The Development of the Brief Spousal Assault Form for the Evaluation of Risk (B-SAFER): A Tool for Criminal Justice Professionals

3. Overview of Risk Assessment

3. Development of the B-SAFER

As discussed in the previous section, structured professional judgment appears to be a viable approach to assessing risk for spousal violence. It also appears to be the method that is best suited to the requirements of criminal justice professionals. Principles of natural justice, as well as those enshrined in Canadian constitutional, statutory, and common law, place a heavy burden on people who make decisions that affect the life, liberty, and security of citizens. On the one hand, these decisions must not be arbitrary or discriminatory; the rationale underlying them must be clear, well reasoned, and reasonable. The use of a checklist or some other tool to enhance the transparency and consistency of decisions is one way to achieve this goal. On the other hand, the decision-making process must allow for some flexibility to reflect the uniqueness and totality of circumstances in the case at hand. The Supreme Court of Canada, in considering a wide range of cases related to violence and violence risk over many decades, has consistently held that the application of discretion by criminal justice and mental health professionals (e.g., police and corrections officers, prosecutors and judges, parole and review boards, psychiatrists and psychologists) is both necessary and appropriate [2].

The Spousal Assault Risk Assessment Guide (SARA), a set of structured professional guidelines for assessing risk of spousal violence, has been used for many years by criminal justice professionals, including police. It comprises 20 risk factors that reflect various aspects of criminal history, social functioning, and mental health. The risk factors were selected based on a comprehensive review of the professional and scientific literatures. Evaluators consider the presence and relevance of individual risk factors, and also make summary judgments of risk. However, the SARA may not be an optimal tool for use by police because it is relatively long and it requires specific judgments regarding mental health, such as major mental illness and personality disorder. Thus, completion of the SARA places a relatively heavy burden on users in terms of the availability of time, technical expertise, and case history information. We therefore saw a need to develop a new tool, which we called the Brief Spousal Assault Form for the Evaluation of Risk, or B-SAFER. In this section, we outline the steps taken in the development of the B-SAFER and describe the tool itself.

3.1 Literature Review

Our first step in developing the B-SAFER was to conduct a comprehensive review of the literature regarding spousal violence and spousal violence risk assessment. We also updated this review continuously during the project to keep abreast of new developments in the field.

Overall, the literature review indicated that there have been relatively few advances in our understanding of risk factors for spousal assault since the development of the SARA in the early 1990s. There has been further research supporting the utility of some risk factors previously identified (for example, see reviews by Dutton & Kropp, 2000; Riggs, Caulfield, & Street, 2000; Schumacher, Feldbau-Kohn, Slep, & Heyman, 2001), but no important new risk factors have been identified.

The literature review also suggested that there have been few advances in the development of specific tools or procedures for spousal violence risk assessment. One exception was the Ontario Domestic Assault Risk Asessment (ODARA), a tool developed for use by the Ontario Provincial Police. As the ODARA is based on the actuarial approach, it is intended to estimate the likelihood of future violence rather than to provide information about risk management. This means that professionals who use the ODARA still need assistance making final decisions that reflect the totality of circumstances in the case at hand and that guide case management.

Another development was an increased focus on victims. Both the ODARA and the Stalking Assessment and Management Guide (SAM), a structured professional judgement tool currently being developed by the BCIFV, include consideration of factors that increase a victim's vulnerability to violence. One potential problem with this advance is that including victim vulnerability factors in a new tool increases the complexity (i.e., length and scope) of the assessment.

In sum, the literature review indicated to us that it would be possible to use the SARA as a basis or starting point for the development of the B-SAFER. It also indicated that the B-SAFER might benefit from consideration of victim vulnerability factors, providing their inclusion did not make the use of the tool unduly complex or resource-intensive.

3.2 Empirical Analyses

Statistical Analysis of SARA Ratings

In October 2002, we asked colleagues in Scotland to conduct statistical analyses of existing data sets to identify possible redundancy among the 20 SARA risk factors. The data sets comprised 2,796 adult male offenders from Canada: 1,786 were offenders on probation in British Columbia, and 1,010 were offenders from federal penitentiaries. The probationers were serving sentences for offences related to spousal assault, whereas the federal offenders were serving sentences for a variety of offences but had a known, documented, or suspected history of spousal assault.

Briefly, Exploratory and Confirmatory Factor Analyses suggested that the statistical association among the ratings of the 20 SARA items could be modeled adequately using 7 factors, with each factor comprising of multiple items. The factors were as follows:

  1. History of spousal violence;
  2. Life-threatening spousal violence;
  3. Escalation of spousal violence;
  4. Attitudes supportive of spousal violence;
  5. General antisocial behaviour;
  6. Failure to obey court orders; and
  7. Mental disorder.

The factors themselves appeared to be non-redundant. Most of the factors had unique predictive power with respect to global judgments of risk for spousal violence or, in a small sub-sample of 102 offenders, with respect to actual spousal violence recidivism. We also used Item Response Theory, which is a framework for analyzing psychological tests, to analyse the SARA data. In our analyses of the B-SAFER, we used Item Response Theory to model the association between spousal assault risk and individual risk factors, as well as the redundancy among various risk factors. The findings regarding redundancy were similar to those yielding from the factor analyses.

