Psychometric Data: Danger Assessment

The Danger Assessment (DA) was originally developed by Co-Investigator Campbell (1986) with consultation and content validity support from battered women, shelter workers, law enforcement officials, and other clinical experts on battering. The initial items on the DA were developed from retrospective research studies of intimate partner homicide or serious injury (Campbell, 1981; Browne, 1987; Berk et al., 1983; Fagan et al., 1983). The first portion of the measure assesses severity and frequency of battering by presenting the woman with a calendar of the past year. The woman is asked to mark the approximate days when physically abusive incidents occurred, and to rank the severity of the incident on a 1 to 5 (1=slap, pushing, no injuries and/or lasting pain through 5=use of weapon, wounds from weapon) scale. The calendar portion was conceptualized as a way to raise the consciousness of the woman and reduce the denial and minimization of the abuse, especially since using a calendar increases accurate recall in other situations (Campbell, 1995; Ferraro et al., 1983). In the original scale development, 38% of women who initially reported no increase in severity and frequency, changed their response to “yes” after filling out the calendar (Campbell, 1986; Campbell, 1995).

The second part of the original DA is a 15-item yes/no dichotomous response format of risk factors associated with intimate partner homicide. Both portions of the instrument take a total of approximately 20 minutes to complete. The DA is scored by counting the “yes” responses and although no cutoff score has been published, a score of 9.3 was found in ED populations versus .75 in nonabused women, supporting discriminant group validity. The ten initial studies using the DA instrument demonstrated internal consistency reliability of 0.60 to 0.86. (Campbell, 1986; Campbell, 1995; McFarlane et al., 1992; McFarlane et. al., 1996; McFarlane et. al., 1998). In two studies in which test-retest reliability was assessed, it ranged from 0.89 to 0.94 (Stuart et al., 1989; Campbell, 1994). All samples included a substantial portion of minority women (primarily African-American) and women from a variety of settings, and internal consistency did not vary significantly among the ethnic groups (Campbell, 1995). The DA instrument has been utilized in a variety of settings including battered woman’s shelters, criminal justice and health care settings.

Convergent construct validity has been supported in the majority of the studies with moderate to strong correlations with instruments measuring severity and frequency of domestic violence (Index of Spouse Abuse, Conflict Tactics Scale and abuse related injury) (Campbell, 1995). Five additional independent studies have been conducted since the published review with additional reliability support (internal consistency of .69 - .78) (Campbell et. al., in press). In further support of construct validity, the DA had significant correlations (r = .56 - 62) with symptoms of Post-Traumatic Stress Disorder (PTSD) (Woods, in press; Silva et. al., 1997) and severity and frequency of domestic violence (r = .75) (McFarlane et. al., 1998) in the newly reported studies. There have been two recent small predictive validity studies with the DA predicting reassault more successfully than the CTS in one and 10 of the 15 DA items from criminal justice record review being moderately successful in predicting reassault, but the woman’s perception even more predictive in the other (Goodman et. al., 1999; Weisz et. al., 2000). In addition, two large prospective studies of reassault have also found independent support for the predictive validity of the DA (Heckert & Gondolf, 2001, N = 499; Williams & Coniff, 2001).

Our multidisciplinary research team recently completed a multi-city case (femicide and attempted femicide cases) control (abused women in the same cities) femicide study (Campbell et. al., in press). All but one of the 15 yes/no items on the original DA were significant predictors of intimate partner femicide (homicide of women). That one item was victim suicidality. Since the item is meant to predict homicide of the perpetrator which was not tested in this study and suicide is also potentially fatal to women and should be assessed, that item has been retained on the revised DA. The item about perpetrator violence toward the children was changed to “Does he threaten to harm your children?” an item taken from the Sheridan HARRASS instrument. This item was more predictive than asking about reported child abuse and also avoids the mandatory reporting of child abuse engendered by a positive response to the original form of the question. The gun item (originally “Is there a gun in the home?”) was changed slightly to “Does he own a gun?” so that cases where he has left the home with his gun or lives elsewhere but owns a gun. The first two items on severity and frequency were combined into one, since the multivariate analysis showed that each of these variables were equally predictive. Five other items were added that were significant in the multivariate analysis. They are: a) Do you have a child that is not his?, b) Is he unemployed?, c) Have you left him during the past year? (If have never lived with him, check here___), and d) Do you currently have another (different) intimate partner? and a stalking item, e) Does he follow or spy on you, leave threatening notes, destroy your property, or call you when you don’t want him to? The resulting revised DA is 20 items long. Since usage is only in its first year, comments on wording, ease of use and other issues with this revision will be gratefully accepted.*

