Could We Have Known? A Qualitative Analysis of Data from Women Who Survived an Attempted Homicide by an Intimate Partner

Abstract

OBJECTIVE

To examine in-depth the lives of women whose partners attempted to kill them, and to identify patterns that may aid in the clinician's ability to predict, prevent, or counsel about femicide or attempted femicide.

DESIGN

Qualitative analysis of 30 in-depth interviews.

SETTING

Six U.S. cities.

PARTICIPANTS

Thirty women, aged 17–54 years, who survived an attempted homicide by an intimate partner.

RESULTS

All but 2 of the participants had previously experienced physical violence, controlling behavior, or both from the partner who attempted to kill them. The intensity of the violence, control, and threats varied greatly, as did the number of risk factors measured by the Danger Assessment, defining a wide spectrum of prior abuse. Approximately half (14/30) of the participants did not recognize that their lives were in danger. Women often focused more on relationship problems involving money, alcohol, drugs, possessiveness, or infidelity, than on the risk to themselves from the violence. The majority of the attempts (22/30) happened around the time of a relationship change, but the relationship was often ending because of problems other than violence.

CONCLUSIONS

Clinicians should not be falsely reassured by a woman's sense of safety, by the lack of a history of severe violence, or by the presence of few classic risk factors for homicide. Efforts to reduce femicide risk that are targeted only at those women seeking help for violence-related problems may miss potential victims.

Femicide,^1–3 the murder of women, is one of the leading causes of premature death for women in the United States,⁴ and the leading cause of death for African-American women aged 15–34 years.⁵ Whereas only 3% to 6% of male homicide victims are killed by an intimate partner,^6,7 30% to 55% of femicide victims are killed by an intimate partner.^6,8–11 Despite widespread efforts to decrease intimate partner violence (IPV), and a steady decrease in the rate of murders by intimate partners where the victim is male, there has been little improvement in intimate partner murder rates where the victim is female.^6,8,11,12

Studies have found that 65% to 80% of intimate partner femicide victims were previously abused by the partner who killed them.^13–15 The question then becomes: among abused women, how does one accurately determine who is at highest risk for serious harm or mortality? Standard medical education curricula on domestic violence teach that clinicians should look for a history of severe or escalating domestic violence, or for classic signs of increased risk such as prior threats to kill or assaults with a weapon. Formal, psychometrically tested lethality assessment tools such as the Danger Assessment (DA) have been shown to improve clinical assessment.¹⁶ The DA is a clinical and research instrument that has been designed to assist battered women in assessing their danger of being murdered by their intimate partner. The original DA measures the total number of “yes” responses by the battered woman on the 15-item risk factors associated with intimate partner homicide and is scored by counting the “yes” responses; a higher number indicates that more of the risk factors for homicide are present in the relationship. The DA has the most published data on risk factors for intimate homicide and concurrent and predictive validity information. However, it is not a standard part of medical practice and is not in widespread use among practicing clinicians. Our objectives were to explore in-depth the lives of 30 women who survived an attempt on their life by an intimate partner. We also wanted to identify themes that may aid in the clinician's ability to predict, prevent, or counsel about femicide and attempted femicide.

METHODS

Participants

This qualitative analysis was performed in conjunction with an 11-city case-control study to determine the risk factors of actual and attempted intimate partner femicide.^13,17 The case-control study compared data on women who had been murdered by an intimate partner and women who had survived a femicide attempt by an intimate partner with age- and location-matched controls who were in violent relationships, but who had not had an attempt on their lives. The attempted femicide sample consisted of 182 consecutive cases that had survived an attempted femicide by an intimate partner for the years 1994–2000. Attempted femicide was defined as: a gunshot or stab wound to the head, neck, or torso; loss of consciousness from strangulation, trauma, or attempted drowning; other severe injuries that could have led to death; and/or verifiable evidence of unambiguous intent to kill the victim. Cases were eligible if the perpetrator was a current or ex-intimate partner and the case was designated as “closed” by the police (an arrest made or adjudication depending on the jurisdiction). Participants also had to be living separately from the perpetrator in a safe environment. There were no age restrictions to participation. At each site, coinvestigators worked with local law enforcement, the district attorney's office, domestic violence shelters, and trauma centers to identify women who would be eligible to participate in the larger case-control study. They mailed an introductory letter to women meeting the inclusion criteria. Because of concerns for safety in case the perpetrator could intercept the woman's mail, the letter did not mention the attempted homicide, but asked the woman if she would be interested in participating in a study of women's health and relationships. Investigators followed up the letter with a phone call, except in cases where women indicated they did not wish to be contacted.