Pilot Testing of the 20-item SARA-PV in Sweden

We then pilot tested a modified version of the SARA, which we called the SARA - Police Version (SARA-PV), with the Swedish National Police between January and December 2002. In the SARA-PV, each of the 20 SARA risk factors was revised and shortened to simplify coding decisions; however, the SARA-PV was still identical to the SARA in terms of the number of risk factors and the general coding procedures. Patrol officers attended 1-day training sessions conducted by one of the authors and then used the SARA-PV when responding to spousal violence incidents. Patrol officers reviewed the completed SARA-PV coding forms with shift supervisors prior to making case management decisions. More specifically, shift supervisors ensured that each risk assessment was thoroughly completed and based on adequate and appropriate information, and that management recommendations were logically linked to the nature and severity of the risks posed.

In total, we received completed SARA-PV coding forms for 430 adult males being charged with perpetrating spousal violence. Analysis of the SARA-PV ratings indicated that it was sometimes difficult for patrol officers to gather the information required to rate some risk factors as part of their usual investigation procedures. In particular, they found it difficult to make specific judgement about the perpetrator's mental disorder and about his history of childhood victimization experiences. In addition, feedback received from police officers revealed two major concerns regarding the use of the SARA-PV. First, they wanted the scheme used to code the presence of individual risk factors to more closely resemble their usual operational procedures and language. Second, they expressed a desire for clarified and simplified coding of overall or summary judgments regarding risk.

In summary, the results of these Analyses empiriques indicated the following:

  1. Some SARA and SARA-PV items may have redundant or overlapping content;
  2. Some SARA and SARA-PV items may be difficult to code when used by police as part of routine investigations, due to specificity of content;
  3. The schemes used in the SARA and SARA-PV to code judgments regarding the presence of individual risk factors and overall risk may not be a good fit for use by law-enforcement.

Overall, these findings were consistent with our anecdotal observations and with informal feedback received when conducting SARA training with police in the past. The findings also suggested that it was both necessary and feasible to shorten, simplify, and revise the SARA and the SARA-PV for use by police.

3.3 Format of the Draft B-SAFER and Related Materials

Based on the statistical analyses of the SARA and the Projets pilotes of the SARA-PV, the draft of the B-SAFER that we developed for Projets pilotes in Canada (and Sweden) comprised 10 risk factors. These risk factors were divided into two sections. The first section, Spousal Assault, contained 5 factors related to the perpetrator's history of spousal violence:

  • 1. Serious physical/sexual violence;
  • 2. Serious violent threats, ideation, or intent;
  • 3. Escalation of physical/sexual violence or threats/ideations/intent;
  • 4. Violations of criminal or civil court orders; and
  • 5. Negative attitudes about spousal assault.

The second section, Psychosocial Adjustment, contained 5 factors related to the perpetrator's history of psychological and social functioning:

  • 6.Other serious criminality;
  • 7. Relationship problems;
  • 8. Employment and/or financial problems;
  • 9. Substance abuse; and
  • 10. Mental disorder.

The risk factors in the latter section are associated with risk for violence in general, in addition to risk for spousal violence.

B-SAFER Worksheet

A B-SAFER Worksheet is included in Annexe A of this report. Some changes were made to the draft form in the final revisions. First, the language was changed to reduce professional jargon and thereby make the B-SAFER easier to read and apply. For example, the title of Item 4 on the B-SAFER was changed from "Violations of Conditional Release" to "Violations of Court Orders" to make the intent of the item more clear. As well, we attempted to use plain language to describe aspects of mental disorder listed as descriptors in Item 10. Second, a section on recommended management strategies was added to facilitate development and documentation of case management plans. Finally, after considering the risk factors and management strategies, the worksheet requires the evaluator to provide judgements of case prioritization, risk for life threatening violence, risk for imminent violence, and the likely victims of violence.

User Manual

The User Manual for the B-SAFER can be found in Annexe B. The Manual includes an overview of the B-SAFER, as well as sections on user qualifications, confidentiality and informed consent, applications, and administration procedures. We have also included a comprehensive section entitled, "Definition of Risk Factors" which includes item definitions, rationales for including items (including Références to supporting literature), specific coding instructions for each B-SAFER item, and a detailed reference list. We also added considerable information regarding the development of case management plans. Finally, the manual includes a semi-structured interview guide for victims.

Victim Interview

The semi-structured interview guide for victims is included in Annexe C. We developed the interview and circulated it among a small number of police officers and victim service workers for feedback. It includes suggested questions that can be asked for each risk factor. The format is semi-structured to allow flexibility and discretion for interviewers.

[2] See Canadian Foundation for Children, Youth and the Law v. Canada (Attorney General), [2004] 1 S.C.R. 76; Penetanguishene Mental Health Centre v. Ontario (Attorney General), [2004] 1 S.C.R. 498; R. v. Johnson, [2003] 2 S.C.R. 357; Smith v. Jones, [1999] 1 S.C.R. 455.

Date modified: