Evaluating the risk assessment tools used by Australian police officers responding to domestic violence incidents: a narrative review

Abstract

Family and domestic violence (FDV) remains a topical issue in Australia. With state incidences of FDV reaching up to 1668 victims per 100,000 persons, the need for tools assessing the likelihood of future offending and recidivism is paramount. The present study reviewed the quality and efficacy of FDV risk assessment tools currently in use by police in each Australian state/territory. The review revealed a large gap in the development and quality assessment of tools employed by Australian police officers. Additionally, when compared to international ‘gold standard’ risk assessment tools, it was evident that the tools currently in use are assessing for too many variables, and that key high-risk factors are absent. It is argued that Australian FDV risk assessment tools need revising, with the development of a nationwide risk assessment tool being the goal of future research in this area.

Family and domestic violence (FDV), often referred to as domestic violence (DV), intimate partner violence (IPV) and wife battering, is a term used to capture abusive behaviours (e.g. physical assault, psychological abuse and rape) within the context of the family setting (World Health Organisation, 2012). It is a social health issue that exists on a global scale (World Health Organisation, 2013). While anybody can be the victim of FDV, IPV – violence specifically perpetrated by and towards an intimate partner – is the most prevalent form, accounting for over one third of all homicides recorded in Australia during 2015 (Australian Bureau of Statistics, 2016).

Despite the use of the word ‘violence’, FDV includes various forms of abusive behaviours that can be absent of any physical abuse, including psychological abuse, financial abuse and sexual abuse. All forms of FDV that an offender engages in are designed to gain control over the targeted family member or result in them being fearful of the offender and future violence (World Health Organisation, 2012). The chronic exposure to the various forms of abuse, labelled ‘coercive control’ sees perpetrators attempting to subjugate family members, forcing them to abide by their demands (Lombard & McMillan, 2012). For the past three decades in Australia, FDV has been a frequent discussion point in political and social sectors, with all governing states and territories committed to reducing the alarming rates of domestic violence occurring at a national level, and to shift societal attitudes around FDV towards zero-tolerance (Bagshaw, Chung, Couch, Lilburn, & Wadham, 2000; Council of Australian Governments., 2010). In addition to this, the federal government of Australia has allocated just under 120 million dollars to FDV over a four-year period from 2015 to 2019, in an attempt to make sustained reductions in the incidences of violence against women (Services DoS, 2015). When examining the statistics in Australia for 2015 (see Table 1), the state-by-state incidences of FDV were as high as 1668 victims per 100,000 persons (Australian Bureau of Statistics, 2016).

Table 1.

Rates of FDV per 100,000 persons in 2015.

State/Territory	FDV incidents per 100,000 persons
New South Wales	400 victims per 100,000 persons.
Western Australia	706 victims per 100,000 persons.
South Australia	456 victims per 100,000 persons.
Northern Territory	1668 victims per 100,000 persons.
Tasmania	232 victims per 100,000 persons.
Australian Capital Territory	177 victims per 100,000 persons.
Victoria	No statistics reported.

Note: FDV = family and domestic violence.

One method currently employed to identify and protect potential victims of FDV is the use of actuarial risk assessment tools. Actuarial risk assessment tools consist of items relating to key factors identified in the literature that indicate an individual is at risk of offending (Singh, Grann, & Fazel, 2011). Items in the assessment tool are assigned a numerical value, responses are then scored, and a corresponding risk estimate is provided, with higher scores indicating the responder being at a greater risk of future exposure to FDV, by the same perpetrator. The risk score is then coupled with professional judgement to determine an appropriate outcome for the offender, victims and other household members; this is termed structured clinical judgement (SCJ; Douglas, Cox, & Webster, 1999). SCJ is a more valid and reliable measure of risk than unstructured clinical judgement (subjective opinion) and is shown to be as equally effective as using a structured tool on its own (Singh et al., 2011). Additionally, SCJ allows for the inclusion of additional information that can be used for risk management, and treatment planning allowing for a superior, holistic approach to intervention (Douglas et al., 1999; Heilbrun, 1997).

In addition to identifying future offenders and victims of FDV, risk assessment tools are also used for assessing the likelihood of offender recidivism. When discussing offender recidivism, the focus is on identified offenders and the likelihood that they will go on to reoffend, which differs from future offenders who have not yet committed any FDV offences, but who are potentially at high risk of doing so. FDV risk assessment tools are useful in instances where FDV offenders are known to police, and information on the likelihood of them continuing to offend is required. Assessing the likelihood of offender recidivism can be useful in determining outcomes such as whether the offender should be allowed to return to the family home and community, or if further interventions such as family violence orders or case management are required (Andrews & Bonta, 2006).