Investigators in 6 cities (Baltimore, Md; Houston, Tex; Kansas City, Mo; Portland, Ore; Tampa, Fla; and Wichita, Kan) agreed to collaborate on the qualitative component of the study. At the end of the interview for the case-control portion of the study, interviewers in these cities asked women if they would be willing to discuss their stories further in an open-ended interview. Recruitment started with women who had most recently experienced a murder attempt and continued until the goal of 30 participants was met. Sample size was based both on the anticipated number needed for saturation and on the desire to obtain approximately 20% participation from each of the cities. We also purposely sampled for ethnic diversity. The study was approved by the institutional review boards of all involved institutions.

Data Collection

Subjects participated in an audiotaped, semistructured, in-depth interview. The purpose of the interview was to allow women to describe, in their own words, their relationship with the partner who had attempted to kill them, and their perceptions of the activities and events that preceded the attempt. In order to standardize procedures across the sites, two of the authors developed guidelines for conducting the semistructured interviews. Interviews were divided into 5 sections focusing on: (1) the intimate relationship; (2) the events preceding the attempted homicide; (3) the event itself; (4) changes since the event; and (5) interactions with health care workers, counselors, and police. Each section began with open-ended questions and was followed by preestablished probes. Interviewers specifically directed women to think about any events in the month, week, or day prior to the event that may have let them know that something was different or that something was going to happen. Interviews lasted 30 to 90 minutes, and were transcribed verbatim.

Coding and Analysis

We analyzed transcripts in accordance with thematic analysis processes as described by Ryan and Bernard.¹⁸ We developed a list of provisional codes based on an initial reading of 5 randomly selected interviews. The list of provisional codes was circulated among 6 team members, who reread the 5 interviews and reached an agreement on the application of these codes to those interviews. Two investigators then coded the entire sample with multiple readings of transcripts, looking for common themes among the participants’ stories. New codes were added as themes emerged. In addition to looking for themes that emerged spontaneously from the interviews, the two authors specifically documented the presence or absence of known risk factors. These included the 15 items on Campbell's DA,¹⁶ an instrument used to assess lethality among abused women, as well as risk factors identified in the associated case-control analyses of intimate partner femicides.¹⁷ The investigators then discussed each interview until agreement was reached regarding the coding of risk factors and common themes. Interview summaries and a list of features most salient for each woman were sent to a team of 7 other coauthors. This team used the summaries to review and validate the identified themes and conclusions. Participants themselves were not available to validate identified themes and conclusions. Formal DA scores¹⁶ were available as part of the case-control portion of the larger study. We compared data on our group of 30 participants with data on all 182 attempted femicide cases in order to detect important sampling differences regarding known risk factors for lethality.

RESULTS

Participants

Thirty women, aged 17 to 54 years, participated in our study: 10 from Baltimore, Md, 4 from Tampa, Fla, 5 from Portland, Ore, 6 from Houston, Tex, 2 from Wichita, Kan, and 3 from Kansas City, Mo. In all cases, the perpetrator was male. Interviews occurred 5 months to 2 years after the femicide attempt. A comparison of the quantitative data obtained on the 30 women in our study and the larger group of 182 women in the attempted femicide arm of the case-control study showed similar demographic characteristics and risk for femicide as measured by formal DA scores (see Table 1).

Table 1.

Demographic Characteristics and Danger Assessment Scores for Women Participating in This Study and the Attempted Femicide Arm of the Larger Case-Control Study

	Qualitative Study (N = 30)	Attempted Femicide Arm of Case-Control Study (N = 182)
Mean age, % (SD)	35 (9.2)	33 (9.3)
Race/ethnicity, %
African American	13 (43)	55
European American	14 (47)	23
Latino	2 (7)	18
Other	1 (3)	3
Danger Assessment score, mean (range)	7.8 (3 to 14)	6.9 (0 to 15)
Prior physical abuse, (N)	20 (67)	129 (71)

Prior Violence, Control Issues, and Known Risk Factors

Twenty women (67%) had a history of repeated physical or sexual abuse by the partner who attempted to kill them, a similar proportion to the 71% prior IPV in the larger sample. The severity and frequency of the abuse ranged dramatically, from those that would occasionally get pushed or slapped to others who suffered frequent or life-threatening injuries. Five additional women (17%) described episodes of minor violence that would be difficult to describe as a pattern of physical abuse. For example, one woman stated her husband had pushed her once during an argument many years earlier, but he had never done that again. Five women (17%) had no prior history of physical or sexual abuse. Women often spontaneously mentioned known risk factors for femicide, such as prior threats to kill, escalating frequency or severity of violence, assaults during pregnancy, the partner's drug use, or the partner's access to a weapon, but the total number of risk factors mentioned varied greatly, with some women only mentioning 1 or 2 factors, and 1 woman mentioning 10 of the 15 items on Campbell's DA.¹⁶

