARA (male-to-female intimate partner violence with a history of assault and cohabitation). The ODARA performed well in predicting further intimate partner physical assault (area under the curve [AUC] = .68), and in predicting the outcome of any further police contact for nonphysical intimate partner abuse (AUC = .72). Despite these positive results, the instrument’s restrictive inclusion criteria meant that it could be appropriately applied in only 23% of family violence cases reported to police during the data collection period, limiting its practicality in this setting.

Keywords

violence risk assessment intimate partner violence family violence actuarial risk assessment ODARA

Over the last two decades, there has been increasing acknowledgment of intimate partner violence (IPV) as a legal and public health issue. The World Health Organization (WHO) defines IPV as “a behavior by an intimate partner or ex-partner that causes physical, sexual or psychological harm, including physical aggression, sexual coercion, psychological abuse and controlling behaviors” (WHO, 2016). Estimates suggest that one in four women and one in five men experience IPV victimization throughout their lifetime (Desmarais, Reeves, Nicholls, Telford, & Fiebert, 2012; Nicholls, Pritchard, Reeves, & Hilterman, 2013). A commonly employed help-seeking behavior for many victims of IPV is to engage police, both to prosecute crimes that have occurred and to try to prevent further abuse (Brookoff, O’Brien, Cook, Thompson, & Williams, 1997; Hutchison & Hirschel, 1998). Given the frequency with which IPV occurs, policing organizations face a significant challenge in identifying and prioritizing cases in which IPV may escalate or persist over time, causing further harm to victims (Nicholls et al., 2013; Rettenberger & Eher, 2013; Williams, 2012).

A number of violence risk assessment instruments have been developed to assist police in this task—each differing in their structure, predictive validity, theoretical assumptions, professional requirements, and the context in which they can be applied (Messing & Thaller, 2013; Nicholls et al., 2013). A systematic review by Nicholls and colleagues (2013) identified 19 violence risk assessment tools that have been either developed for, or applied to, IPV risk assessment. Some of these tools require clinical training or extensive knowledge of the perpetrator’s history and psychological characteristics, making them unsuitable for use by frontline police officers (e.g., Spousal Assault Risk Assessment [SARA]; Kropp, Hart, Webster, & Eaves, 1995; Violence Risk Appraisal Guide [VRAG]; Harris, Rice, & Quinsey, 1993; Domestic Violence Risk Appraisal Guide [DVRAG]; Hilton, Harris, Rice, Houghton, & Eke, 2008). Other instruments are designed to assess the risk of severe injury or homicide in a range of frontline settings (e.g., victim shelters, emergency department, police responders), and are intended to be administered with information from victims rather than perpetrators of IPV. These instruments, such as the Danger Assessment (Campbell, 1986; Campbell, Webster, & Glass, 2009) and Lethality Screen (Messing et al., 2015) are used by responding police in some jurisdictions in the United States (Messing & Campbell, 2016).

Other tools, designed specifically for police officers, assess broader risks associated with IPV, and are based on a combination of current and historical information about the victim, the perpetrator, and their relationship. The Brief Spousal Assault Form for the Evaluation of Risk (B-SAFER; Kropp, Hart, & Belfrage, 2005, 2010) is a structured professional judgment risk assessment designed to help law enforcement personnel assess and manage the risk of IPV among adult couples with a history of IPV (irrespective of sexual orientation or gender). There are several studies that provide evidence for the construct validity, concurrent validity, interrater reliability, and predictive validity of the B-SAFER (e.g., Au et al., 2008; Belfrage & Strand, 2008; Serie, van Tilburg, van Dam, & de Ruiter, 2015; Storey, Kropp, Hart, Belfrage, & Strand, 2014; Thijssen & de Ruiter, 2011). However, the B-SAFER requires in-depth assessment of each case and is, therefore, somewhat time consuming to complete, meaning it is not appropriate for first responders attending a report of IPV.

In contrast, the Ontario Domestic Assault Risk Assessment (ODARA; Hilton et al., 2004) is designed specifically for a frontline policing setting, being relatively simple and quick to use. The ODARA is an actuarial instrument designed to assist frontline police officers in assessing and managing the risk of male-to-female–perpetrated IPV, based on information that is usually readily available to them from the victim, perpetrator, and their databases. The ODARA has demonstrated good interrater reliability (Hilton, Harris, Popham, & Lang, 2010; Hilton et al., 2008). Perhaps, more important, the instrument has demonstrated accuracy in predicting further intimate partner physical assault (Hilton et al., 2010; Hilton et al., 2008; Hilton et al., 2004) and is able to appropriately classify perpetrators into categories differentially associated with both risk of reoffending and severity of injury to victims (Hilton, Harris, & Holder, 2008; Nicholls et al., 2013). There is preliminary evidence suggesting that the ODARA is effective in categorizing the majority of eventual intimate partner homicide offenders into its highest risk category (Eke, Hilton, Harris, Rice, & Houghton, 2011). Hilton and Harris (2009) have also shown that the ODARA can distinguish between IPV recidivists and nonrecidivists within a sample of offenders with less extensive criminal records. A meta-analysis by Messing and Thaller (2013) identified that the ODARA had the best predictive accuracy of all the IPV risk instruments examined, with an average area under the curve (AUC) of .67 and a range of .64 to .72 (moderate to large effect sizes).