The reliability and validity of actuarial risk assessment tools are an important factor for determining which tools to use to predict future-based risk. Tools that lack strong psychometric properties run the risk of misclassifying risk, which can have adverse effects as extreme as death, should a genuine threat be misclassified. Risk assessment tools tend to be more reliable when they are measuring the presence of risk in specific populations, and the effectiveness of the tools used can be reduced when psychological risk factors are included, due to the subjective and dynamic nature of these (Singh et al., 2011) This presents as a limitation to the assessment of risk within the context of FDV, as victims of FDV tend to experience adverse psychological effects, and psychological abuse is a defined category of FDV, making its assessment necessary (Wolfe, Crooks, Lee, McIntyre-Smith, & Jaffe, 2003). Some risk assessment tools disregard the assessment of psychological risk factors for improved predictive validity; however, in doing so they ignore the potential long-term psychological damage that can occur as a result of FDV. Research has demonstrated the psychological harm experienced by women that results from FDV, with victims being much more likely to experience depression and anxiety than populations that don’t experience FDV (Pico-Alfonso et al., 2006) and that up to half of all women and child victims of FDV experience post-traumatic stress disorder (PTSD; Chemtob & Carlson, 2004). Therefore, there is a need for risk assessment tools to accurately assess the presence of psychological risk, rather than avoiding assessment of this factor, in order to protect individuals from serious psychological harm.

On the frontline of managing Australian FDV cases are members of the police force, who, based on data can be inferred, respond to incidents of domestic violence every 2 minutes (Australian Bureau of Statistics, 2016). When responding to incidents of FDV, the use of actuarial risk assessment tools help to determine the level of risk that an individual might be at, in relation to causing or experiencing harm within the family context. The effectiveness of the tools used by police officers is largely dependent on their ability to correctly identify future offenders and victims of FDV. The need for risk assessment tools to accurately identify at-risk individuals is paramount, as the misclassification of risk can have potentially damaging consequences. Individuals who are at risk of perpetrating future FDV, but are not classified as such, may not be provided with the appropriate assistance and interventions to ensure future violence doesn’t occur. If this occurs, the implications can range from experiencing future violence, to the worst-case scenario of death. On the other hand, incorrectly identifying an individual as being at risk of future perpetration when the risk is low can result in enforced separation from family and children, along with becoming involved in the legal system. Both instances have the potential to cause significant harm to an individual, their family and the broader community, and serve to highlight the importance of using effective risk assessment tools to screen for FDV.

Currently, research on the effectiveness of risk assessment tools used worldwide suggests that there is considerable variability in the effectiveness of the tools available. However, from the research that is available, there are a number of tools that have been identified as being effective for identifying offenders as being at risk of committing future FDV. In their systematic review of FDV risk assessment tools, Nicholls, Pritchard, Reeves, and Hilterman (2013) identified the Ontario Domestic Assault Risk Assessment (ODARA), the Spousal Assault Risk Assessment (SARA) and the Danger Assessment (DA) as effective actuarial assessment tools for the appraisal of FDV risk. These tools have been developed from what the literature dictates as being significant risk factors associated with FDV. Further to this, they are used by members of the police to help guide intervention decisions for offenders and victims of FDV (e.g. arrest/removal of offender, and/or the relocation of victims). The predictive validity of these tools is what sets them apart from other risk assessment tools that are used to determine FDV risk. Research on these tools indicates a 70–77% likelihood of correctly identifying risk within the context of FDV, indicating fair predictive validity (Tape, 2017). When considering that most risk assessment tools used worldwide for FDV exhibit a 60–70% success rate when correctly identifying risk of future FDV exposure (Andrews & Bonta, 2006), it becomes clear that these best practice tools are indeed above the norm, indicating they are the best available risk assessment tools available for identifying individuals who are at risk of future perpetration of FDV.

Given the politico-legal constraints that come with Australia being a federated state, there is currently no single nationwide screening tool used to assess for risk of FDV, with each state determining which tools are used. Having FDV risk assessment managed at a state level has resulted in the use of multiple screening tools, all with different degrees of effectiveness. The lack of a uniform risk assessment tool used by Australian police officers raises questions regarding the effectiveness of the screening tools currently being used. If there is a superior screening tool currently being used by a state within Australia, then its lack of use at a nationwide level puts families elsewhere in this country at unnecessary risk of experiencing further FDV. Additionally, the use of multiple screening tools results in unnecessary replication of efforts being made to identify appropriate and effective screening tools, while also limiting resource sharing between police jurisdictions. The lack of a nationwide risk assessment framework for FDV leads to several questions:

1. Which assessment tools are being used by police to assess risk of FDV in Australia?

2. Of the assessment tools being used, which ones are most effective? And

3. How do the tools used in Australia compare to what research indicates as best practice for risk assessment of FDV?

Given these questions, the aim of this study was to identify the screening tools used by state police departments in Australia, and to compare them in terms of content and predictive utility to the ODARA, SARA and DA, to determine their effectiveness at accurately predicting risk.

Method

Participants

The following Australian state/territory police research departments participated in the study: New South Wales, Tasmania, Western Australia and Queensland.

Procedure

Police departments from each state and territory in Australia were invited to participate in the study. Applications were submitted to each state’s police research department, requesting information on the specific FDV risk assessment tool used by police officers. The information requested included a copy of the tool used and any research regarding the development or effectiveness of the tool. Of the eight states and territories in Australia, New South Wales, Tasmania, Western Australia and Queensland chose to participate in the current study. Victoria and South Australia police departments chose not to participate, and the Northern Territory and Australian Capital Territory did not return contact. The state and territory police departments were asked to provide any available documents regarding the FDV risk assessment tools in use within their jurisdiction. Information on risk assessment tools used by the remaining states and territories were obtained through the McCulloch, Maher, Fitz-Gibbon, Segrave, and Roffee (2016) review on the Family Violence Risk Assessment and Risk Management Framework (CRAF), with copies of the identified screening tools obtained through a literature search, where available. From here, a review of the available literature was conducted on each state/territory’s FDV risk assessment tool.

Search strategy

A systematic search was conducted using Elsevier, Scopus, Ovid and Google Scholar. The search criteria consisted of the full names of each identified screening tool, along with their abbreviations. As the search was for very specific literature regarding the FDV risk assessment tools used by Australian police officers, it was determined that names and abbreviations of the tools used would be enough to cover search terms. Studies were limited to the English language, and those not published in peer-reviewed journals (grey literature) were used in addition to those contained in peer-reviewed journals. Studies that were included in the review reported evidence of measuring the predictive validity of the FDV risk assessment tools used in Australia.

Data analysis

Statistical appraisal of tools

Risk assessment literature commonly uses the area under the receiver operating characteristic curve (AUC) to report the effectiveness of a given tool. The AUC is represented as a single outcome value with 95% confidence, which summarises the overall effectiveness of a risk assessment screening tool by plotting the false-positive and true-positive rates of the tool across a score threshold. The AUC is therefore an indication of the risk assessment tool’s ability to accurately discriminate between offenders and non-offenders across all cut-offs. It is worth noting that the ability to discriminate between offenders and non-offenders forms only one piece of the puzzle toward reducing incidents of FDV. Douglas and Kropp (2002) outline the important role of effective risk and behavioural management in reducing violence, drawing on a preventative approach, rather than one of risk identification (Douglas & Kropp, 2002).

Risk assessment tool ranking

The FDV risk assessment tools used by Australian police officers were ranked according to the number of risk factors present that were consistent with ‘gold standard’ assessment tools, along with the tool’s predictive utility. The Australian FDV risk assessment tools that aligned closely with ‘best practice’ tools and sound predictive utility were ranked greater than those that did not align and demonstrated poor predictive utility.

International ‘gold standard’ FDV risk assessment tools

The Ontario Domestic Assault Risk Assessment (ODARA)

The ODARA is a 13-item actuarial risk assessment tool (see Table 2) that was developed to assess the risk of recidivism in FDV offenders (Hilton et al., 2004). Items on the ODARA are a result of research on 589 FDV offenders, exploring offender characteristics and factors that contribute to the likelihood of reoffending. The 13 resulting ‘yes/no’ items that make up the ODARA were offender characteristics that correlated most with the likelihood of reoffending.

Table 2.

Risk factors measured by the ODARA.

ODARA items
Previous assault (domestic/non-domestic).
Previous jail term ≥30 days.
Failed to adhere to previous conditional release.
Threats to harm/kill in current offence.
Unlawful restraint of victim during current offence.
Victim fears continued violence.
Offender and victim have multiple children together.
Offender is a stepfather to the children.
Offender is non-domestically violent.
Multiple indications of drug/alcohol abuse.
Victim assaulted when pregnant.
Barriers to victim receiving support.

Note: ODARA = Ontario Domestic Assault Risk Assessment.

The ODARA has been demonstrated to have an AUC ranging from .64 (95% CI ± .09, where CI = confidence interval) to as high as .77 (95% CI ± .04; Hilton et al., 2004). The success of the ODARA as an actuarial risk assessment tool for FDV is the result of rigorous empirical methodology and developers assessing only the variables that correlate to offender recidivism. Singh et al. (2011) highlighted the importance of risk assessment tools in refraining from trying to assess too many variables as it reduces the tool’s overall predictive validity, and the ODARA has successfully managed to achieve that with a 13-item risk assessment tool. Therefore the 13 risk factors in the ODARA should be included in all other FDV risk assessment tools to enhance their predictive validity.

The Spousal Assault Risk Assessment (SARA)

The SARA was developed as a checklist to screen for the risk factors of FDV (Kropp, Hart, Webster, & Eaves, 1999). Consisting of 20 items (see Table 3), the SARA assesses FDV risk factors across five key domains: criminal history, psychosocial adjustment, spousal assault history, alleged/most recent offence (index offence) and other considerations. Items for each section were determined after an extensive literature review into risk factors associated with FDV, with the aim of keeping the SARA brief and with moderate specificity relative to other FDV assessments available.

Table 3.

Risk factors measured by the SARA.

SARA items
Past assault on family members. Past assault on strangers.
Previous violation of parole/conditional release or no contact orders.
Recent relationship problems.a Recent employment problems.a
Offender exposed to FDV at a young age.
Offender substance use/dependence.a
Offender exhibits suicidal/homicidal ideation or intent.a
Offender exhibits psychotic/manic symptoms.a
Offender exhibits the presence of a personality disorder with aggression.a
Offender previously convicted of physical assault. Offender previously convicted of sexual assault.
Previous use of a weapon or death threats.
Increase in frequency or severity of abuse by offender.a Previous violation of ‘no contact’ orders.
Offender minimises or denies spousal assault.a
Offender exhibits attitudes condoning assault.a Most recent offence included severe and/or sexual assault. Most recent offence included the use of weapons and/or credible death threats. Most recent offence violates current ‘no contact’ orders.

