You are on page 1of 11

TRAUMA,

Dutton, VIOLENCE,
Kropp & ABUSE / April 2000
/ RISK INSTRUMENTS

A REVIEW OF DOMESTIC VIOLENCE RISK INSTRUMENTS

DONALD G. DUTTON
University of British Columbia

P. RANDALL KROPP
British Columbia Forensic Psychiatric Services Commission

The problem of domestic violence has been well documented with respect to its so-
cial, psychological, and economic costs. Proactive arrest and sentencing policies
have resulted in an increasing, and in some cases, overwhelming number of
spousal batterers being processed through the criminal justice system. Scarce cor-
rectional and treatment resources necessitate that difficult decisions be made about
the management of domestic violence perpetrators. In an effort to make better deci-
sions, many jurisdictions have adopted a risk assessment approach. Spousal as-
sault risk assessment now serves to inform those making decisions about
sentencing (e.g., community release vs. incarceration), treatment placement, and
supervision intensity. With these developments, researchers and clinicians have
begun to discuss the appropriate content and process of spousal assault risk assess-
ment. There have been a number of efforts in recent years to develop theoretically
and scientifically sound risk assessment instruments and procedures. This article
attempts to review state-of-the-art instruments in this rapidly expanding field.

Key words: spouse assault, domestic violence, risk prediction

THERE HAS BEEN A LONG-EXISTING DE- impoverished predictor variables, weak crite-
BATE about the validity of violence prediction rion variables, and constricted validation sam-
(Cocozza & Steadman, 1976; Grove & Meehl, ples. These authors note that if such methodo-
1996; Monahan, 1981; Quinsey, Harris, Rice, & logical problems are addressed, “the rejuve-
Cormier, 1998). Much of the controversy has fo- nated field of risk assessment may yield results
cused on the distinction between actuarial and quite different than those to which we have
clinical prediction, with many authors express- become inured” (Monahan & Steadman, 1994,
ing pessimism about the ability of clinicians to p. 13). Borum (1996) observed that, despite the
predict violence accurately (Cocozza & Stead- debate about the ability of mental health profes-
man, 1976; Grove & Meehl, 1996; Quinsey et al., sionals to predict violence, the courts have con-
1998). However, recent reviews have taken a tinued to rely on such assessments. He de-
more optimistic tone. Monahan and Steadman scribed several advances in the field that have
(1994) discussed problems that have hobbled re- provided a foundation for enhancing the valid-
search on the prediction of violence, including ity of risk assessment. Although Melton, Petrila,

TRAUMA, VIOLENCE, & ABUSE, Vol. 1, No. 2, April 2000 171-181


© 2000 Sage Publications, Inc.

171
172 TRAUMA, VIOLENCE, & ABUSE / April 2000

KEY POINTS OF THE lighted the traditional limitation of weak crite-


RESEARCH REVIEW rion variables that are either too broadly defined
• History of violence prediction shows gradual im- (e.g., all violent behavior) or unreliably mea-
provements with time. sured. They noted that there might be unique
• Relatively high baserates for spousal assault sets of predictors for various subtypes of vio-
makes prediction more accurate. lence. Indeed, research over the past 20 years
• Predicting spousal homicide, a much more rare has identified several risk factors that are
event, is not nearly as accurate. uniquely related to spousal violence.
• The Danger Assessment Scale, the Spousal As-
sault Risk Scale, the propensity for Abusiveness
Scale, and the Psychopathy Checklist–Revised all Risk Factors for Spousal Assault
predict risk for future spouse assault with some
accuracy. There have been a number of studies that
have attempted to identify factors associated
with spousal violence (Cooper, 1993). Many
studies have reported risk factors that discrimi-
Poythress, and Slobogin (1997) caution against nated those who were violent toward their
using conclusive terms about dangerousness, spouses from those who were not (e.g., Hotaling
they suggest that if clinicians express their judg- & Sugarman, 1986; Tolman & Bennett, 1990).
ments in relative terms (e.g., relatively high risk), Other studies highlight factors associated with
with reference to empirically valid risk factors, risk for recidivistic violence among known
risk assessment can play a valuable role for the spousal assaulters—those arrested, convicted,
courts. Many have suggested that the renewed or in treatment (e.g., Gondolf, 1988; Saunders,
optimism is due to improvements in research 1992a, 1992b, 1993). Many of these studies inter-
methodology employing more comprehensive sect with those discussing risk for violence in
assessment techniques (S. D. Hart, 1998) and general, with many of the same factors emerg-
better follow-up procedures to measure vio- ing (e.g., Hall, 1987; Monahan, 1981; Monahan &
lence in the community (Borum, 1996; Douglas, Steadman, 1994; Webster, Harris, Rice, Cormier, &
Cox, & Webster, 1999; Steadman et al., in press). Quinsey, 1994). There are also several important
works that have discussed the assessment of
Assessing Future Probability risk for future violence in spousal assaulters,
of Spousal Assault sometimes described as a lethality or a need-to-
warn assessment (e.g., Goldsmith, 1990; Saun-
It is within this atmosphere of rejuvenation, ders, 1992b, 1995; Sonkin, 1987; Sonkin, Martin, &
as Monahan and Steadman (1994) described it, Walker, 1985). Finally, there have been some
that efforts to design spousal assault risk instru- extremely useful studies on factors associated
ments have arisen. When the task of assessing with the more specific act of domestic homicide
risk for general violence is narrowed specifi- (Campbell, 1995, 1999).
cally to spousal assault, there are additional rea- There is considerable consensus among these
sons to be optimistic about achieving success. studies with regard to the important factors to
First, the base rates for repeated spousal physi- consider when assessing risk for spousal
cal assaults are relatively high, probably 25% to assault. Many risk assessment lists have been
50% (Dutton, 1995a; Dutton, Ogloff, Hart, Bod- published. Most of these include factors related
narchuk, & Kropp, 1997). This serves to reduce to a history of assaultive behavior, generally
the false-positive error rate or the rate of falsely antisocial behaviors and attitudes, stability of
predicting that violence will occur. Second, in relationships, stability of employment, mental
the course of a spousal assault risk assessment, health and personality disorder, childhood
the evaluator often has access to the potential abuse, motivation for treatment, and attitudes
victim, who is a source of critical information toward women (B. Hart, 1990; Pence & Lizdas,
about the history and personality of the perpe- 1998; Roehl & Guertin, 1998; Sonkin, 1987; Sug-
trator. Third, Monahan and Steadman high- arman, Aldorondo & Boney-McCoy, 1995; Sug-
Dutton, Kropp / RISK INSTRUMENTS 173

