Domestic violence risk assessment is the evaluation of characteristics of an offender, offense, or other circumstances considered relevant to the risk that an offender will commit a domestic violence offense in the future. Whether information is considered relevant and how it is weighed depend on the approaches taken to risk assessment. In the two leading approaches, these decisions are based either on empirical research leading to actuarial data that permit individual risk to be estimated using the strongest predictors known for the offender population or on a broader assessment of the individual leading to an overall judgment of risk or a clinical case formulation. Domestic violence risk assessment is important to the extent it aids the risk principle of the Risk-Need-Responsivity model to be carried out. All approaches to domestic violence risk assessment have in common the intention to identify priorities for responding to offender risk, whether in terms of criminal justice intervention, offender treatment, or other forms of risk management.
Domestic violence risk assessment is different from screening, which is used to identify individuals who might be experiencing abuse. It is also different from the function of risk management, in which attempts are made to mitigate the likelihood of an offense occurring among offenders identified at risk of domestic violence.
This article provides a context for domestic violence in relation to other criminal behavior, describes the most widely researched specialized risk assessment instruments, and looks to future research avenues for further applications.
Domestic Violence and Other Criminal Behavior
Developments in the assessment of risk of domestic violence followed on the success of research constructing and validating tools to predict violent and sexual recidivism (reoffending). The two fields share certain challenges, including reliance on official charges or convictions (an exception being some research on the Danger Assessment, described in the next section, which was developed for its ability to assess risk of self-reported victimization). Criminal charges substantially underrepresent the actual occurrence of offending. In addition, the qualifier of sexual or spousal is often omitted or dropped from assault and similar charges, making it difficult to determine whether or not the outcome criteria have been met. When cases with ambiguous recidivism (e.g., the victim of a new offense cannot be determined from available evidence) are excluded from analysis, predictive accuracy increases. Another approach is to compare only offenders known to recidivate domestically with those who only reoffend against nonintimate partners. However, this approach might obscure domestic offenders’ criminal diversity, for which there is growing empirical evidence that might be pertinent to assessing domestic violence risk.
Domestic Violence Risk Assessment Instruments
Danger Assessment
The first specialized domestic violence risk assessment to be published and evaluated was the Danger Assessment, which appeared in the mid1980s. The intention in creating the Danger Assessment was to help women evaluate their risk of being killed by an abusive male partner. The instrument was designed to be used as part of a structured interview, combining a calendar-based assessment of the history of domestic violence with a 20-item risk assessment tool, revised from an original 15 items. The items cover information such as the history of violence, threats, and other abuse, weapons, substance use, and suicidal histories of the offender and victim. Scores on this tool are used to classify risk as variable, increased, severe, or extreme danger. Several studies have shown that scores on the Danger Assessment are associated with subsequent domestic violence recidivism as well as with the severity and type of abuse (e.g., intimate partner sexual assault). The Danger Assessment is also associated with victim outcomes, including depression, post-traumatic stress disorder, substance use, and taking precautions for safety from the violence. The Danger Assessment has been modified based on research for different populations, including immigrant women and women in same-sex relationships. In case–control research, the Danger Assessment score has been found to distinguish between women who were killed and women who were abused but not killed.
Spousal Assault Risk Assessment Guide (SARA)
The SARA, intended for assessing the risk of domestic violence perpetration, first appeared in the 1990s. This instrument was designed to be used after gathering information on the offender from official reports as well as offender and possibly victim interviews. The 20-item risk assessment tool covers information such as the offender’s history of violence and criminal justice violations, childhood abuse, mental disorder, threats, weapons, substance use, and suicidal history. The assessor uses this tool, as well as items or other information determined to be critical, to classify risk as low, moderate, or high. The SARA has been validated as a predictor of domestic violence recidivism in several studies using the total score or the summary risk judgment. In 2015, a revision was released that added items pertaining to victim vulnerability and omitted the item scoring procedure in favor of a clinical case formulation approach. In this SARA-V3, assessors are encouraged to identify factors that might motivate, disinhibit, or destabilize violent behavior as well as the scenarios under which violence might occur. Opinions are then offered as to management plans, case prioritization, and the risk of violence or other outcomes. This theoretical conceptualization approach requires more expertise and analysis than might be practicable for frontline policing or other first responders such as those working in shelters or emergency departments.
Brief Spousal Assault Form for the Evaluation of Risk (B-SAFER)
Related to the SARA is the B-SAFER that was designed for frontline use in the criminal justice section. The original 10-item risk assessment tool covers information such as the offender’s history of violence and criminal justice violations, mental disorder, threats, substance use, and employment; 5 items were later added concerning victim vulnerability. Initial research showed that scores on the B-SAFER can predict subsequent domestic violence.