Note: SARA = Spousal Assault Risk Assessment.

^aDynamic risk factors.

The SARA has been demonstrated to have an AUC ranging from .52 (95% CI ± .11) to .70 (95% CI not reported; Grann & Wedin, 2002; Heckert & Gondolf, 2004; Kropp & Hart, 2000; Williams & Houghton, 2004). The success of the SARA can be attributed to the extensive review of the literature in guiding the formulation of items to assess FDV; however, the inclusion of dynamic factors associated with psychosocial adjustment (refer to items in Table 3 marked with superscript a) could attribute to the tool’s reduced predictive validity, relative to the ODARA (Singh et al., 2011). The SARA’s inclusion of psychological variables and its ability to contend with the high predictive utility of the ODARA demonstrates that the inclusion of such variables can occur in effective FDV risk assessment tools. Given the significance that psychological factors can have on FDV and the SARA’s proven ability to effectively incorporate these into a risk assessment tool, it is arguably a benchmark tool suitable for comparisons to be made in this study.

The Danger Assessment (DA)

The DA is a self-report risk assessment for adult female victims consisting of 19 yes/no items (see Table 4) and one item assessing the frequency of IPV that a woman has experienced over the past year (Campbell, 2001). The DA differs from the ODARA and SARA in that it assesses the risk of the female being a victim of intimate partner homicide (IPH). Items on the DA were determined after consultation with shelter workers, law enforcement officials, battered women and experts on battering (Campbell, 1986). The revised, and current, version of the DA came as a result of further multivariate research into IPV and IPH indicating significant factors absent or in need of revision in the original DA. Reliability and validity of the original DA has been demonstrated to be acceptable, and for the version currently in use, the AUC ranges from .59 (95% CI ± .05) to .92 (CI not reported) depending on the criterion variable assessed (Campbell, Webster, & Glass, 2009; Heckert & Gondolf, 2004).

Table 4.

Risk factors measured on the DA.

DA items
Offender violence increased in past 12 months.
Offender use, or threats of using weapon.
Offender violence includes choking.
Offender possesses a gun.
Offender has engaged in previous rape.
Offender uses illicit drugs.
Offender threats to kill, or is capable of murder.
Offender is drunk every/most days.
Offender engages in controlling behaviours.
Victim assaulted while pregnant.
Offender prone to jealousy.
Victim has threatened/attempted suicide.
Offender has threatened/attempted suicide.
Offender has threatened to harm children.
Offender is not the father of the children.
Offender is unemployed.
Victim has left during the past year.
Victim has a different intimate partner.
Offender engages in stalking behaviours.

Note: DA = Danger Assessment.

The success of the DA has been attributed to it specifically being designed to measure risk of lethal violence, an area that is considered under-represented (Nicholls et al., 2013). However, despite its purpose, the DA tends to receive more use for assessing the risk of IPV recidivism, which can potentially result in increased risk of homicide being misclassified as risk of violence. This highlights a limitation of research on the DA, as there are discrepancies between variables that contribute to IPV versus IPH, resulting in potentially misleading results from research.

Comparative variables

As demonstrated by high AUCs, the ODARA is argued to be the most accurate FDV risk assessment tool currently in use, globally, by police officers (Hilton et al., 2004; Nicholls et al., 2013). Given the high quality of the ODARA, and the empirical approaches used during development, all items on the ODARA were used to assess the quality of the risk assessment tools used by Australian police officers.

Research by Kropp and Hart (2000) has shown that items on the SARA specifically exploring partner violence risk are most useful for determining IPV recidivism. Additionally, further research has found that five items on the assessment tool have a positive relationship with partner violence recidivism (Grann & Wedin, 2002). These items are: past violation of conditional release, the presence of a personality disorder with unstable characteristics, past physical assault, violation of no-contact orders and minimisation or denial of spousal assault history. As some of these items overlap with items used on the ODARA, only the presence of a personality disorder with unstable behaviour and minimisation or denial of spousal assault were included as comparative items.

As the DA assesses for intimate partner homicide, which is the most extreme consequence of FDV, factors that assess for risk of homicide should be included in FDV risk assessment tools. Research on risk factors associated with intimate partner homicide has highlighted several factors included in the DA that contribute to the increased likelihood of intimate partner homicide. These are: the abuser being unemployed, having access to a firearm, recent relationship separation and the abuser being the step-father to children in the household (Campbell et al., 2003). As some of these items are included in both the ODARA and the SARA, the following DA items were included for comparison: abuser being unemployed, abuser having access to a firearm and separation from a previously cohabitating relationship.

Additionally, when risk assessment tools try to assess for too many risk factors, it comes at the cost of the tool’s predictive utility (Singh et al., 2011). As a result, the additional comparison variable ‘includes ≤20 items’ was included when assessing the effectiveness of Australian FDV risk assessment tools (see Table 5).

Table 5.

Variables included for comparison to Australian FDV risk assessment tools and ‘gold standard’ tools they derive from.