TABLE 1: List of Criteria for Identifying Life-Threatening toward you by your partner” (for a critique, see
Risk Among Violent Men
Dutton, 1995a).
Three or more instances of violence in the previous year The second example of a risk factor list is one
Two or more instances of violence in the previous year initiated developed by Daniel Sonkin (1997) in The Perpe-
by the man
trator Assessment Handbook, intended for use by
Wife needed medical treatment for assault
Police called during previous 12 months treatment providers. As
Man was drunk more than three times a year part of this guide, Sonkin There is considerable
Man abused drugs in past year outlines factors that dis- consensus among
Man threatened to kill
Man threatened his partner with a weapon in his hand criminate lethality risk, these studies with
Man owns a gun and threatens to use it including the man’s fre- regard to the
Extreme male dominance or attempts to use it quency of violence, sever- important factors to
Physical abuse of a child
ity of violence, frequency consider when
Attitudinal acceptance of violence
Physically forced sex of intoxication, drug use, assessing risk for
Extensive destruction of property
a
threats to kill, forced or spousal assault. Many
Threats or actually killing a pet risk assessment lists
threatened sexual acts,
History of psychological problems
Assault on a nonfamily person or other violent crime and the woman’s suicide have been published.
Severe violence between parents threats. However, Sonk- Most of these include
Verbal aggression toward partner in’s factors were derived factors related to a
SOURCE: Straus (1991). on the basis of factors that history of assaultive
a. Cruelty to pets should probably also be considered. behavior, generally
g e n e ra te d le th a l
responses by battered antisocial behaviors
arman & Hotaling, 1986). It is important to women in the Browne and attitudes, stability
remember that these risk markers are not neces- (1987) study. Hence, it of relationships,
sarily causal predictors; rather, they are factors a g a in re qu ire s th e stability of
that consistently co-occur with abusiveness. a s s u mp tion th a t th e employment, mental
woman’s fear, which led health and
Two examples of risk marker lists are offered
her to kill, was a veridical personality disorder,
here. Straus (1991) developed a “list of criteria
predictor of impending childhood abuse,
for identifying life threatening risk among vio-
lethal violence by her motivation for
lent men” from the 1985 National Family Vio-
male partner. As compel- treatment, and
lence Survey (see Table 1). These criteria were attitudes toward
ling as Browne’s case
associated with severe violence as measured by women.
studies are, we do not
the Conflict Tactics Scale (CTS) (Straus & Gelles,
know whether this ulti-
1990) and thus have some concurrent construct mate outcome was always inevitable. In a recent
validity from one sample. The problem of study, Dutton and Kerry (1999) found that
course is that Straus is calling these “life threat- spousal homicide perpetrators were more likely
ening risk criteria” when, in fact, they are crite- to be overcontrolled dependent men than psy-
ria for severe assault. Most severe assaults are chopathic or borderline men. These latter
not lethal; hence, the term life threatening is groups with more flagrant personalities may be
somewhat misleading. more likely to terrorize but not necessarily to
Straus (1991) offers no weighting formula for kill.
this lengthy list of factors, which, on closer Sonkin (1997) rightly alerts his readers to two
inspection, is a list of extreme scores from the points about the clinical assessment of risk: The
CTS used to predict another extreme score on chances of being correct are about 50-50, but the
the CTS. This form of tautological prediction is best predictor of future behavior is past behav-
common in family violence literature, especially ior. He thus advocates a comprehensive assess-
in the literature on dating violence, in which ment of both psychological and physical abu-
predictors of dating violence are listed as “pre- siveness. Sonkin emphasizes that his checklist is
vious dating violence” or “violence directed a guide, not a psychometric test. In effect, it
174 TRAUMA, VIOLENCE, & ABUSE / April 2000

TABLE 2: Domestic Violence Risk Assessment: Man’s


combines the frequency and severity of physical Frequency of Violence
and sexual violence, threats, intoxication, drug
use, Diagnostic and Statistical Manual of Mental Man’s severity of violence
Disorders (American Psychiatric Association, Man’s frequency of intoxication
Man’s drug use
1994) diagnosis, severity of current psychoso- Man’s threats to kill
cial stressors, global assessment of functioning, Man’s forced or threatened sexual acts
proximity of victim and offender, prior criminal Woman’s suicide threats
convictions of offender, prior criminal behav- SOURCE: Sonkin (1997).
iors and abuse of children, and attitudes toward
violence (see Table 2). Of all the risk assessment
inventories reviewed, it was probably the Houghton, 1999), the Kingston Screening
lengthiest and most comprehensive. Instrument for Domestic Violence (K-SID)
(Gelles, 1998), the Navy Risk and Safety (as cited
by Campbell, 1999), and the Pre-Sentence Inves-
RISK ASSESSMENT INSTRUMENTS
tigation (PSI) Domestic Violence Supplement
Until recently, there associated with the Domestic Abuse Interven-
Until recently, there were very few efforts to tion Project in Duluth, Minnesota (Roehl &
were very few efforts empirically validate any Guertin, 1998). We had difficulty locating many
to empirically of the risk assessment of the instruments in the first two categories, so
validate any of the schemes. This state of it is difficult to provide a detailed analysis. The
risk assessment affairs is changing rap- interested reader is referred to the Roehl and
schemes. This state of idly, with several evalua- Guertin (1998) review for more information on
affairs is changing tion studies underway. some of these projects.
rapidly, with several Roehl and Guertin (1998) The remainder of this article will focus on a
evaluation studies and Campbell (1999) third category of instruments selected because
underway. have conducted useful we were able to locate published validity data.
surveys of dangerous- We have focused on four such instruments: the
ness assessment instru- Danger Assessment (DA) Scale (Campbell,
ments currently in use in North America. It 1995, 1998, 1999), the Spousal Assault Risk
appears that there has been a proliferation of Assessment (SARA) (Kropp, Hart, Webster, &
instruments and scales in the past several years, Eaves, 1995, 1998), the Propensity for Abusive-
ranging in their degree and scope of develop- ness Scale (PAS) (Dutton, 1995b, 1999), and the
ment. To the best of our knowledge, there Psychopathy Checklist–Revised (PCL-R) (Hare,
appear to be three general categories of instru- 1991).
ments. The first includes instruments for which
the developers report no past or current The DA Scale
attempts to establish validity. These include the
Domestic Violence Risk Assessment Form used Campbell (1995) developed the DA Scale,
in Delaware, the Mosaic-20 developed by Gavin which she described as a “form of statistical pre-
de Becker (Trone, 1999), and the Risk Assess- diction, contrasted with clinical prediction,
ment and Lethality Assessment used in Hawaii because it is based on prior research and has
(Roehl & Guertin, 1998). The second category of some preliminary evidence of reliability and
instruments contains those that are currently validity” (p. 103). Campbell’s criterion was
included in evaluation and validity studies but “women’s perception of the danger of being
have not yet been reported in the scientific lit- killed by their partners.” However, the relation-
erature. These include the Domestic Violence ship of the fear of their partner to the actual dan-
Inventory (DVI) (as cited in Roehl & Guertin, ger is unknown. Furthermore, Campbell’s
1998), the Domestic Violence Screening Instru- validity data appear not to be measures of part-
ment (DVSI) used in Colorado (Williams & ner violence taken subsequent to the interview
Dutton, Kropp / RISK INSTRUMENTS 175

with the female respondent but measures of TABLE 3: Areas Covered by the Spousal Assault Risk
Assessment
prior severity of partner violence. Hence, her
study is retrospective and presents some inter- Childhood abuse and neglect experiences
pretative problems. Goodman, Dutton, and Occupational and social history
Relationship history
Bennett (2000) have completed a study using Physical and mental health history
the DA to predict short-term misdemeanor wife Current mental status
assaults. An increment of one standard devia- History of assaultive or abusive behavior
Criminal history
tion on the DA was related to a fourfold increase
Current life stressors
in the likelihood of repeat abuse (22% of the Current social support network
entire sample reoffended). Some methodologi-
cal problems were pointed out by the authors.
The sample was small (n = 92), and only 53% review of the empirical literature on wife
could be recontacted. This smaller group (n = assault and the literature written by clinicians
47) may, of course, not be representative of the that evaluates male wife abusers. The authors
original sample. The authors called for replica- point out that the SARA is not a test. Its purpose
tion with a larger sample and more detailed out- is not to provide absolute or relative measures
come measures. of risk using cutoff scores or norms but rather to
Weisz, Tolman, and Saunders (2000) studied structure and enhance professional judgments
the victims’ prediction of future violence as a about risk. Because the SARA is not a formal
legitimate predictor. Data were supplied by 177 psychological test, professionals other than psy-
women partnered with men in court-ordered chologists can use it. The authors list several
batterers’ treatment. The authors found that vic- potential applications of the SARA: pretrial
tims who predicted no violence or a strong like- assessment, presentence, correctional intake,
lihood of subsequent violence were likely to be correctional discharge, civil justice matters,
correct for a 4-month period following the pre- warning third parties, and as an instrument to
diction. A bivariate analysis of women’s reports review spousal risk assessments given by
found that the following items were also signifi- others.
cantly related to recidivism: kicked, bit or hit The item selection for the SARA was carefully
with a fist (from the CTS), forced sex, choked or based on relevant factors reported in the litera-
strangled, woman obtained protection order ture. The SARA procedure includes interviews
before man’s first arrest, woman sustained inju- with the accused and the victims, standardized
ries before first arrest, violent disputes between measures of physical and emotional abuse and
first arrest and court date, man threatens her to drug and alcohol abuse, and a review of collat-
drop charges, man tells her she cannot leave or eral records such as police reports, victim state-
see certain people, man restricts her use of ments, criminal records, and other psychologi-
phone or car, man accuses her of an affair. The cal procedures. If the user is not a psychologist,
authors included the DA in their prediction the authors recommend appropriate referrals or
package. A multiple regression comprising the a review of existing psychological or psychiatric
DA and the above items yielded an R2 of .22. In reports. The authors recommend that any risk
other words, the combined instruments assessment interview should cover the areas
accounted for 22% of the variance in repeat listed in Table 3.
violence. The Manual for the Spousal Assault Risk Assess-
ment Guide (Kropp et al., 1995) describes the
The SARA rationale and scoring system for each of these
areas. The items can be scored 0 (absent), 1
The SARA (Kropp et al., 1995; Kropp, Hart, (subthreshold), or 2 (present). The SARA has an
Webster, et al., 1998) is a set of guidelines for the override and makes use of critical items, which
content and process of a thorough risk assess- are defined as follows: “those, given the circum-
ment. It is composed of 20 items identified by a stances of the case at hand, are sufficient on their
176 TRAUMA, VIOLENCE, & ABUSE / April 2000