The majority of research evaluating the ODARA has been undertaken in Canada by the instrument’s authors, although it has also been evaluated in some other jurisdictions. Rettenberger and Eher (2013) investigated the use of the ODARA in a sample of high-risk sexually motivated offenders released from the Austrian prison system. The results were consistent with Canadian studies, showing that the ODARA risk categories were a good predictor of sexual and physical intimate partner assault recidivism (AUC = .71). Similarly, in a New Zealand study examining ODARAs completed by trained frontline police officers, increases in risk category were positively associated with the likelihood of an intimate partner assault reoffense (Nimmo, 2012). However, a recent retrospective study in a Swiss sample of 185 domestic violence offenders registered by police obtained weaker results (Gerth, Rossegger, Bauch, & Endrass, 2017). Although the ODARA still significantly differentiated between IPV recidivists and nonrecidivists in this sample, the effect size was small (AUC = .61). Given the markedly different findings in the Swiss study, it is possible that there were sample characteristics that affected the performance of the ODARA. The authors suggested that the effects of the interventions implemented with the offenders, in addition to societal and jurisdictional differences between Canada and Switzerland, may have led to the different results. Further research is needed in a wider variety of jurisdictions to determine whether the ODARA can reliably be used outside of Canada.

Additional research is also needed in which the ODARA is administered as intended in real-life settings. With the exception of the New Zealand study, which was not subject to peer review (Nimmo, 2012), all ODARA evaluation research to date has used retrospective designs, in which ODARAs are scored from file by experienced researchers and trained research assistants (e.g., Eke et al., 2011; Gerth et al., 2017; Hilton et al., 2010; Hilton et al., 2008; Rettenberger & Eher, 2013). Given that the instrument is intended for use by responding police officers, it is essential that the ODARA be shown to be equally effective when used in a dynamic field environment by less specialized raters.

A further shortcoming of existing ODARA evaluation research is the relatively restrictive outcome variable used in most studies: repeat physical assault of the partner. Given that IPV can involve a range of behaviors that are not necessarily physically aggressive (WHO, 2016), it would be worthwhile to know whether the ODARA performs as well in predicting the broader outcome of repeat police-reported intimate partner abuse, rather than only physical violence. In a recent study, Hilton and Eke (2016) provide preliminary support for the use of the ODARA in predicting any general offense (AUC = .75), any violent offense (AUC = .65), any nonviolent offense (AUC = .74), assault on any intimate partner (AUC = .67), and recidivism relating to any intimate partner offense including stalking, harassment, violence, and release failures (AUC = .71). These results suggest that a specialized IPV risk assessment such as the ODARA predicts a range of criminal behaviors, in addition to a wider range of intimate partner abuse outcomes, at least in samples of offenders in contact with the police. This likely reflects the fact that there are a core group of risk factors common to violent and general reoffending that are somewhat similar and are associated with general antisociality (Hanson, Bourgon, & Helmus, 2007; Hilton & Eke, 2016; Nicholls et al., 2013). Similar findings have been shown with tools developed to specifically predict violent recidivism, such as the Historical-Clinical-Risk Management-20 (HCR-20) also predicting general recidivism (de Vogel, Ruiter, Hildebrand, Bos, & van de Ven, 2004; Gray, Fitzgerald, Taylor, MacCulloch, & Snowden, 2007), and tools developed to specifically predict general recidivism also predicting violent recidivism (Hanson et al., 2007).

This study aimed to overcome some of the shortcomings of existing ODARA research literature by prospectively examining the ODARA’s predictive validity when coded by Australian police officers at the time of report. Expanding on previous literature, the predictive validity of the ODARA was assessed against two different IPV outcomes: (a) further police-reported physical violence toward the same significant intimate partner by the same offender as at index report and (b) further police-reported abuse of the same significant intimate partner by the same offender as in the index report, which did not involve physical violence.

Method

The study was conducted in a police division in metropolitan Melbourne, Australia. Melbourne is a large, multicultural city on the southeast coast of Australia, with a population of 4.6 million people, 25% of whom were born overseas (Australian Bureau of Statistics [ABS], 2016). ODARA items were scored by police officers in a nonconsecutive sample of 854 family violence incidents (FVIs) involving unique dyads that occurred between January 1, 2015, and June 13, 2015. This represented 43.86% of all FVIs that occurred in this division during the data collection period. In Victoria, police record an FVI whenever they attend an incident between two family members (including intimate partners) that involves physical, sexual, emotional, or psychological abuse, regardless of whether the incident results in charges (Victoria Police, 2014).