Ontario Domestic Assault Risk Assessment (ODARA)
The first specialized domestic violence risk assessment tool to be developed using statistical methods was the ODARA. This instrument was designed to be used by police officers during a domestic violence investigation and as a guide to conducting a victim interview to obtain the information that was subsequently developed for use in victim support services. The 13-item risk assessment tool covers information such as the offender’s history of violence and criminal justice involvement, threats, weapons, substance use, and victim barriers to support. Each item is scored 0 or 1, and the total score is used to compare an offender with others in terms of risk of domestic violence recidivism, using an actuarial table that contains statistical data from a large follow-up study of outcomes. Several validation studies have been conducted that support the ODARA’s ability to assess domestic violence recidivism risk in police and correctional cases. Some research has found that the ODARA score is also associated with co-occurring child abuse and with nonviolent, nondomestic violence, and other criminal outcomes by domestic offenders.
Domestic Violence Risk Appraisal Guide
Related to the ODARA is the Domestic Violence Risk Appraisal Guide that was designed for use when greater time and resources are available. It was created through research revealing that predictive accuracy could be improved by combining the ODARA items with the Psychopathy Checklist–Revised. In cross-validation research, the Domestic Violence Risk Appraisal Guide was associated with domestic violence recidivism and with the subsequent number of incidents, victim injury, severity, and seriousness of criminal charges.
Domestic Violence Screening Instrument–Revised (DVSI-R)
The algorithmic and professional judgment methods of assessing risk are combined in the DVSI-R. This instrument was designed to be used by probation officers. The original 12-item risk assessment tool was developed using large-scale empirical research identifying characteristics of offenders with a history of domestic violence; these characteristics were selected as items if they were also associated with repeated domestic violence in a literature review. It was revised to 11 items following further empirical research. The resulting DVSI-R covers information such as the offender’s history and current violence, arrests and convictions, criminal justice violations, weapons, substance use, and employment status. Items are scored on a 3- or 4-point scale and summed for a total score. The assessor also uses this tool to classify risk as low, moderate, or high. Validation research has replicated the predictive effect of the DVSI-R for family violence and other domestic incidents, and the score has yielded a significantly larger effect size than the professional judgment.
A Different Approach to Domestic Violence Risk
There is a growing capacity for criminal justice systems to develop risk assessment procedures based on their offender processing and tracking databases. Richard Berk and colleagues reported in 2016 that data from over 28,000 cases involving arrest for domestic violence were used to generate an algorithm to forecast those who would have no further detected domestic violence incidents in a 2-year follow-up. Most offenders met the criterion for nonrecidivism, and they could be identified with nearly 90% accuracy. The tool is used to classify, at the time of arraignment, offenders who can be swiftly processed for bail in order to allow more resources to be allocated to further assessment or intervention for offenders who pose a risk of recidivism. This approach permits the use of statistical modeling techniques more robust than those typically used to develop actuarial risk assessment tools, and the big data set also means that the process can be continually refined. Because it is constructed using data from the intended population and used to inform decisions in that same population, this approach is a good example of implementing the Risk-Need-Responsivity risk principle.
Cross-Validation and Generalization of Domestic Violence Risk Assessment
Empirically derived risk assessment instruments are sometimes criticized for the expected loss of predictive accuracy between construction and cross-validation. Nonalgorithmic assessments have been opposed for their inclination toward unstructured clinical opinion. For any risk assessment approach, there is a need to evaluate and replicate the extent to which a measure is related to the outcome of interest. Research has been conducted to demonstrate the generalizability of most of the major published domestic violence risk assessment instruments for male offenders. Research with some of the tools described in this article has shown that they are associated with police decision-making, but field-testing of domestic violence risk assessment remains in its infancy.
Limited research, in contrast, has been conducted with female offenders. One study of police using the B-SAFER found that fewer than 10% of women were judged to be high risk. Research with women in same-sex relationships led to a revision of the Danger Assessment to improve predictive accuracy for this group. A small study of the ODARA and a large study of the DVSI-R found that these instruments predicted women’s domestic violence recidivism, but their overall reoffense rates were lower than men’s.
Risk assessment for victims has been particularly concerned with their risk of homicide by an intimate partner. Both the Danger Assessment and ODARA have been reported to yield relatively high scores in domestic homicide cases. However, statistical modeling of homicide risk is impeded by the low base rate of such cases. Cooperation among sectors responding to domestic violence, and the sharing of information from validated domestic violence risk assessment, has been recommended as the most promising approach.
References:
- Helmus, L., & Bourgon, G. (2011). Taking stock of 15 years of research on the Spousal Assault Risk Assessment Guide (SARA): A critical review. International Journal of Forensic Mental Health, 10, 64–75. doi:10.1080/14999013.2010.551709
- Hilton, N. Z., & Eke, A. W. (2017). Assessing risk of intimate partner violence. In J. C. Campbell (Ed.), Assessing dangerousness: Violence by batterers and child abusers (3rd ed., pp. 139–178). New York, NY: Springer.
- Williams, K. R., & Stansfield, R. (2017). Disentangling the risk assessment and intimate partner violence relation: Estimating mediating and moderating effects. Law and Human Behavior, 41, 344–353. doi:10.1037/ lhb0000249