own to compel the evaluator to conclude that The PAS


the individual poses an imminent risk of harm.”
The manual provides case management strate- The PAS (Dutton, 1995b) was empirically
gies appropriate to various risk factors. developed by assessing the items in a compre-
The authors have evaluated the reliability hensive assessment given to convicted batterers
and validity of judgments concerning the risk and nonviolent controls (n = 206) that best pre-
for violence made using the SARA (Kropp & dicted their female partners’ reports of abusive-
Hart, in press; Kropp, Hart, Webster, et al., 1998; ness. It is a self-report scale that, at face value,
Kropp, Hart, Whittemore, & Bodnarchuk, appears unrelated to abusiveness because it
1998). The SARA ratings were analyzed in six taps into background factors such as parental
samples of adult male offenders (N = 2,681). The treatment, attachment style, anger response,
distribution of ratings indicated that offenders trauma symptoms, and stability of self-concept.
were quite heterogeneous with respect to the The scale has focused primarily on emotional
presence of individual risk factors and to the abuse and can correctly discriminate abusive
overall perceived risk. Structural analyses of the men with 82.2% accuracy. Social desirability fac-
risk factors indicated moderate levels of inter- tors do not appreciably alter its relationship to
nal consistency and item homogeneity. Interra- women’s scores of abusiveness.
ter reliability was high for judgments concern- The PAS has been cross-validated on men in
ing the presence of individual risk factors and community groups (n = 500+), and its predictive
for the overall perceived risk. The SARA ratings accuracy remained at the same level as with the
significantly discriminated between offenders original sample. Although the PAS is a good
with and without a history of spousal violence predictor of abusiveness, its main research focus
in one sample and between recidivistic and non- has been on emotional abuse. The PAS corre-
recidivistic spousal assaulters in another sam- lated .51 with victims’ reports of dominance and
ple. Finally, the SARAratings showed good con- isolation by the perpetrator and .37 with emo-
vergent and discriminant validity with respect tional abuse. Using the Severity of Violence
to other measures related to the risk for general Against Women Scale (Marshall, 1992), the PAS
and violent criminality (Kropp & Hart, in press). correlated .76 with use of threats toward the vic-
Another large-scale study involving the tim and .30 with physical violence. All correla-
SARA is underway in Colorado. The project is tions were statistically significant. The PAS cor-
jointly supported by the Department of Justice, related significantly with men’s anger and
Violence Against Women Grants Office, the general dysphoria ratings when listening to
Colorado State Court Administrator’s Office, audiotaped abandonment scenarios (Dutton &
and the Center for the Study and Prevention of Parkin, 1999). It should be noted that the PAS is
Violence at the University of Colorado. The probably more suitable for noncriminal popula-
Colorado group has developed and imple- tions in which the main objective is to predict
mented a process for the evaluation of domestic emotional abuse and the risk for physical abuse.
violence offenders. The process uses two instru- Instruments such as the SARA, on the other
ments, the DVSI and the SARA. The DVSI is a hand, include factors related to prior assaultive
structured criminal history screening com- behavior and may therefore be better suited for
pleted before an offender’s first appearance in assessment in a different population than the
court. The SARA is completed as a presentence PAS (see Figure 1).
evaluation. Data have not yet been published,
but the project will test the reliability and valid- The Psychopathy Checklist
ity of the instruments standing alone and in
combination. Preliminary results indicate that Psychopathy is a personality disorder charac-
the SARA is superior to the DVSI in predicting terized by deficits in empathy and persistent,
future assaults, but both instruments predict frequent, and varied antisocial behavior start-
reoffense above chance levels (Williams & ing at an early age (S. D. Hart & Hare, 1997). Psy-
Houghton, 1999). chopathy is assessed using the Psychopathy
Dutton, Kropp / RISK INSTRUMENTS 177