Power and control issues were prominent in the majority of the relationships. Twenty-five of the 30 women (83%) described examples of their partners using stalking, extreme jealousy, social isolation, physical limitations, or threats of violence. The intensity of the control varied greatly. For some, it resembled romance. For example, one man would constantly show up to his wife's work to bring her flowers and gifts. It was only in retrospect that she realized he was checking up on her all the time. For others, it was so extreme that the women were prisoners in their own homes, with partners who did not allow them to go anywhere by themselves, constantly stalked them, and still accused them of somehow finding a way to sleep with other men.

Many of the women who had not experienced significant prior physical abuse did experience some form of control, leaving only 2 women (7%) who experienced neither physical abuse nor controlling behavior from the perpetrator. In both cases, the perpetrator's jealousy or prior violence toward other women did ultimately become apparent, but only after the attempted femicide. In 1 case, which involved 2 teenagers, the victim's boyfriend heard that she had been out with someone else, confronted her about it, and shot her when she attempted “to get him out of [her] face” by telling him that she “would rather be with the other boy.” In the other case, the victim found out after the attack that her husband had previously served a 7-year sentence for attempting to kill his first wife.

Wide Spectrum of Abuse

Women's stories fell along a wide spectrum of abuse, as defined by the severity of violence and control, and the number of classic risk factors mentioned. At the severe end of the spectrum, women and anyone who heard their story were aware of the extreme danger, but the system failed to provide adequate protection. At the low end of the spectrum, it may have been impossible for anyone to suspect the potential for lethality. Most women fell somewhere in the middle of this continuum. Even though clues existed of heightened danger, few recalled discussing this risk with healthcare workers or counselors.

The following is an example of a case at the severe end of the spectrum. This woman suffered years of repeated severe physical assaults by her husband, as well as forced sex and constant threats with guns. Her partner displayed extremely controlling behavior. He would not let her leave the house without him, and she had to page him to get permission to take a bath. Her abuse was not hard to detect. In fact, her obstetrician had to admit her to the hospital in order to keep her partner from forcing her to have sex in the face of a threatened premature delivery. The final attack came after a period of escalating violence. Many people tried to help her, but it was her father-in-law's message that rang true: “There is no way out. The system can’t help you. These guys beat women for years and then kill them when they try to leave.” Here the issue was not of recognition of risk, but of failure of adequate resources and protection.

This next woman was at the other extreme of the spectrum. She had already broken up with a man whom she described as a sweet, but possessive “gentle giant.” There had been no violence or abuse in the relationship. The break up was calm and she had no contact with him for 6 weeks. When he called asking to meet as friends she agreed with no hesitation. He then came to her home, tied her up, and beat her for 12 hours. At this extreme, the problem comes when providers assume all attacks are predictable. As this woman describes, “So, [the counselor] was like ‘Women should be more aware.’ Excuse me. If there are no signs to you—no previous throwing, hitting, screaming— how are you supposed to know what's going to happen?”

The last example represents someone at the middle of the spectrum. She had been married for 12 years to a man she described as a “withdrawn workaholic.” She felt her partner was “mean when drunk” but said he was not drunk often. Once, many years earlier, he had shoved her down the stairs when she was pregnant. Since then, he pushed or shoved her about once per month, but never hit her. She did not mention extreme jealousy or stalking during the relationship, but did say, “He was pretty much the boss of most things.” Her husband had recently moved out, but he wanted to get back together. In the month before the attack, his escalating stalking made her scared to stay at home alone. One night he went through her home with a sledgehammer, chainsaw, and toxic chemicals, destroying everything the family owned. Though neighbors called the police, the police felt they could not intervene as he was an owner of the house and there was no restraining order against him. The next morning, he was waiting for her in the house with a loaded crossbow. During her relationship, she had wanted help, but did not tell health care workers about the abuse. “I didn’t tell her because [my husband] was always in the room with me … I mean, if she could have just looked at my face or my eyes she would have known. You know how you give those little signals.” By the time of her attack this woman recognized her risk, as would have an astute clinician, but she had never been assisted in systematically assessing her risk or accessing orders of protection, safety planning, shelters, and other available resources.