Item	Tool
Previous assault (domestic/non-domestic).	ODARA/SARA
Previous jail term ≥30 days.	ODARA
Failed to adhere to previous conditional release.	ODARA/SARA
Threats to harm/kill in current offence.	ODARA
Unlawful restraint of victim during current offence.	ODARA
Victim fears continued violence.	ODARA
Multiple children together.	ODARA
Offender is stepfather to the children.	ODARA
Non-domestically violent.	ODARA
Multiple indications of drug/alcohol abuse.	ODARA
Victim assaulted when pregnant.	ODARA
Barriers to victim receiving support.	ODARA
Personality disorder with unstable behaviours.	SARA
Minimisation or denial of spousal assault.	SARA
Abuser unemployed.	DA
Abuser access to firearm.	DA
Separation from previously cohabitated relationship.	DA
Includes ≤20 items.	ODARA/SARA/DA
Research demonstrating modest predictive utility (AUC ≥ .7).

Note: FDV = family and domestic violence; ODARA = Ontario Domestic Assault Risk Assessment; SARA = Spousal Assault Risk Assessment; DA = Danger Assessment; AUC = area under the receiver operating characteristic curve.

The last comparison variable to be included regarded the predictive utility of the tool according to available research. As Hosmer and Lemeshow (2000) state AUC scores ≥.7 (95% CI) to be indicative of acceptable predictive utility and not a result of chance, literature demonstrating Australian FDV risk assessment tools to reach this threshold was also included for comparison.

Results

Tools used by Australian police officers

In Australia, there is no single tool used to assess for future risk of FDV; instead, each state has developed and/or implemented their own tools. The tools used at a state level are as detailed below. Table 6 provides a state-by-state summary of the results obtained in this study. In-depth, descriptive results (including the risk assessment tool name), are contained within the written results section.

Table 6.

State-by-state summary of results.

State	Number of items	Provides risk score	Structured clinical judgement	Supporting research	Comparison score	Ranking
South Australia (SAPRAF)	60	Yes	Yes	No	12	1
Northern Territory (CRAF–NT)	59	Yes	Yes	No	12	1
Western Australia (CRARMF)	35	No	No	No	12	1
Victoria (CRAF–V)	26	No	No	Yes	11	2
Tasmania (RAST)	34	Yes	Yes	Yes	9	3
Queensland (DV-PAF)	22	No	No	No	9	3
New South Wales (DVSAT)	34	No	No	No	8	4
Australian Capital Territory	0	N/A	N/A	N/A	0	5

Note: SAPRAF = South Australian Police Risk Assessment Form; CRAF–NT = Common Risk Assessment Form – Northern Territory; CRARMF = Common Risk Assessment and Risk Management Framework; CRAF–V = Family Violence Risk Assessment and Risk Management Framework – Victoria; RAST = Risk Assessment Screening Tool; DVSAT = Domestic Violence Safety Assessment Tool.

Note: there is currently no state-wide risk assessment tool used to assess for FDV in the Australian Capital Territory (McCulloch et al., 2016). Assessment of risk in the Australian Capital Territory is at the discretion of police judgement. When addressing this in the current research, the Australian Capital Territory received a comparison score of zero, and was ranked last as they had no items available to map onto the comparison variables.

The Risk Assessment Screening Tool (RAST) – Tasmania

The RAST consists of 34 items, with 18 high-risk factors and 16 other risk factors. The RAST is completed by police officers attending incidents of FDV and produces a risk score according to the number of risk factors that have been identified. High-risk factors are scored 3, and other risk factors receive a score of 2. Total scores are used to determine the risk of an individual being subject to future incidents of FDV by the perpetrator in question. Risk levels are assessed as being low (0–13), medium (Bureau of Crime Statistics and Research, 2017; Campbell, 1986, 2001; Campbell et al., 2009; Chemtob & Carlson, 2004; Douglas & Kropp, 2002; Grann & Wedin, 2002; Heckert & Gondolf, 2004; Hilton et al., 2004; Kropp & Hart, 2000; Kropp et al., 1999; Nicholls et al., 2013; Tape, 2017; Williams & Houghton, 2004), or high (28+). There is also a section on the RAST where the police officer uses professional judgement to determine the reliability of the information source and the accuracy of information provided.

Family Violence Risk Assessment and Risk Management Framework – Victoria (CRAF–V)

The CRAF–V consists of 26 yes/no items associated with increased risk of FDV reoccurring in the household (McCulloch et al., 2016). In addition to the 26 yes/no items, the CRAF–V also has response sections regarding the victim’s presentation, assessment of their own safety and current protective factors, to aid in risk management planning. The CRAF–V does not provide a total risk value to officers, and is used to guide professional judgement. This professional judgement results in the officer determining members in the household as being at risk, at elevated risk or requiring immediate attention regarding future exposure to FDV by the same perpetrator.

The Domestic Violence Safety Assessment Tool (DVSAT) – New South Wales

The DVSAT is a two-part FDV risk assessment tool, consisting of a total of 34 items. Part A: The Intimate Partner Risk Identification Checklist consists of 25 yes/no items, while Part B: Other Identification/Professional Judgement consists of nine items with a mix of closed and open-ended questions. The DVSAT provides no total risk score, and risk factors identified are used to guide officer’s professional judgement of offender’s recidivism risk.