In a meta-analytic review of the prediction of


recidivism, Hemphill, Templeman, Wong, and
Hare (1998) compared the predictive ability of
the PCL with other predictor variables. They
conclude that “the PCL and PCL-R by them-
selves are significant predictors of re-offending.
However, an important issue is their ability to
add to the predictive validity of other more tra-
ditional variables” (p. 381). The PCL, in other
words, contributed unique information beyond
that offered by demographic and criminal his-
tory variables.1
The PCL and PCL-R are more strongly related
Figure 1: PAS in a Demographically Controlled Sample to recidivism than any personality diagnosis
NOTE: PAS = Propensity for Abusiveness Scale. EMBU = Enga including the antisocial personality. Finally, the
Minnen Betraffande Uppfostran (Recollections of Early Childhood
Rearing). BPO = Borderline Personality Organization. MAI = Multi- PCL and PCL-R have
dimensional Anger Scale. PMWI = Psychological Maltreatment of shown incremental pre- Psychopathy is a
Women Inventory. TSC = Trauma Symptoms Inventory. RSQ = Re-
lationship Style Questionnaire. N = 45.
dictive validity in com- personality disorder
**r < .40, p < .01. ***r < .44, p < .001. parison to the following characterized by
actuarial risk scales: Base deficits in empathy
Checklist (PCL) (Hare, 1980), which was later Expectancy Score (Gott- and persistent,
revised to the 20-item PCL-R (Hare, 1991) fredson & Bonds, 1961), frequent, and varied
designed for male forensic populations and the Salient Factor Score antisocial behavior
eventually to the 12-item Psychopathy Check- (Hoffman, 1983), the Gen- starting at an early
list: Screening Version (PCL: SV). It is included eral Statistical Informa- age.
in this discussion because it has been demon- tion on Recidivism Scale
strated to be a robust predictor of violent behav- (G-SIR) (Nuffield, 1982),
ior in general, with many of its validity studies and the Violence Risk Appraisal Guide (V-RAG)
including domestic violence perpetrators (S. D. (Harris, Rice, & Quinsey, 1993).
Hart & Hare, 1997; Quinsey et al., 1998). The Dunford (in press) assessed 850 men who
PCL-R is a structured interview and set of rat- completed their treatment for wife assault in
ings based on the interview and corroboration San Diego. A random sample of treatment suc-
based on case history reviews, institutional files, cesses (no violence reported by the wife) and
interviews with family members and employ- treatment failures revealed that unsuccessful
ers, and criminal and psychiatric records (see men were more than six times as likely to have
Table 4). The PCL-R has generated two factors PCL: SV scores above 12 (the scale ranges from 0
labeled affective and social deviance. The first has to 24).
to do with glibness, lack of empathy, and patho- Kropp, Hart, Whittemore, Webster, and Eaves
logical lying. The second has to do with antiso- (1998) included the PCL: SV in a study following
cial behavior. A newer version of the scale, the back 50 spousal assault recidivists and 50 nonre-
PCL: SV, has been developed for use with both cidivists. Raters coded the SARA and the PCL: SV
criminal and noncriminal populations. blind to the outcome status. Although the SARA
Psychopaths have criminal careers that start discriminated between the two groups, the PCL:
early (in their teens) and generate high rates of SV did not. This is an intriguing finding given
violent offending until at least their mid-40s. how effective the PCL has proven to be when pre-
They are the most likely group to engage in vio- dicting violence in general. In some ways, how-
lence while incarcerated, to recidivate after ever, we do not expect the psychopath to be the
release, and to constitute treatment failure from most likely type of spousal assaulter to recidivate.
prison programs. Fundamental to the psychopath’s personality is
178 TRAUMA, VIOLENCE, & ABUSE / April 2000

TABLE 4: Items in the Psychopathy Checklist–Revised

Factor 1: Interpersonal of Affective Factor 2: Social Deviance Additional Itemsa

Glibness or superficial charm Need for stimulation or proneness to boredom Promiscuous sexual behavior
Grandiose sense of self-worth Parasitic lifestyle Many short-term marital relationships
Pathological lying Poor behavioral controls Criminal versatility
Conning or manipulative Early behavioral problems
Lack of remorse or guilt Lack of realistic long-term goals
Shallow affect Impulsivity
Failure to accept responsibility Irresponsibility
for own actions Juvenile delinquency
Revocation of conditional release
SOURCE: Hare (1991).
NOTE: The rater uses specific criteria, interview, and file information to score each item on a 3-point scale ranging from 0 to 2.
a. These are items that do not load on either factor.

the tendency to not form attachments to other attempts to predict domestic violence based on
people: He may roam from partner to partner. these factors.
Psychopaths do not form the pathological cling- Empirically validated risk assessment instru-
ing attachments that some other batterers ments should constitute one part of the constel-
exhibit (e.g., the dysphoric or borderline bat- lation of decision factors used to make determi-
terer, see Dutton, 1995a; Holtzworth-Monroe & nations of risk. Quality risk assessment,
Stuart, 1994). Regardless, the Kropp, Hart, instruments that are actuarially based, inter-
Whittemore, Webster, and Eaves data suggest nally and externally valid and reliable, longitu-
that the PCL may not be the best instrument to dinally verified, peer reviewed, and based on
use, at least not on its own, when assessing the quality research should be an important part,
risk for repeated spousal violence. but not all, of the assessment process. The qual-
ity of the information on which the assessment
is based is also extremely critical. Assessments
CONCLUSIONS
should use multiple methods and multiple
Clinicians, correctional personnel, police, sources—including the victim(s)—for data col-
and victim service workers have for many years lection. Although validation studies are needed,
been asked to make judgments about risk and longitudinal outcome research is difficult due to
dangerousness in spousal assaulters. Until practical and ethical considerations. For exam-
recently, however, there were few guidelines or ple, to truly know if our predictions for violence
tools to assist them in this task. This has are accurate, we would need to release high-risk
changed with the proliferation of spousal offenders to the community, without interven-
assault risk instruments in production. Unfor- tion or restraints, to determine if they reoffend
tunately, there is still relatively little published (i.e., that they are true-positive predictions).
research on the reliability and validity of these This strategy is ill advised, therefore any predic-
tools. This situation is likely to improve if the tive study in this area must attempt to control
number of in-progress validation studies for the level of supervision, access to treatment,
located is any indication. The results of these and so forth. In the absence of such carefully
studies will be welcome, as the science and controlled research, it might still be useful to
practice of spousal assault risk assessment is have risk assessment guidelines that (a)
obviously still in its infancy. There exists no sin- improve the reliability of assessments, (b) intro-
gle assessment instrument that would infallibly duce accountability into the assessment process,
predict recidivist domestic violence, and there and (c) allow more effective communication
never will be. Yet there are, among the instru- with regard to the risk to the courts, victims, and
ments we reviewed above, several that show so forth. Many of the reviewed instruments,
promising advances in recognizing factors rele- especially those that offer structured guidelines
vant to domestic violence and making empirical for gathering and reporting information, can
Dutton, Kropp / RISK INSTRUMENTS 179

