Only in recent years has the widening scope of domestic violence achieved such national prominence. Highly publicized cases like the Nicole Brown Simpson murder, and the subsequent disclosure of her frantic 911 calls to police alleging beatings by her celebrity husband O.J. Simpson, have brought the issue of domestic violence to worldwide attention. Last year, according to the U.S. Department of Justice's (DOJ) Violence Against Women Office (1998), over 91,000 Americans-primarily women-accessed the National Domestic Violence Hotline, averaging over 280 calls a day.
Overall rates of violence against women, including both lethal and nonlethal violence, are undoubtedly much higher: one widely cited epidemiological study reported that 11.6% of the women surveyed had been the victims of physical assault by spouses or sexual partners at some time during the preceding year, and 3.4% were the victims of severe violence (Straus and Gelles, 1990). At least one-third of these women reported being assaulted by a partner over the course of their relationship (Straus and Gelles, 1990).
According to data from the DOJ Bureau of Justice Statistics' redesigned National Crime Victimization Survey of 1992 and 1993, women were victims in more than 4.5 million violent crimes-29% of them perpetrated by husbands, former husbands and boyfriends. During 1992, approximately 28% of female homicide victims (1,414 women) were killed by these same lone offenders (Bachman and Saltzman, 1995).
Most domestic assaults go unreported for a variety of reasons, including feelings of shame and fear of the consequences of disclosure, e.g., fear of retribution (Straus and Gelles, 1990). Studies have found that women who are battered are often psychologically intimidated as well (Walker, 1984).
Thus, partner assault often occurs in the context of an emotionally abusive relationship in which women are demeaned, threatened and isolated from social supports, such as friends and family members (Walker, 1984). As a result of prolonged physical and psychological victimization, women may lack the emotional and social resources to extricate themselves from what may be potentially life-threatening situations. Therefore, despite the heightened publicity surrounding the issue of domestic violence, the assault of women by their partners remains largely a hidden problem, with most victims continuing to go unidentified, sometimes with tragic consequences.
The detection of "hidden" victims of domestic violence may be improved through the adoption of routine screening procedures (Aldarondo and Straus, 1994). Victimized women often come into contact with medical, legal, mental health care or social service professionals, but their victimization status frequently goes unrecognized because it is not inquired about routinely. Although obvious signs of physical assault such as bruises, marks or broken bones may provoke questions about partner violence, such inquiries are less likely in the absence of such physical sequelae, even when known risk factors for domestic violence are present (Table 1).
For example, recent studies suggest that many women are battered during pregnancy, thus jeopardizing the unborn as well (Helton et al., 1987). Routine screening might help to prevent some of these assaults. Substance abuse is another known risk factor for partner violence, with drug or alcohol(Drug information on alcohol) abuse being implicated in 50% or more of domestic violence cases (Kantor and Straus, 1989). Yet many substance abuse treatment programs do not incorporate screening questions about domestic violence in their standard intake assessments. Even more surprisingly, couple and family therapists frequently do not ask their clients about partner violence, despite studies showing that as many as two-thirds of couples in such treatments engage in physically assaultive behavior (Aldarondo and Straus, 1994).
Victimized women may also be more likely to develop medical problems that bring them into contact with health care providers. For example, women who are forced to work in the sex trade may seek treatment for sexually transmitted diseases (Jenny et al., 1990), but domestic violence may not always be asked about in these cases. Finally, women with histories of partner violence are more likely to present for medical treatment for unexplained somatic conditions such as fibromyalgia, a pain syndrome of unknown etiology (Walker et al., in press). In fact, such women are often high health care utilizers whose somatic complaints are sometimes misdiagnosed by physicians who are unaware of their traumatic histories (Walker et al., in press).
These considerations suggest that the implementation of routine screening procedures, particularly in settings where women present with physical or psychological conditions that are often associated with partner abuse, may dramatically increase the detection of victims of domestic violence.
One approach to implementing domestic violence screening is to conduct in-service training for health services and other professionals designed to increase awareness about partner violence issues and provide practical guidelines for conducting such assessments. For example, professionals can be trained to ask questions in objective, neutral terms (e.g., "Do your fights with your husband/wife ever get physical?") and to avoid using potentially pejorative terms such as victim and assault, which may be distressing to victims and therefore inhibit disclosure (Stordeur and Stille, 1989). Another approach is to develop standardized screening instruments that can be quickly and easily self-administered, for example, in a doctor's office waiting area. Although both approaches are desirable, the latter has several potential advantages, including its efficiency and standardization. Moreover, self-report screening measures may be relatively less threatening or embarrassing than face-to-face interviews, and may therefore result in increased rates of disclosure.
To that end, my colleagues and I are currently developing and validating such a screening questionnaire, the Partner Violence Inventory (PVI), to improve the detection of domestic violence victims in a variety of clinical and community settings.
The current version of the PVI consists of 37 items forming seven scales and takes about five minutes to administer. It is introduced by an inquiry about "your feelings and experiences in your relationship with your current sexual partner" or, for respondents not currently in a sexual relationship, about "the most significant recent sexual partner you have had." Items are written using objective, nonjudgmental language, and terms such as assault and abuse that may be threatening to respondents are avoided (Table 2). Items are neutral with respect to gender and sexual orientation, so they can be administered to either men or women and to partners in either opposite sex or same sex relationships. All items are responded to on a five-point Likert scale, with response options ranging from "Never True" to "Very Often True."
Although other measures have been developed that might be used to detect victims of partner violence, the PVI incorporates several features that make it a desirable screening instrument, including its brevity, specificity and assessment of multiple forms of victimization. The PVI assesses three types of victimization by partners-physical, sexual and emotional assault-as well as partner drug and alcohol abuse. The questionnaire also inquires about forms of physical fighting that may appear to be mutually instigated (e.g., mutual pushing, shoving or slapping) and less dangerous than outright assault, but which may nevertheless escalate into more extreme violence. The questionnaire also contains items reflecting more positive, affiliative aspects of partner relationships, such as warmth and affection, to provide a balance between positively and negatively balanced items. Finally, the questionnaire contains a minimization scale to detect respondents who may be underreporting their victimization experiences.
I am conducting ongoing validation research of the PVI with colleagues from several academic institutions: Edward Walker, M.D., and Ann Gelfand, R.N., at the University of Washington School of Medicine; Judith A. Stein, Ph.D., at the University of California, Los Angeles, Department of Psychology; Martha Medrano, M.D., David Desmond, M.S.W., and William Zule, Dr.P.H., at the University of Texas Health Science Center at San Antonio; and Murray Stein, M.D., at the University of California, San Diego, School of Medicine.
In a preliminary study presented at the May 1997 meeting of the American Psychiatric Association in San Diego (Bernstein et al., 1997), we presented a pilot version of the PVI containing 121 items to 234 respondents at four sites: a community-based sample of 87 male and 51 female substance abusers in San Antonio; a general psychiatric sample of 21 male and 43 female patients in Seattle; a sample of 19 women residing at a domestic violence shelter in Seattle; and a sample of 13 women attending a domestic violence program in San Diego.
The respondents were quite diverse in terms of age, gender and ethnicity. Exploratory and confirmatory factor analyses were performed to reduce the number of items on the scale and derive subscales that had comparable psychometric properties when given to both female and male respondents. When the validity scale items were excluded from the analysis, a six-factor solution involving 34 items was found to be optimal:
I. Partner drug and alcohol problems (six items)
II. Sexual assault (five items)
III. Warmth and affection (six items)
IV. Physical assault (five items)
V. Physical fights (four items)
VI. Emotional assault (eight items)
Thus, the factor analytic findings corresponded to the hypothetical partner violence constructs for which the items had been written. The internal consistency of the six factors was excellent, as indicated by Cronbach's alphas ranging from 0.83 to 0.90 (Table 3). Fit indices for the confirmatory model were good for both male and female respondents (females: Satorra-Bentler x²=703.24/509 df, Robust Comparative Fit Index=0.93, males: x²=623.50/509 df, RCFI=0.90), suggesting that male and female subjects responded to the items in an equivalent manner (i.e., that the items held the same meaning for both male and female respondents). Finally, the PVI showed excellent test-retest reliability when the scale was readministered to a subgroup of 64 respondents from the San Antonio sample after a one-week interval, with intraclass correlations ranging from 0.84 to 0.90 (Table 3). Thus, responses to the scale's items were highly stable over a brief test-retest interval.
These preliminary findings supported the reliability and validity of the PVI as a screening measure for victims of domestic violence. My colleagues and I have subsequently collected data on several hundred additional respondents, including a normative community sample of women, that we are currently analyzing.
Our goals are to make further refinements in the questionnaire, examine the consistency of its factor structure across a broad range of respondents varying in age, gender and ethnicity; and investigate its external validity, for example, in relationship to victimization histories gathered by structured interview from the same respondents. We also hope to develop and validate a short form of the PVI that is no more than 10 items in length for very rapid screening. These validation studies are necessary to ensure that the PVI can be utilized across a broad variety of settings and populations. Ultimately, however, the success of this endeavor will depend on the willingness of clinicians and institutions to implement routine screening procedures for victims of partner violence. Although such screening requires time and effort, its benefits can be calculated in terms of the hundreds of lives it may save.