Sample

Eligibility for inclusion was based on the inclusion criteria for the ODARA as described by Hilton and colleagues (Hilton, Harris, & Rice, 2010; Hilton et al., 2004). Therefore, FVIs were excluded from analyses if they involved family members who were not current or former intimate partners (n = 273), if they involved a female perpetrator and male victim (n = 86) or same sex intimate partners (n = 16), if the perpetrator was below the age of 18 years (n = 4), if the victim and perpetrator had not currently or previously cohabited (n = 64), or if there had been no act of physical assault or threat with weapon in hand during the index incident or a prior incident between the same two people (N = 211). This resulted in a final sample of 200 cases that met inclusion criteria (23.42% of all FVIs).

The mean age of the 200 perpetrators at the time of the initial police response was 37.73 years (SD = 10.68 years, range = 18-85 years) and the mean age of victims was 34.74 years (SD = 9.85 years, range = 18-75 years). According to police records, 71 cases (35.50%) involved a former intimate relationship, 63 (31.50%) were in current common law relationships,¹ 57 (28.50%) were currently married, and in the remaining nine (4.50%) cases, the parties reported to police that they were dating, but had cohabited at some point in their relationship.

Measures

The ODARA predicts repeat physical assault (including threats or sexual violence) of female intimate partners by male perpetrators. It was developed actuarially from retrospective coding of 689 IPV cases recorded in the Ontario police database (100 used for cross validation) with a mean follow-up period after the index offense of almost 5 years (Hilton et al., 2004). It comprises 13 statistically derived items, including information about the history of IPV in the relationship, the victim’s view on the potential for future violence, and the presence of other antisocial behaviors of the perpetrator. Each ODARA item is scored 0 or 1, with scores added together for a total score that reflects the individual’s overall risk category relative to other male IPV perpetrators. The risk categories, or “bins,” range from 0 (lowest risk) to 7+ (highest risk), with research showing incremental increases in the probability of IPV recidivism with each increase in category (Hilton et al., 2010; Hilton et al., 2004).

Procedure and Variables

FVIs dealt with by first responders are referred for further assessment to a specialist family violence team (FVT) within each Victoria Police division. The FVT triages the case within 5 days and identifies the need for any follow-up and ongoing risk management. Triage involves reviewing collateral information from the Law Enforcement Assistance Program (LEAP) database for past and current incident reports, in addition to speaking to the victim and perpetrator in some cases. During this project, FVT members used this information to rate the presence or absence of the 13 ODARA items. The items were presented as part of a list including demographic information and information about the perpetrator’s history of physical violence and family violence. The officers did not know they were scoring the ODARA, nor did they calculate a total score.² Due to the demands on police time and lack of researcher access to police information, interrater reliability could not be investigated.

Follow-up data were collected on June 30, 2016, with follow-up time ranging from 16 to 216 days, depending on when the original FVI occurred (M = 111.62 days, SD = 47.10 days). Two mutually exclusive types of recidivism were identified:

Physical assault recidivism was defined as the presence of another FVI between the same dyad, which led to at least one charge for physical violence or involved police-reported physical violence. The following charges were defined as physical violence: assault (actual), assault (threat), assault (sexual), attempted murder, homicide, and abduction.

Nonphysical abuse recidivism was defined as any further FVI involving the same dyad that was reported to police during the follow-up period, excluding subsequent incidents where physical violence was present (using the above definition of physical violence).

Data Analyses

Data were analyzed using IBM SPSS Statistics Version 23. Chi-square analyses were used to ascertain the unique contribution of each of the ODARA items as a predictor of both kinds of recidivism, with receiver operator curve (ROC) analyses as an effect size measure. The ability of the ODARA total score to discriminate between recidivists and nonrecidivists was assessed using ROC analyses. The ROC analysis provides an AUC value by charting the proportion of true positives against false positives at each possible score. AUC values range from 0 to 1.0, with a value of .50 signifying that the score differentiates between recidivists and nonrecidivists no better than chance, and a value of 1.0 signifying perfect differentiation, with all recidivists positively identified and all nonrecidivists negatively identified (Messing & Thaller, 2013; Rice & Harris, 2005). An AUC value of .71 and above is considered a large effect size (Rice & Harris, 2005). To determine the predictive validity of the ODARA at all possible cut points, sensitivity, specificity, positive predictive values (PPVs), and negative predictive values (NPVs) were calculated for each risk category.