South Australian Police Risk Assessment Form (SAPRAF) – South Australia

The SAPRAF consists of 60 items that are totalled to produce an overall risk score that falls into three categories: standard (0–23), medium (24–44) and high (45+). The total risk score is used to give an indication of the likelihood of reoffending and may be overridden by the officers’ professional judgement if they believe victims to be at high risk.

The Common Risk Assessment Form – Northern Territory (CRAF–NT)

The CRAF–NT consists of 59 items similar to the SAPRAF. The CRAF–NT and the SAPRAF consist of nearly all the same items, save for items in the CRAF–NT assessing offenders’ history of trauma. Like the SAPRAF, the CRAF–NT also provides a total risk score that falls into the same three categories: standard (0–23), medium (24–44) and high (45+). The total risk score is used to give an indication of the likelihood of reoffending and may be overridden by the officers’ professional judgement if they believe victims to be at high risk.

The Common Risk Assessment and Risk Management Framework (CRARMF) – Western Australia

The CRARMF consists of multiple components, the first of which is the common screening tool, which consists of three close-ended questions and one additional information section designed to detect the possible presence of FDV in the household. If the use of the screening tool detects the possible presence of FDV, the 35-item risk assessment tool is administered. Level of risk is determined by the officer as being at risk of harm or at high risk of serious harm, determined according to the officer’s judgement of risk factors present.

The Domestic Violence Protective Assessment Framework (DV-PAF) – Queensland.

The DV-PAF is a 22 item FDV risk assessment tool used to guide clinical judgement by Queensland Police Officers in response to FDV call outs. The DV-PAF also includes a section on the victims fear of the perpetrator. While the DV-PAF does not provide a risk score, items are used to assess the severity of risk across 4 domains. Level of risk is determined by the officer as being unknown, medium, high, or extreme, determined according to the officers’ judgement of risk factors present.

Literature review

The names of each state’s FDV risk assessment tools were used to review existing literature assessing predictive utility. The review yielded two publicly available studies that assessed the effectiveness of the FDV risk assessment tools used by Australian police officers. Mason and Julian (2009) reported the effectiveness of the Tasmanian RAST to be of modest predictive utility, with an AUC of .60 (CI not reported). Millsteed and Coghlan (2016) assessed the predictive utility of the L17, which is a component of the Victorian CRAF–V, and report an AUC of .72 (±.01), indicating good predictive utility.

Comparison matrix

The quality of FDV risk assessment tools used by Australian police officers was determined according to how closely their features mapped onto features seen in the ODARA/Domestic Violence Risk Appraisal Guide, SARA, and DA. The features of each FDV risk assessment tool used by Australian police officers were explored to determine how many of the 19 previously determined factors were present for each risk assessment tool. Tools used by Australian police officers were then ranked according to how many of the 19 features they exhibited, with higher rankings indicating the greater presence of features. Table 7 shows a visual matrix representing the features exhibited by each state’s screening tool, marked with an X.

Table 7.

Matrix depicting the degree of overlap between Australian FDV risk assessment and international ‘best practice’ tools.

	RAST (Tasmania)	CRAF– V(Victoria)	DVSAT (New South Wales)	SAPRAF (South Australia)	CRAF– NT(Northern Territory)	CRARMF (Western Australia)	DV-PAF – (Queensland)	Australian Capital Territory – no FDV risk assessment identified
Previous assault (domestic/non-domestic)	X	X		X	X	X	X
Previous jail term ≥30 days
Failed to adhere to previous conditional release.	X	X	X	X	X	X
Threats to harm/kill in current offence.	X	X	X	X	X	X	X
Unlawful restraint of victim during current offence.				X	X	X
Victim fears continued violence.						X	X
Multiple children together.
Offender is stepfather to the children.			X	X	X	X
Non-domestically violent.	X	X		X	X	X
Multiple indications of drug/alcohol abuse.	X	X	X	X	X	X	X
Victim assaulted when pregnant.	X	X	X	X	X	X	X
Barriers to victim receiving support.		X		X	X		X
Personality disorder with unstable behaviours.
Minimisation or denial of spousal assault.
Abuser unemployed.	X	X	X	X	X	X	X
Abuser access to firearm.	X	X	X	X	X	X	X
Separation from previously cohabitated relationship.	X	X	X	X	X	X	X
Includes ≤20 items
AUC ≥ .7		X
Total	9	11	8	12	12	12	9	0

Note: X=Presence of the comparative feature in the relative risk assessment tool;

blank cells=absence of the feature in the relative tool;

total=combined number of features the risk assessment tool possesses; FDV = family and domestic violence; RAST = Risk Assessment Screening Tool; CRAF–V = Family Violence Risk Assessment and Risk Management Framework – Victoria; DVSAT = Domestic Violence Safety Assessment Tool; SAPRAF = South Australian Police Risk Assessment Form; CRAF–NT = Common Risk Assessment Form – Northern Territory; CRARMF = Common Risk Assessment and Risk Management Framework; DV-PAF = Domestic Violence Protective Assessment Framework; AUC = area under the receiver operating characteristic curve.