Women's Sense of Surprise

Almost half of the women (N = 14), said they were completely surprised by the attack. As one woman stated, “I didn’t really realize what big trouble I was in until I was to the point of where I thought I was going to die.” This particular woman had 6 of the risk factors from the DA instrument, and though not at the extreme of the abuse spectrum, she had a higher than average level of prior abuse. Still, she saw her marital problems as revolving around her partner's alcohol use, bipolar disorder, and “anger control” problems. In reading her story, one gets the sense that both she and her support community started to normalize his frequent violence and threats as part of a typical bad marriage. As she states, her “family's attitude was ‘get out of the relationship or quit whining about it.’” No one ever made the connection between his behavior and the risk to her own life.

Timing of Attack Around Relationship Change

In 22 of the 30 cases (73%), the femicide attempt occurred just around the time of a significant relationship change. In most of these cases, as expected, the woman was trying to leave the relationship. But her reason for ending the relationship was rarely (3 cases) because of his violence or threats. More often, she was ending the relationship owing to arguments over money (5 cases), his use of drugs or alcohol (6 cases), his infidelity (6 cases), his possessiveness (3 cases), or a combination of these. For example, despite suffering frequent violence from her partner, one woman stated, “Well, we had a good relationship. We argued sometimes, but other than that it was a good relationship with me and him and my children.” Their final argument, like many previous ones, was over drugs and money. “And I said ‘You need to leave because, you know, if you are going to trip out about money that belongs to me and my church, and blah blah, you need to go on and move out.’ And that is when the argument got really deep, and he said, ‘I am not going to ever leave you. I’ll kill you before I leave you.’ And the next thing you know, I was shot.”

Often, there seemed to be a sequence of events which included frequent arguments over his behavior, her decision to end the relationship, his pleading to get her back, and then finally the attack when he realized that she really meant it. For example, this is how 1 woman describes her final interaction with her partner, whom she had kicked out of the house because of his drug use: “I sit on my bed and he looks at me and says, ‘So what you are telling me is that you do not love me anymore.’ I said, ‘No.’‘You want me out of the house?’ I said, ‘Yes, I do.’ He says, ‘O.K., well then I am going to kill you.’ That is when he lunged at me.” Another woman, who left her husband because of his cheating, explains, “No matter what, he owned me and he owned the kids and he’d kill us all before he would let us go.” In four cases, it was the man who had initiated the separation, but he then became violent when she began seeing other people or refused to have him back. In 3 additional cases, the attack occurred in the context of the man wanting to end the relationship. In these cases the man's reasons for attempting to kill his partner were unclear, but all potentially involved financial gain, a new partner, or both.

DISCUSSION

Our intent was to describe the danger in the lives of women who had been almost killed by an intimate partner, their perceptions of risk, and the circumstances that led up to the near-fatal attack. We were not trying to determine the validity of particular risk factors for intimate partner femicide—that was the purpose of the larger case-control study. Our study demonstrates the complexity and variety of women's stories, offers insights as to why many do not perceive their risk, and warns of the pitfalls of expecting all victims of intimate partner femicide to fit into a classic picture of severe abuse.

Our study has several limitations. First, we used a convenience sample of women who could be located after the attack, lived in a safe environment apart from their abuser, and agreed to a second interview after completing the case-control portion of the study. As is appropriate in qualitative studies, we were more interested in obtaining a sample of key informants who could safely offer in-depth insights than we were in randomly selecting participants. As such, we cannot generalize our results to those women who could not be located, who were still living with their abusers, or who did not wish to further discuss their experiences. The similarity in our sample's DA scores and those of the women in the attempted femicide arm of the case-control study suggests that our sample was similar to the larger group of attempted femicide survivors in terms of prior risk. It is possible, however, that women with less typical histories of abuse or with fewer classic features of abuse were more likely to participate, thus skewing our conclusions.

Second, in keeping with the principles of qualitative research, we attempted to learn how the participants saw their lives and relationships, and what was most salient to them about the attack. As such, the qualitative data may not reflect all the information pertinent to each case. Even so, the women's perceptions and stories are important, and they may best reflect the information obtainable by a clinician, especially if the clinician is not using a formal lethality assessment instrument. Given that these women had already participated in a lengthy multiple-choice survey, we were only able to perform one additional in-depth interview per participant. As such, our results only represent their views at one moment in time. It is possible that women would describe their experiences differently over time.