Results

In total, 37 (18.50%) unique dyads had at least one further FVI reported during the follow-up period. Nineteen (9.50%) police reports involved physical assault recidivism and 23 (11.50%) involved nonphysical abuse recidivism. Some cases had multiple repeat FVIs. In these cases, any reoffense that involved physical violence was coded as physical assault recidivism and any reoffense that did not involve any physical violence was coded as nonphysical abuse recidivism. The average time to recidivism was 111.91 days (SD = 73.27 days) for physical assault and 97.06 days (SD = 53.20 days) for nonphysical abuse recidivism. ODARA scores were approximately normally distributed with a mean of 3.22 (SD = 2.09). Preliminary analyses revealed that ODARA Item 10 (any prior violence against other victims) had been coded identically to Item 2 (police record of nondomestic violence), based only on information about previous violence that was available in police databases, rather than also from information gathered from the victim.

As shown in Table 1, ODARA total scores were effective in discriminating between perpetrators who reoffended with a further physical assault and those who did not (AUC = .68, 95% CI = [0.55, 0.81]). The total score had a similar ability to discriminate between those with and without any further nonphysical abuse (AUC = .72, 95% CI = [0.61, 0.82]). The difference between the AUC for physical assault and nonphysical abuse recidivism was not significantly different as the confidence intervals overlap substantially. The relationship between each ODARA item and each recidivism outcome are also shown in Table 1. Only three of the 13 ODARA items were significantly related to physical assault recidivism at an individual level, although all with small to moderate effect sizes. Items 2 (police record of nondomestic violence) and 10 (any prior violence toward another victim) were the strongest predictors of physical assault recidivism (noting that they were identically scored in this study and so have an identical relationship to the outcome), along with prior custodial sentence of 30 days, which had a smaller effect size. In contrast, nonphysical abuse recidivism was predicted by level of victim concern, threat to harm or kill at the index incident, and substance abuse, all with moderate effect sizes.

Table 1:

Chi-Square Coefficients and Predictive Accuracy (AUC) for Each ODARA Item Across Physical Assault and Nonphysical Abuse Recidivism

ODARA items	N with item present (%)	Physical assault		Nonphysical abuse
ODARA items	N with item present (%)	χ²(1)	AUC	χ²(1)	AUC
1. Prior domestic incident	29 (14.50)	2.33	.57	1.07	.54
2. Prior nondomestic incident	36 (18.00)	8.26**	.63	1.15	.55
3. Prior custodial sentence of 30 days or more	27 (13.50)	5.88*	.60	1.51	.55
4. Failure on prior conditional release	46 (23.00)	0.87	.55	3.81	.59
5. Threat to harm or kill at the index assault	50 (25.00)	0.17	.48	4.73*	.61
6. Confinement of the partner at the index assault	27 (13.50)	1.00	.54	3.24	.52
7. Victim concern	73 (36.50)	1.07	.56	9.25**	.66*
8. More than one child	98 (49.00)	0.66	.55	0.59	.55
9. Victim’s biological child from a previous partner	13 (6.50)	1.46	.46	1.83	.54
10. Violence against others	36 (18.00)	8.26**	.63	1.15	.55
11. Substance abuse	56 (28.00)	2.07	.58	5.06*	.61
12. Assault on victim when pregnant	9 (4.50)	0.99	.48	1.06	.52
13. Barriers to victim support	145 (72.50)	0.44	.54	0.43	.53
Total ODARA score			.68*		.72***

Note. AUC = area under receiver operating characteristic curve; ODARA = Ontario Domestic Assault Risk Assessment.

p < .05. **p < .01. ***p < .001.

Sensitivity, specificity, PPV, and NPV were calculated for each ODARA bin. Figures 1 and 2 display the trade-off between sensitivity, specificity, positive predictive power, and negative predictive power at all possible cut points on the ODARA for physical assault and nonphysical abuse recidivism. Positive predictive power ranged from .10 to .24, and negative predictive power ranged from .92 to 1.0 for physical assault recidivism, depending on the cut point chosen. For nonphysical abuse recidivism, positive predictive power ranged from .12 to .29, and negative predictive power ranged from .90 to 1.0, depending on the cut point chosen. These figures show that an ODARA total score of 4 or more maximized the predictive accuracy of the ODARA for both the physical assault and nonphysical abuse outcome, correctly identifying 68% of physical assault and 70% of nonphysical abuse recidivists, while correctly excluding 64% of both physical assault and nonphysical abuse recidivists.

Figure 1:

Area Under Curve and Other Accuracy Statistics for the ODARA Predictions of Physical Assault Recidivism

Figure 2:

Area Under Curve and Other Accuracy Statistics for the ODARA Predictions of Nonphysical Abuse Recidivism

Discussion

This study found evidence for the validity of the ODARA in an Australian police setting. Moreover, the ODARA performed equally well in this real-life prospective study as in previous research, in which the ODARA was scored retrospectively from files by researchers, notwithstanding the fact that police officers did not receive any specific training in completing the ODARA. The ODARA performed well not only in predicting its stated outcome, further intimate partner physical assault, but also in predicting the outcome of further police contact for nonphysical intimate partner abuse.