Following the mapping of common features found in ‘best practice’ FDV risk assessment tools and how the tools used in Australia compare, the eight state/territories were ranked according to the total number of features similar to ‘best practice’ tools (Table 8).

Table 8.

Screening tool rankings according to total number of features present.

FDV risk assessment tool	Rank	Score
South Australian Police Risk Assessment Form (SAPRAF) – South Australia	1	12
Common Risk Assessment Form (CRAF–NT) – Northern Territory	1	12
Common Risk Assessment and Risk Management Framework (CRARMF) – Western Australia	1	12
Family Violence Risk Assessment and Risk Management Framework (CRAF–V) – Victoria	2	11
Risk Assessment Screening Tool (RAST) – Tasmania	3	9
Domestic Violence – Protective Assessment Framework (Queensland)	3	9
Domestic Violence Safety Assessment Tool (DVSAT) – New South Wales	4	8
Australian Capital Territory: No tool used.	5	0

Note: FDV = family and domestic violence.

Discussion

This narrative review investigated the presence of ‘best practice’ features for FDV risk assessment in actuarial tools used by Australian police officers and the predictive utility of the FDV risk assessment tools used by Australian police officers. Data were collected from six of the eight states and territories in Australia surrounding the FDV risk assessment tools used, and any published or unpublished literature they had in their possession. Further to this, data were collected from two studies identified in a review of the literature on the predictive utility of the risk assessment tools provided by Australia’s state police departments. The focus of this review was to explore the following questions: (a) Which assessment tools are being used by police to assess risk of FDV in Australia? (b) Of the assessment tools being used, which ones are most effective? (c) How do the tools used in Australia compare to what research indicates as best practice for risk assessment of FDV?

When comparing the FDV risk assessment tools used by Australian police officers with best practice tools used internationally, several characteristics stand out. Firstly, there appears to be common foundational factors that each assessment tool measures: previous assaults, alcohol and drug abuse, victim pregnancy when abused, abuser unemployed, the offender having access to a firearm and previous relationship breakdown between offender and victim. However, in addition to these features, all FDV risk assessment tools used by Australian police officers exceeded 20 items. There are multiple implications of having risk assessment tools assessing for too many items, such as reducing the predictive utility of the tool used (Singh et al., 2011). Further to this, when police officers attend FDV call outs, the situation is one of high stress, and the need for a risk assessment measure that is easy to administer and time friendly is important for all parties involved. Research on police attitudes towards the use of FDV risk assessment tools in England and Wales highlights the importance of using a tool that considers and caters to the practical role of police duties and to ensure officers are well trained in administration. As high as 86% of police officers endorse the need for a shorter tool, with some officers concerned that asking a large number of risk-related questions to a victim is not appropriate, given the crisis situation they have experienced (Robinson, Myhill, Wire, Roberts, & Tilley, 2016). For victims and offenders, FDV situations are distressing, and the need for efficient tools is important to ensure the individuals they are administered to are not spending too much time recalling the distressing situation. Further to this, McCulloch et al. (2016) highlight concerns that members of Victorian Police have reported their current risk assessment tool (CRAF–V) as too time-consuming to complete, resulting in officers completing it at the station, rather than at the incident. Increasing the time between receiving the information from the offender or victim and when this information is documented increases the likelihood of inaccurate recording that can result from delayed recall. The normal cognitive processes that occur as time increases between an event and its recall include: retrospective bias, the availability heuristic and the fundamental attribution error (Robinson-Riegler & Robinson-Riegler, 2012). The normal cognitive processes that result in inaccurate recall of events, coupled with police officers waiting until returning to the station before filling out FDV risk assessments, raises concerns for the validity of information derived from risk assessment tools being used by Australian police officers. This situation further highlights the importance of encouraging completion of risk assessment tools at the incident, which necessitates brevity in item selection.

Secondly, none of the FDV risk assessment tools used by Australian police officers measure whether the offender has a personality disorders, or has received a previous jail term of more than 30 days. Additionally, the number of children in the relationship and whether the offender denies or minimises the spousal assault is also absent in all the FDV risk assessment tools used in Australia. The absence of factors that have been proven to be predictive of high risk is concerning, especially considering personality disorders can result in poor emotional and behavioural regulation (Casey, Rogers, Burns, & Yiend, 2013). Reduced emotional and behavioural regulation inhibits an individual’s ability to control their temper and think through situations and makes them more likely to engage in rash behaviours that can have negative consequences (Gross, 2007). For FDV offenders, reduced behavioural and emotional regulation increases the likelihood that they will reoffend given their tendency to act on impulse, rather than in a way that is well considered and socially, emotionally and behaviourally appropriate.