serve these purposes presently. That said, any assaulters. It can serve as the basis for release
risk assessment device must be appropriate to planning, treatment placement, and safety
the level of expertise of the assessment users, assessment for the victim. Improperly applied,
who in turn should be sensitive to any legal it can mislead the courts, victims, and offenders
issues that could arise from its use. into falsely believing in an infallible science that
We emphasize that, regardless of which risk does not yet exist. However, we believe that the
measure is being used, the evaluator should accumulation of 25 years of research on domes-
always consider the implications for risk man- tic violence allows us to make informed deci-
agement. The process of risk assessment should sions about factors related to spousal violence.
not conclude with a prediction of dangerous- The judicious, ethical, and professional applica-
ness or lethality—this is only the beginning. The tion of risk assessment instruments has a valu-
evaluator should then formulate a risk manage- able role to play in the protection of spousal vio-
ment plan aimed at specific risk assessment
lence victims.
variables, particularly those dynamic variables
that might change over time. Such a plan can be
linked to the risk factors that are present and NOTE
absent (i.e., the perpetrator’s strengths or pro-
tective factors). It is important to remember that 1. This was determined by forcing demographic and criminal
history variables into a hierarchical multiple regression and al-
the true goal of the evaluator is to prevent vio- lowing the Psychopathy Checklist and Psychopathy Checklist–
lence, not predict it. This can only be achieved Revised to enter the regression equation only if it added unique in-
formation to the prediction of the outcome.
through sound planning based on a comprehen-
sive and informed risk assessment.
For the purposes of this review, we attempted REFERENCES
to find studies that addressed the risk for any American Psychiatric Association. (1994). Diagnostic and
form of spousal violence. Therefore, our defini- statistical manual of mental disorders (4th ed.). Washing-
tion of spousal violence was any actual, ton, DC: Author.
attempted, or threatened violence in the context Borum, R. (1996). Improving the clinical practice of vio-
of an intimate relationship. As the field of lence risk assessment. American Psychologist, 51,
945-953.
domestic violence risk assessment evolves, it
Browne, A. (1987). When battered women kill. New York:
will be useful to make distinctions with regard
Free Press.
to the risk for various forms of spousal assault.
Campbell, J. C. (1995). Prediction of homicide of and by
Campbell (1999) has made advances in our battered women. In J. C. Campbell (Ed.), Assessing dan-
understanding of factors associated with the gerousness: Violence by sexual offenders, batterers, and child
more narrow issue of domestic homicide, but it abusers (pp. 96-113). Thousand Oaks, CA: Sage.
might prove possible to isolate specific risk indi- Campbell, J. C. (1998, October). Commentary. In D.
cators for different levels of violence severity, McGrogan (Chair), Lethality and risk assessment for family
type (e.g., emotional abuse vs. sexual violence), violence cases. Symposium presented at the Fourth Inter-
national Conference on Children Exposed to Family
frequency, and imminence. We believe that it is Violence, San Diego, CA.
premature to make such Campbell, J. C. (1999, November). Issues in risk assessment
Risk assessment has distinctions given the in the field of intimate partner violence. Keynote address
become a well-used state of the literature. presented at the Conference on Risk Assessment and
term in the literature. Risk assessment has Risk Management: Implications for the Prevention of
Properly applied, the become a well-used term Violence, Vancouver, British Columbia, Canada.
practice of risk in the literature. Prop- Cocozza, J. J., & Steadman, H. J. (1976). The failure of psy-
assessment can serve erly applied, the practice chiatric predictions of dangerousness: Clear and con-
vincing evidence. Rutgers Law Review, 29, 1084-1101.
as a paradigm for of risk assessment can
Cooper, M. (1993). Assessing the risk of repeated violence
effective case serve as a paradigm among men arrested for wife assault: A review of the litera-
management of for effective case man- ture. Vancouver, Canada: British Columbia Institute on
spousal assaulters. a g e men t of sp o u s a l Family Violence.
180 TRAUMA, VIOLENCE, & ABUSE / April 2000