Further reliability and validity support was found in the multicity femicide study. The results demonstrated acceptable reliability of the DA among femicide victims (r = .80), attempted femicide victims (r = .75) and abused women (r = .74) in 12 US cities. Discriminant group validity was supported by significant (p = .004) differences between the unweighted mean scores of cases (femicide victims = 220; X = 7.4) and controls (abused women from the same cities = 343; X = 3.2) for the controls. Sensitivity and specificity of the original DA without weighting was marginally acceptable. Sensitivity was acceptable (83.4% of women killed correctly identified) at a relatively low score (4 or more), but at that score, specificity is relatively low, with almost 40% (39.2%) of the abused controls (not killed) also at this score. At a cutoff score of 9 or higher, specificity is good, with 94% of the women who were below that score in the control group. However, only 40% of the femicides cases scored that high on the DA (sensitivity). At a cutoff of 7, both sensitivity (58%) and specificity (87%) are fairly good, but one is concerned about the 42% of women in extreme danger who would be missed at that cutoff. However, using weightings based on the risk factor analysis (Campbell and Webster, 2004) plus the new items added to the DA (see DA instrument) resulted in a very good ROC analysis with 90.8% of the cases below the curve. Thus, although the original DA and even better, the revised DA, has support for reliability and both discriminant and predictive validity with a simple additive scoring, the most accurate prediction is achieved with a weighted scoring. We have now completed a weighted scoring procedure with ranges of scores considered more or less acutely dangerous.

* We have considered carefully the “choking” language (vs. strangulation) and have decided that women are much more likely to respond to the word “choking” as reflecting their experience, although strangulation is a more accurate medical (and criminal justice) term. We have also carefully thought about the women’s perceptions issues and have continued to include it in the “do you believe he is capable of killing you” item along with his threats of killing based on our analysis. We also urge people to consider that women’s perceptions of threat of reassault, shown to be important in many studies, may be more accurate than their perception of risk of homicide. Only 47% of our femicide victims (according to proxy informants) and more importantly 53% of our attempted femicide victims accurately predicted their risk before the lethal or near lethal event.

Pet Abuse

In our femicide study (Campbell et al, 2003), we specifically tested pet abuse as a variable. We asked first if there had been a pet in the home during the relationship and then if the batterer had abused that pet. Ten per cent of our attempted/actual homicide cases combined reported pet abuse while 8.5% of the abused women in the same cities (controls) reported pet abuse, a nonsignificant difference. As far as we know, ours is the only controlled study that has systematically investigated this issue. Thus an item on pet abuse has not been included on the Danger Assessment. We did find a greater percentage of pet abuse among the attempted femicide victims than was reported among the actual femicides, suggesting that pet abuse might not be known to a proxy informant while the actual victim of attempted femicide was more aware of this issue. In addition, there were a few cases where abuse to a pet was used as a clear symbolic threat of death to the woman. Thus, the issue warrants further investigation in future research but the need to control for this occurrence among abused women not killed is a very important strategy in order to be sure about its importance as a risk factor for femicide over and above prior domestic violence. We did find that pet abuse was the strongest risk factor with an adjusted odds ratio of 7.59 (1.61, 35.96) for domestic violence in comparison to the nonabused women in the same cities in the femicide study (Walton-Moss et al, 2005).

Campbell, J. C., Webster, D., Koziol-McLain, J., Block CR, Campbell, D., Curry, MA, Gary, F, Sachs, C. Sharps, PW, Wilt, S., Manganello, J., Xu, X. (2003). Risk factors for femicide in abusive relationships: Results from a multi-site case control study. American Journal of Public Health, 93, 1089-1097.

Walton-Moss, B. J., Manganello, J., Frye, V. & Campbell, J.C. (2005) Risk factors for intimate partner violence and associated injury among urban women. Journal of Community Health, 30, 377-389

Ascione,F. R. (2005).  Children and Animals:Exploring the Roots of Kindness and Cruelty. West Lafayette, IN: Purdue University Press.  Sponsored by the Kenneth A. Scott Charitable Trust and American Humane.

Ascione, F. R. (Ed.), International handbook of theory, research, and application on animal abuse and cruelty. West Lafayette, IN: Purdue University Press.