Thirdly, there is a significant difference in the amount of research conducted assessing the effectiveness of ‘best practice’ risk assessment tools and those used by Australian police officers. When reviewing the effectiveness of the FDV risk assessment tools used by Australian police officers, it became clear that there is currently a significant gap in the literature. While it does appear the assessment tools were developed from the basis of existing FDV literature, follow-up studies exploring the effectiveness of the FDV risk assessment tools were only identified for the RAST (Tasmania) and CRAF–V (Victoria). This is an alarming finding for the remaining states and their FDV risk assessment tools, as there is no clear statistical evidence as to the effectiveness of the tools used by police officers at assessing the risk of offender recidivism. Having no indication of a tool’s effectiveness has dangerous implications if the tools being used are ineffective. If the FDV risk assessment tools being used by police officers are not effective, then offenders could be consistently and repeatedly misclassified in terms of the risk of future perpetration. Without assessing the effectiveness of the tools being used, individuals around Australia may be at risk of serious harm or death.

Tasmania’s RAST was reported by Mason and Julian (2009) to have an AUC of .60 (CI not reported), indicating that the tool currently in use by Tasmanian police officers performs only slightly better than chance (50%). Further to this, the RAST was empirically very close (0.4%) to be considered a failed tool (AUC < .60), indicating significant room for improvement. The Victorian CRAF–V, on the other hand, demonstrated greater predictive utility, with an AUC of .72 (±.01), indicating good predictive utility; however, as the CRAF–V is missing some key factors that assess FDV risk, there is still room for this tool to be improved. It is worth noting that the reason Victoria Police decline to participate in the study is because they are currently developing a new tool to assess for FDV risk (Victoria Police, personal communication, June 21, 2017), which will necessitate further research be conducted to ensure the new tool’s effectiveness. Available research on the risk assessment tools used by Australian police officers and their comparison to ‘best practice’ tools indicates that the current state of Australian FDV risk assessment needs to be improved and reveals an area for future research to explore. Although having state administrative subdivisions in Australia results in FDV risk assessment being managed at a state, rather than at nationwide, level the sheer number of FDV risk assessment tools currently circulating in Australia is problematic, as it makes assessing their effectiveness from a research standpoint difficult. Multiple researchers assessing up to six assessment tools is less effective than a collective of researchers assessing and advancing a single risk assessment tool. Hence the nationwide development and use of a single FDV risk assessment tool would make future research regarding the quality of the tool easier and would subsequently have a greater impact in terms of assessment quality, compared to a state-level approach. The use of a single nationwide tool would also allow for the sharing and easy interpretation of findings between states, allowing better communication to occur between police jurisdictions. Additionally, a single research-driven tool with enhanced predictive utility would also benefit offenders and victims in that it would be more effective at accurately assessing the risk of offender recidivism, allowing appropriate action to be taken.

In addition to more accurate and effective assessment of risk, an improved, nationwide tool would allow for better interventions to be implemented for both the offender and victim in FDV incidents. Effective planning and interventions rest on the accurate identification of target problems; and the identification of target problems is difficult when high-risk variables are absent in assessment, or when assessing for too many variables provides an inaccurate range of variables for interventions to map onto. The development of a single nationwide tool would allow for interventions to be developed that could be implemented in all states and territories in Australia. This would be beneficial for high-risk individuals and families that move interstate, as there would be consistency in interventions, and police officers and other professionals involved in protection and rehabilitation within FDV contexts (e.g. support workers, psychologists) in the new state would be able to continue with the interventions already in place from other jurisdictions.

As with the existing FDV tools in Australia, the current study is not without limitation. The ranking of risk assessment tools according to best practice features was not done so according to any statistical analysis, making the interpretation of the actual effectiveness of the risk assessment tools examined impossible. While the rankings according to best practice features may correlate with potential effectiveness of the risk assessment tool, it is important that interpretations according to best practice ranking are made with caution, serving to further highlight the broader problem regarding the lack of research conducted in the area of Australian FDV risk assessment.

Conclusion

In conclusion, this study aimed to identify and assess the effectiveness of FDV risk assessment tools used by Australian police officers by comparing them with best practice international tools and reviewing existing literature on each state’s tool. All the FDV risk assessment measures currently in use by Australian police officers are assessing for too many variables, which increases the risk of them being completed incorrectly or sometimes not at all. In addition to this, the high-risk FDV variables – personality disorder with unstable behaviours, previous jail term ≥30 days, offender and victim having multiple children together and the offender’s minimisation/denial of spousal abuse – appear to be absent in all risk assessment tools used by police in Australia. When reviewing existing literature on the effectiveness of FDV risk assessment tools being used in Australia, it becomes clear that there is not enough research being conducted assessing whether the tools being used actually do what they are designed to. Future research around FDV risk assessment in Australia should explore the development of a nationwide risk assessment measure that incorporates features from best practice risk assessment tools, is easy to administer and is regularly assessed post-development for effectiveness.

Correction Statement

This article was originally published with errors, which have now been corrected in the online version. Please see Correction (http://dx.doi.org/10.1080/13218719.2022.2057785)

Ethical standards

Declaration of conflicts of interest

Jesse Richardson has declared no conflicts of interest

Kimberley Norris has declared no conflicts of interest

Ethical approval

All procedures performed in studies involving human participants were in accordance with the ethical standards of the Tasmanian Social Sciences Human Research Ethics Committee (ethics number: H0015332) and with the 1964 Helsinki declaration and its later amendments or comparable ethical standards.