Douglas, K. S., Cox, D. N., & Webster, C. D. (1999). Violence Hart, B. (1990). Assessing whether batterers will kill. Read-
risk assessment: Science and practice. Legal and Crimino- ing, PA: Pennsylvania Coalition Against Domestic
logical Psychology, 4, 149-184. Violence.
Dunford, F. W. (in press). The San Diego Navy Experiment: Hart, S. D. (1998). The role of psychopathy in assessing risk
An assessment of interventions for men who assault for violence. Conceptual and methodological issues.
their wives. Journal of Consulting and Clinical Psychology. Legal and Criminological Psychology, 3, 121-137.
Dutton, D. G. (1995a). The domestic assault of women. Van- Hart, S. D., & Hare, R. D. (1997). Psychopathy: Assessment
couver, Canada: UBC Press. and association with criminal behavior. In D. Stoff, J. Breil-
Dutton, D. G. (1995b). Ascale for measuring the propensity ing, & J. D. Maser (Eds.), Handbook of antisocial behavior
for abusiveness. Journal of Family Violence, 10, 203-221. (pp. 32-40). New York: John Wiley.
Dutton, D. G. (1999). Perpetrator personality effects on Hemphill, J. F., Templeman, R., Wong, S., & Hare, R. D.
post-separation victim reactions in abusive relation- (1998). Psychopathy and crime: Recidivism and crimi-
ships. Journal of Family Violence, 14, 193-204. nal careers. In D. Cooke, A. Forth, & R. D. Hare (Eds.),
Dutton, D. G., & Kerry, G. (1999). Modus operandi and per- Psychopathy: Theory, research and implications for
sonality disorder in incarcerated spousal killers. Inter- society (pp. 375-397). Dordrecht, the Netherlands: Klu-
national Journal of Law and Psychiatry, 22, 287-300. wer Academic Publishers.
Dutton, D. G., Ogloff, J.R.P., Hart, S. D., Bodnarchuk, M., & Hoffman, P. B. (1983). Screening for risk: A revised salient
Kropp, R. (1997). Wife assault treatment and criminal factor score. Journal of Criminal Justice, 11, 539-547.
recidivism: An eleven-year follow up. International Holtzworth-Monroe, A., & Stuart, G. (1994). Typologies of
Journal of Offender Therapy and Comparative Criminology, male batterers: Three subtypes and the differences
41, 9-23. among them. Psychological Bulletin, 116, 476-497.
Dutton, D. G., & Parkin, C. M. (1999). Personality predictors Hotaling, G. T., & Sugarman, D. B. (1986). An analysis of
of dysphoria in response to intimate conflict scenarios. risk markers in husband-to-wife violence: The current
Unpublished manuscript, Department of Psychology, state of knowledge. Violence and Victims, 1, 101-124.
University of British Columbia, Canada. Kropp, P. R., & Hart, S. D. (in press). The Spousal Assault
Gelles, R. J. (1998, October). Commentary. In D. McGrogan Risk Assessment guide (SARA): Reliability and validity
(Chair), Lethality and risk assessment for family violence in adult male offenders. Law and Human Behavior.
cases. Symposium conducted at the 4th International Kropp, P. R., Hart, S. D., Webster, C. W., & Eaves, D. (1995).
Conference on Children Exposed to Family Violence, Manual for the Spousal Assault Risk Assessment Guide
San Diego, CA. (2nd ed.). Vancouver, Canada: B.C. Institute on Family
Goldsmith, H. R. (1990). Men who abuse their spouses: An Violence.
approach to assessing future risk. Journal of Offender Kropp, P. R., Hart, S. D., Webster, C. W., & Eaves, D. (1998).
Counseling, Services and Rehabilitation, 15, 45-56. Spousal Assault Risk Assessment: User’s guide. Toronto,
Gondolf, E. W. (1988). Who are those guys? Toward a Canada: Multi-Health Systems.
behavioral typology of batterers. Violence and Victims, 3, Kropp, P. R., Hart, S. D., Whittemore, K., & Bodnarchuk,
187-203. M. A. (1998, March). Spousal Assault Risk Assessment
Goodman, L. A., Dutton, M. A., & Bennett, L. (2000). Pre- Guide: Validation and implementation research. Paper pre-
dicting repeat abuse among arrested batterers. Journal of sented at the biannual meeting of the American Psy-
Interpersonal Violence, 15, 63-74. chology & Law Society, Redondo Beach, CA.
Gottfredson, D. M., & Bonds, J. A. (1961). A manual for intake Kropp, P. R., Hart, S. D., Whittemore, K., Webster, C. W., &
base expectancy scoring. San Francisco: California Eaves, D. (1998). The SARA guide: Validation and imple-
Department of Corrections. mentation research. Redondo Beach, CA: American Psy-
Grove, W. M., & Meehl, P. E. (1996). Comparative efficiency chology & Law Society.
of informal (subjective, impressionistic) and formal Marshall, L. (1992). Development of the Severity of Vio-
(mechanical, algorithmic) prediction procedures: The lence Against Women Scale. Journal of Family Violence, 7,
clinical-statistical controversy. Psychology, Public Policy, 103-121.
and Law, 2, 293-323. Melton, G. B., Petrila, J., Poythress, N. G., & Slobogin, C.
Hall, H. V. (1987). Violence prediction: Guidelines for the foren- (1997). Psychological evaluations for the courts: A handbook
sic practitioner. Springfield, IL: Charles C Thomas. for mental health professionals and lawyers (2nd ed.). New
Hare, R. D. (1980). A research scale for the assessment of York: Guilford.
psychopathy in criminal populations. Personality and Monahan, J. (1981). Predicting violent behavior: An assess-
Individual Differences, 1, 111-119. ment of clinical techniques. Beverly Hills, CA: Sage.
Hare, R. D. (1991). The Hare Psychopathy Checklist-Revised. Monahan, J., & Steadman, H. J. (1994). Violence and mental
Toronto, Canada: Multi-Health Systems. disorder: Developments in risk assessment. Chicago: Uni-
Harris, G. T., Rice, M. E., & Quinsey, V. L. (1993). Violent versity of Chicago Press.
recidivism of mentally disordered offenders: The Nuffield, G. L. (1982). Parole decisions making in Canada:
development of a statistical prediction instrument. Research towards decision guidelines. Ottawa: Ministry of
Criminal Justice and Behavior, 20, 315-335. Supply and Services Canada.
Dutton, Kropp / RISK INSTRUMENTS 181