Even with a relatively low base rate for physical assault recidivism, the ODARA total score was still able to accurately discriminate between recidivists and nonrecidivists 68% of the time. This finding is consistent with studies conducted by the research team that developed the ODARA (Hilton & Harris, 2009; Hilton et al., 2004). However, a closer look at the univariate performance of each ODARA item revealed that items pertaining to the perpetrators’ history of acting violently toward others (i.e., Items 2 and 10) accounted for the majority of the ODARA’s prediction of further assault, whereas most other predictors added relatively little to the performance of the instrument. This finding is consistent with Klein and Tobin’s (2008) longitudinal study of IPV perpetrators, which indicated that perpetrators with a history of general violent behavior were more likely to reoffend toward their intimate partner. These results are also consistent with the broader literature showing that a history of violent offending and of criminal offending generally (as indicated by having a custodial sentence of more than a month) are among the strongest predictors of violent recidivism (see Bonta, Law, & Hanson, 1998). The findings of this study suggest that these general risk factors for violence are equally important to the prediction of physical violence in an intimate partner context.

In addition to demonstrating good discriminant ability for its intended purpose of recidivistic physical assault, the ODARA showed similar ability to discriminate between those with and without any further police contact for nonphysical abuse. With a base rate for nonphysical abuse recidivism of 11.5%, there was a 72% probability that a recidivist would have a higher ODARA score than a nonrecidivist. These results indicate that the ODARA may have utility in assessing the risk of any further police contact for intimate partner abuse within the same relationship, and add to small literature suggesting the ODARA has use in assessing a range of intimate partner abuse outcomes (Hilton & Eke, 2016). For this outcome variable, victim concern contributed most to the predictive power of the instrument, with small, but significant contributions from other items, though none specifically relating to previous violence. Notably, in this study, victims were asked about their fear of further IPV, not specific fear of severe or fatal violence by their partner, which has been used in other studies of victim fear (Nicholls et al., 2013). This finding might suggest that it is not fear of one’s partner engaging in a specific violent outcome, per se, but rather a general fear of one’s partner that is a risk factor for subsequent intimate partner abuse.

The present study provides preliminary support for the notion that an actuarial instrument developed to predict a specific outcome of repeat physical assault can predict a more general IPV recidivism outcome. However, it is noteworthy that different ODARA items were related to physical assault and nonphysical abuse recidivism. The physically violent outcome was predicted by risk factors associated with physical violence and significant antisociality (incarceration for more than 30 days), whereas risk factors for nonphysical abuse recidivism appeared to be somewhat more victim specific (victim concern/fear and threats to harm or kill), and relate to other criminogenic needs (substance use). This may reflect different “types” or patterns of IPV, such as “antisocial” or “family-only” IPV offenders, with different relevant risk factors and different risk profiles (Petersson, Strand, & Selenius, 2016). Future research including a wider range of information about the IPV perpetrators’ other offending could investigate this finding further. It should be noted that comparison of this sample with the ODARA development sample showed higher prevalence of all risk factors that could be directly compared, possibly explaining some of the nonsignificant univariate relationships with subsequent assault.

Analysis of test utility statistics (i.e., sensitivity, specificity, PPV, NPV) shows the expected differences in predictive performance for each ODARA bin. An ODARA total score of 4 or above appeared to maximize the predictive accuracy of the ODARA for the physical assault outcome. Of those who scored in this bin or above, 16.45% went on to violently reoffend against their partner (almost twice the base rate of 9.50%), whereas 95% of those scoring below did not. Moreover, scores in this bin or above maximized the sensitivity and specificity of the instrument, with 68% of physical assault recidivists being correctly identified and 64% of nonrecidivists being correctly identified.

When considering the predictive accuracy of various bins for the nonphysical abuse outcome, a total score of 4 or above also appeared to produce the best outcome, maximizing sensitivity and specificity (.70 and .64, respectively). Positive predictive power for a score in this bin or above was 24%, approximately twice the base rate of recidivism of 11.5%. Conversely, 94% of people with a score below four did not have further nonphysical abuse during the follow-up period, an incremental improvement on the base rate of 88.6% who did not reoffend. These findings are comparable with those in the ODARA development sample, which also concluded that a total score of 4 or above maximizes the predictive accuracy of the ODARA for physical assault recidivism (Hilton et al., 2004). The findings of this study show an improvement in sensitivity (.59 in the development sample) and a decline in specificity (.79 in the development sample) at the selected threshold in comparison with the original Canadian study.

Although the ODARA performed well on all validity measurements, its restrictive inclusion criteria have significant implications for its implementation as a screening measure in an Australian policing setting. The ODARA could be legitimately scored and used in an evidence-based way in only 23% of all FVIs recorded during the data collection period (although we acknowledge that there may have been additional cases in which previous physical assault was present but not known to police). In Victoria, as in many Australian jurisdictions, family violence is defined broadly in relevant legislation as involving abuse between people in any sort of family-like relationship, regardless of gender, sexuality, or nature of the relationship (Family Violence Protection Act 2008, 2017). This reflects literature showing that family violence can involve abuse not only by male perpetrators of female victims but also by female perpetrators toward male victims, same sex partners, and between family members in other kinds of relationships (Carney & Barner, 2012; Langhinrichsen-Rohling, Misra, Selwyn, & Rohling, 2012). This means that intimate partner abuse accounts for only 56% of police-recorded FVIs in Victoria, and male-to-female intimate partner abuse for approximately 47% of all incidents (State of Victoria, 2016). Therefore, even without applying the cohabitation and previous physical assault inclusion criteria, the ODARA would be relevant in fewer than half of all FVIs requiring a police response in Victoria.

This study’s results must be interpreted in light of its limitations. Police officers were not trained in how to administer the ODARA because they were unaware that they were administering it in addition to their local standard risk assessment protocol (that protocol and its lack of evidence base and predictive validity is discussed in McEwan, Bateson, & Strand, 2017). This may be why Items 2 and 10 were scored identically, based only on police records rather than also from information gathered from the victim. This finding potentially has limited impact given the overlapping nature of the items, whereby Item 10 would automatically be scored present if Item 2 were present (Hilton et al., 2010). The other significant limitation was marked variability in individual follow-up periods, limiting the comparability of the present results with the original ODARA validation, where the follow-up period was a standard 5 years. However, this likely had relatively limited impact given research showing that most IPV reoffending occurs within 100 days of the index offense (Stansfield & Williams, 2014), and, if anything, would have underestimated recidivism. Finally, given the relatively small sample size and low base rate of physical assault recidivism, it was not possible to undertake multivariate analyses to further explore the interrelationships among ODARA items.

Conclusion

This study provides strong support for the validity of the ODARA when administered by police in an Australian frontline setting. The findings provide support for the use of the ODARA not only in predicting its stated outcome, further intimate partner physical assault by the same perpetrator, but also in predicting the outcome of further police contact for nonphysical intimate abuse within the same relationship. The findings suggest that some specific items have a much stronger predictive relationship than others, differing depending on the risk outcome evaluated. The role of particular risk factors and their relationship to different kinds of risk should continue to be explored in further studies with larger samples and a wider range of variables. The only caveat on the use of the ODARA in an Australian police setting was that it could not be applied in accordance with the manual in the majority of FVIs requiring a police response. This is a significant problem for implementation, as it would necessitate the use of another screening instrument in addition to the ODARA. For many police organizations, having multiple screening instruments for different types of family violence would introduce a significant training burden and administration challenge, meaning that a single instrument that can effectively triage all family violence reports for further, detailed assessment is more desirable. It may be that the ODARA can be shown to be valid in family violence cases outside of its restrictive inclusion criteria in future research, or it may be possible to develop an equally effective, but more general, instrument for police settings, where family violence definitions are broader than only IPV (e.g., Williams, 2012).

Footnotes

Acknowledgements

The authors would like to acknowledge Superintendent Stuart Bateson and his staff, and Dr. David Ballek and the staff at Victoria Police Corporate Statistics for their assistance with this research. The authors would also like to acknowledge the Australian Government’s support through the Research Training Program for its assistance with this research.

This research was supported by funding from the Macedon Ranges and North-West Melbourne Medicare Local.

The authors have no conflicts of interest to declare.

Notes

Ilana Lauria is a doctoral candidate in clinical and forensic psychology at Swinburne University in Melbourne, Australia. Her research interests focus on the risk assessment and management of intimate partner violence.

Troy E. McEwan is a senior lecturer and a registered clinical and forensic psychologist. Her research focuses on improving assessment and understanding of complex criminal behaviors such as stalking, family violence, and deliberate firesetting. Her clinical work is with adult offenders who engage in these and other forms of interpersonal violence.

Stefan Luebbers is a lecturer in forensic behavioral science and a clinical forensic psychologist. His clinical background includes experience in general and forensic mental health with children, adolescents, and adults, and military psychology. His research focuses on youth mental health and violence, victimization and abuse, and adolescent service delivery.

Melanie Simmons is a doctoral candidate in clinical and forensic psychology at Swinburne University in Melbourne, Australia. Her research interests focus on understanding and assessing violence within the family and in the community.

James R. P. Ogloff is professor and director of the Centre for Forensic Behavioural Science, and executive director of psychological services and research for the Victorian Institute of Forensic Mental Health, Victoria’s statewide mental health service. His research interests are in the area of psychology and law, mental illness and offending, complex criminal behavior, and service development and delivery.

References

Cheung

Kropp

Yuk-chung

Lam

G. L. T.

Sung

(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, 727-735.

Australian Bureau of Statistics. (2016). Regional population growth Victoria, Australia 2016 (Catalogue No. 3218.0). Canberra, Australia: Author. Retrieved from http://www.abs.gov.au/AUSSTATS/abs@.nsf/mf/3218.0

Belfrage

Strand

(2008). Structured spousal violence risk assessment: Combining risk factors and victim vulnerability factors. International Journal of Forensic Mental Health, 7, 39-46.

Bonta

Law

Hanson

(1998). The prediction of criminal and violent recidivism among mentally disordered offenders: A meta-analysis. Psychological Bulletin, 123, 123-142.

Brookoff

O’Brien

K. K.

Cook

C. S.

Thompson

T. D.

Williams

(1997). Characteristics of participants in domestic violence: Assessment at the scene of domestic assault. Journal of the American Medical Association, 277, 1369-1373.

Campbell

(1986). Nursing assessment for risk of homicide with battered women. Advances in Nursing Science, 8, 36-51.

Campbell

J. C.

Webster

D. W.

Glass

(2009). The danger assessment: Validation of a lethality risk assessment instrument for intimate partner femicide. Journal of Interpersonal Violence, 24, 653-674.

Carney

M. M.

Barner

J. R.

(2012). Prevalence of partner abuse: Rates of emotional abuse and control. Partner Abuse, 3, 286-335.

Desmarais

S. L.

Reeves

K. A.

Nicholls

T. L.

Telford

Fiebert

M. S.

(2012). Prevalence of physical violence in intimate relationships: Part 1. Rates of male and female victimization. Partner Abuse, 3, 170-198.

10.

de Vogel

Ruiter

C. D.

Hildebrand

Bos

van de Ven

. (2004). Type of discharge and risk of recidivism measured by the HCR-20: A retrospective study in a Dutch sample of treated forensic psychiatric patients. International Journal of Forensic Mental Health, 3, 149-165.

11.

Eke

Hilton

N. Z.

Harris

Rice

Houghton

(2011). Intimate partner homicide: Risk assessment and prospects for prediction. Journal of Family Violence, 26, 211-216.

12.

Family Violence Protection Act 2008. (2017, May 17). Retrieved from http://www6.austlii.edu.au/cgi-bin/viewdb/au/legis/vic/consol_act/fvpa2008283/

13.

Gerth

Rossegger

Bauch

Endrass

(2017). Assessing the discrimination and calibration of the Ontario Domestic Assault Risk Assessment in Switzerland. Partner Abuse, 7, 168-189.

14.

Gray

N. S.

Fitzgerald

Taylor

MacCulloch

M. J.

Snowden

R. J.

(2007). Predicting future reconviction in offenders with intellectual disabilities: The predictive efficacy of VRAG, PCL-SV, and the HCR-20. Psychological Assessment, 19, 474-479.

15.

Hanson

R. K.

Bourgon

Helmus

(2007). The validity of risk assessments for intimate partner violence: A meta-analysis. Ottawa, Ontario, Canada: Public Safety Canada.

16.

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 Behavior, 20, 315-335.

17.

Hilton

N. Z.

Eke

A. W.

(2016). Non-specialization of criminal careers among intimate partner violence offenders. Criminal Justice and Behavior, 43, 1347-1363.

18.

Hilton

N. Z.

Harris

G. T.

(2009). Criminal justice responses to partner violence: History, evaluation and lessons from interventions for criminal conduct. In Lutzker

Whitaker

(Eds.), Preventing partner violence: Research and evidence-based intervention strategies (pp. 219-243). Washington, DC: American Psychological Association.

19.

Hilton

N. Z.

Harris

G. T.

Holder

(2008). Actuarial assessment of violence risk in hospital-based partner assault clinics. Canadian Journal of Nursing Research, 40, 56-70.

20.

Hilton

N. Z.

Harris

G. T.

Popham

Lang

(2010). Risk assessment among incarcerated male domestic violence offenders. Criminal Justice and Behavior, 37, 815-832.

21.

Hilton

N. Z.

Harris

G. T.

Rice

M. E.

(2010). Risk assessment for domestically violent men: Tools for criminal justice, offender intervention, and victim services. Washington, DC: American Psychological Association.

22.

Hilton

N. Z.

Harris

G. T.

Rice

M. E.

Houghton

R. E.

Eke

A. W.

(2008). An in depth actuarial assessment for wife assault recidivism: The domestic violence risk appraisal guide. Law and Human Behavior, 32, 150-163.

23.

Hilton

N. Z.

Harris

G. T.

Rice

M. E.

Lang

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, 267-275. doi:10.1037/1040-3590.16.3.267

24.

Hutchison

I. W.

Hirschel

J. D.

(1998). Abused women: Help-seeking strategies and police utilization. Violence Against Women, 4, 436-456.

25.

Klein

Tobin

(2008). A longitudinal study of arrested batterers, 1995-2005: Career criminals. Violence Against Women, 14, 136-157.

26.

Kropp

P. R.

Hart

S. D.

Belfrage

(2005). Brief Spousal Assault Form for the Evaluation of Risk (B-SAFER) manual. Vancouver, British Columbia, Canada: ProActive ReSolutions.

27.

Kropp

P. R.

Hart

S. D.

Belfrage

(2010). Brief Spousal Assault Form for the Evaluation of Risk (B-SAFER), version 2: User manual. Vancouver, British Columbia, Canada: ProActive ReSolutions.

28.

Kropp

P. R.

Hart

S. D.

Webster

C. D.

Eaves

(1995). Manual for the Spousal Assault Risk Assessment Guide (2nd ed.). Vancouver, British Columbia, Canada: The British Columbia Institute Against Family Violence.

29.

Langhinrichsen-Rohling

Misra

T. A.

Selwyn

Rohling

M. L.

(2012). Rates of bidirectional versus unidirectional intimate partner violence across samples, sexual orientations, and race/ethnicities: A comprehensive review. Partner Abuse, 3, 199-230.

30.

McEwan

T. E.

Bateson

Strand

(2017). Improving police risk assessment and management of family violence through a collaboration between law enforcement, forensic mental health and academia. Journal of Criminological Research, Policy and Practice, 3, 1-12.

31.

Messing

J. T.

Campbell

(2016). Informing collaborative interventions: Intimate partner violence risk assessment for front line police officers. Policing, 10, 328-340.

32.

Messing

J. T.

Campbell

Webster

D. W.

Brown

Patchell

Wilson

J. S.

(2015). The Oklahoma lethality assessment study: A quasi-experimental evaluation of the lethality assessment program. Social Service Review, 89, 499-530.

33.

Messing

J. T.

Thaller

(2013). The average predictive validity of intimate partner violence risk assessments. Journal of Interpersonal Violence, 28, 1537-1558.

34.

Nicholls

T. L.

Pritchard

M. M.

Reeves

K. A.

Hilterman

(2013). Risk assessment in intimate partner violence: A systematic review of contemporary approaches. Partner Abuse, 4, 76-168.

35.

Nimmo

(2012, May). Police family violence response; strengthening systems to enhance prevention and intervention. Paper presented at the Family Violence Symposium, Wellington, New Zealand. Retrieved from https://nzfvc.org.nz/content/strengthening-systems-eliminate-family-violence-family-violence-symposium-2012

36.

Petersson

Strand

Selenius

(2016). Risk factors for intimate partner violence: A comparison of antisocial and family-only perpetrators. Journal of Interpersonal Violence. Advance online publication. doi:10.1177/0886260516640547

37.

Rettenberger

Eher

(2013). Actuarial risk assessment in sexually motivated intimate-partner violence. Law and Human Behavior, 37, 75-86.

38.

Rice

M. E.

Harris

G. T.

(2005). Comparing effect sizes in follow-up studies: ROC, Cohen’s d and r. Law and Human Behavior, 29, 615-620.

39.

Serie

C. M.

van Tilburg

C. A.

van Dam

de Ruiter

(2015). Spousal assaulters in outpatient mental health care: The relevance of structured risk assessment. Journal of Interpersonal Violence, 32, 1658-1677.

40.

Stansfield

Williams

K. R.

(2014). Predicting family violence recidivism using the DVSI-R: Integrating survival analysis and perpetrator characteristics. Criminal Justice and Behavior, 41, 163-180.

41.

State of Victoria. (2016). Royal commission into family violence report and recommendations (Parliamentary Paper No. 132, 2014-2016). Melbourne, Australia: Author.

42.

Storey

J. E.

Kropp

P. R.

Hart

S. D.

Belfrage

Strand

(2014). Assessment and management of risk for intimate partner violence by police officers using the brief spousal assault form for the evaluation of risk. Criminal Justice and Behavior, 41, 256-271.

43.

Thijssen

de Ruiter

(2011). Identifying subtypes of spousal assaulters using the B-SAFER. Journal of Interpersonal Violence, 26, 1307-1321.

44.

Victoria Legal Aid. (2016). De facto and same sex relationships. Retrieved from https://www.legalaid.vic.gov.au/find-legal-answers/separation-divorce-and-marriage-annulment/de-facto-and-same-sex-relationships

45.

Victoria Police. (2014). Victoria police code of practice for the investigation of family violence. Melbourne, Australia: Author.

46.

Williams

K. R.

(2012). Family violence risk assessment: A predictive cross-validation study of the Domestic Violence Screening Instrument-Revised (DVSI-R). Law and Human Behavior, 36, 120-129.

47.

World Health Organization. (2016). Violence against women (Fact sheet No. 239). Retrieved from http://www.who.int/mediacentre/factsheets/fs239/en/