Pence, E., & Lizdas, K. (1998). The Duluth safety and account- Webster, C. D., Harris, G. T., Rice, M., Cormier, C., & Quin-
ability audit: A guide to assessing institutional responses to sey, V. L. (1994). The violence prediction scheme: Assessing
domestic violence. Minneapolis, MN: MPDI. dangerousness in high risk men. Unpublished manuscript,
Quinsey, V. L., Harris, G. T., Rice, M. E., & Cormier, C. Centre for Criminology, University of Toronto, Canada.
(1998). Violent offenders: Appraising and managing risk. Weisz, A., Tolman, R. M., & Saunders, D. G. (2000). Assess-
Washington, DC: American Psychological Association. ing risk of severe domestic violence. Journal of Interper-
Roehl, J., & Guertin, K. (1998). Current use of dangerousness sonal Violence, 15, 75-90.
assessments in sentencing domestic violence offenders: Final Williams, K., & Houghton, A. (1999, November). Com-
report. State Justice Institute: CA mentary. In P. R. Kropp (Chair), Developments in Spousal
Saunders, D. G. (1992a). A typology of men who batter Assault Risk Assessment. Symposium conducted at the
women: Three types derived from cluster analysis. Conference on Risk Assessment and Risk Management:
American Journal of Orthopsychiatry, 62, 264-275. Implications for the Prevention of Violence, Vancouver,
Canada.
Saunders, D. G. (1992b). Woman battering. In R. T. Ammer-
man & M. Hersen (Eds.), Assessment of family violence: A
clinical and legal sourcebook (pp. 208-235). New York: Donald G. Dutton received his doctorate
Wiley. in social psychology from the University of
Saunders, D. G. (1993). Husbands who assault: Multiple Toronto in 1970. He cofounded the Assaul-
profiles requiring multiple responses. In N. Z. Hilton tive Husbands Project, a court-mandated
(Ed.), Legal responses to wife assault: Current trends and treatment program for men convicted of wife
evaluation (pp. 9-34). Newbury Park, CA: Sage. assault. In the course of providing therapy for
Saunders, D. G. (1995). Prediction of wife assault. In J. Camp- these men, he drew on his background in both social and
bell (Ed.), Assessing dangerousness: Violence by sexual clinical psychology to develop a psychological model for
offenders, batterers, and child abusers (pp. 173-196). Thou- perpetrators of intimate abuse. He has published more
sand Oaks, CA: Sage. than 100 papers and three books, including The Domes-
Sonkin, D. G. (1987). Assessment of court mandated batter- tic Assault of Women, The Batterer: A Psychological
ers. In D. G. Sonkin (Ed.), Domestic violence on trial: Psy-
Profile, and The Abusive Personality. Dutton has
chological and legal dimensions of family violence (pp. 174-
served as an expert witness in civil trials involving domes-
196). New York: Springer.
tic abuse and in criminal trials involving family violence,
Sonkin, D. G. (1997). The perpetrator assessment handbook.
Sausalito, CA: Volcano.
including his work for the prosecution in the O. J. Simpson
Sonkin, D. G., Martin, D., & Walker, L.E.A. (1985). The male trial. He is currently a professor of psychology at the Uni-
batterer: A treatment approach. New York: Springer. versity of British Columbia.
Steadman, H. J., Silver, E., Monahan, J., Appelbaim, P. S.,
Mulvey, E. P., Grisso, T., Roth, L. H., & Banks, S. (in P. Randall Kropp is a clinical and forensic
press) Aclassification tree approach to the development psychologist specializing in the assessment
of actuarial violence risk assessment tools. Law and and management of violent offenders. He
Human Behavior. works for the Forensic Psychiatric Services
Straus, M. A. (1991, July). Factors predicting serious violence Commission of British Columbia, is a
towards wives. Paper presented at the International Con- research consultant with the British
ference for Research on Family Violence, Durham, NH. Columbia Institute against Family Violence,
Straus, M. A., & Gelles, R. J. (1990). Physical violence in and is an adjunct professor of psychology at Simon Fraser
American families: Risk factors and adaptations to violence in University. He has conducted over 100 training
8145 families. New Brunswick, NJ: Transaction. workshops for mental health professionals, police officers,
Sugarman, D., Aldorondo, E., & Boney-McCoy, S. (1995).
and corrections staff throughout North America. This
Risk marker analysis of husband to wife violence: A contin-
training focuses on risk for violence, psycholegal
uum of aggression. Unpublished manuscript, Family
assessments (including psychopathy), and criminal
Violence Research Lab, University of New Hampshire.
Sugarman, D., & Hotaling, G.T. (1986). An analysis of risk
harassment (stalking). He has frequently consulted with
markers for husband to wife violence. Violence & Vic- provincial, state, and federal government ministries on
tims, 1, 101-124. matters related to violence against women and children,
Tolman, R. M., & Bennett, L. W. (1990). A review of research and the assessment and treatment of violent offenders. He
on men who batter. Journal of Interpersonal Violence, 5, has published journal articles, book chapters, and research
87-118. reports; and he is the coauthor to two works on risk
Trone, J. (1999). Calculating intimate danger: Mosaic and the assessment, the Manual for the Spousal Assault Risk
emerging practice of risk assessment. New York: Vera Insti- Assessment Guide and the Manual for the Sexual Violence
tute of Justice. Risk-20.

You might also like