Even though they would have been eligible, our study did not include any same-sex couples. When thinking about the stories of women at risk for femicide, it is unclear how similar or different the lives of women in same-sex relationships would be from those in heterosexual relationships. Despite our attempts to include women with a variety of racial and ethnic backgrounds, and our success in recruiting African-American women, we had a relative underrepresentation of Latina, Asian, or Native-American women. These results may be most applicable to European-American and African-American women. Also, by interviewing only victims, we could only get half of the story, and cannot make any conclusions as to the perpetrators’ perceptions or motivations. Lastly, it is possible that women who were killed by their partners lived different lives prior to the attack than those who survived a femicide attempt. However, the multivariate analysis in the larger case-control study found that the data for actual and attempted femicides were remarkably similar in cases where there had been prior domestic violence.¹⁹ This finding lends support for the attempted femicide cases being important sources of information for actual femicides.

Despite these limitations, our study adds important insights into the lives of women at risk of femicide. It is well known that a history of prior physical abuse is the primary risk factor for intimate partner femicide.¹⁰ Still, prior research has found that between 20% and 30% do not have such history.^14,15,20 Elements such as stalking, extreme jealousy, physical or economic limitations, and social isolation are generally thought of as features of intimate partner violence,^20–24 and have been shown to correlate with a higher incidence of severe or lethal violence.^21,25,26 Our study suggests that an important proportion of femicide victims who do not have a history of physical abuse may still have experienced other features of intimate partner abuse such as controlling behavior. Similarly, although we did not assess how common controlling behaviors are in relationships that are not otherwise abusive, the pervasiveness of control issues in the survivors’ stories lends support to the notion that clinicians who are trying to evaluate if a patient is experiencing intimate partner violence may gain important insights by asking questions about control.

In the quantitative analysis, the victims of homicide and attempted homicide had significantly higher DA scores than the abused controls in the same cities,²⁷ confirming that these individual factors are associated with an increased risk of femicide. However, our study challenges the notion of femicide being the consequence of the most severe abuse only, and makes one consider the lethality risk of women anywhere along the spectrum of abusive relationships. At the severe end, where anyone can see the risk, women most need help accessing resources. Clinicians or other members of the health care team may need to play an active role in calling the police, filing orders of protection, or otherwise helping women connect to resources. At the low end of the abuse spectrum, it may be impossible to detect the increased risk, and thus survivors of femicide attempts must not be made to feel that they should have been able to predict the event. The majority of women, however, lie somewhere in the middle of this spectrum, with some but not all the classic risk factors for femicide, and at least a theoretical chance that their risk can be modified.

Approximately half the time, both in our study and in the larger case-control study,¹⁷ the victims did not suspect that their lives were in danger. Though a clinician should never discount a woman's fear of femicide or serious injury if she is afraid,²⁸ it is important to recognize that one cannot necessarily be reassured by the patient's own sense of safety. One must question why women had not perceived themselves to be at risk. Our qualitative data analysis suggests that for some, the violence they were experiencing may have been perceived as a “normal” part of life. Others seemed to be more focused on relationship problems aside from violence—in particular the partner's alcohol or drug abuse, his financial problems, or his infidelity. These women were often trying to help him or force him to change, and did not necessarily see themselves as frightened victims.

Similarly, we found that in the majority of cases, the attack occurred when the woman was trying to leave the relationship—a time that is known to carry an increased risk of femicide.¹⁶ These results support the notion that clinicians should not push victims of violence to leave their relationships prior to having dealt with issues of safety, as ending the relationship may potentially increase the risk. Additionally, however, we noted that the victim was often trying to leave for reasons other than violence. This finding is worrisome as it suggests that at the time of the relationship change, many women may not be thinking of their problems in terms of domestic violence. It is possible that these women may be less inclined to get help from domestic violence agencies about safety planning than those who are leaving because of the violence.

Our study suggests that all victims of intimate partner violence, not just those who fit a classic picture of severe abuse or those who seek out domestic violence resources, should be educated about the risk of femicide. Women with physically abusive or highly controlling partners need to know that the risk is heightened around the time of relationship change, even if the main issues prompting the break-up are not related to domestic violence. Efforts that are targeted only toward those women who are seeking help for violence-related problems may miss an important proportion of potential victims. Our qualitative study cannot make any conclusions as to the utility of risk factor assessments, the benefit of routine domestic violence screening, or the effectiveness of interventions to reduce the risk of femicide. Our findings, however, highlight the complexity in the lives of women experiencing a femicide attempt, and warn against thinking of intimate partner femicide as the consequence of only the most severe violence. Further research needs to address clinical strategies to predict risk of femicide and the efficacy of interventions to decrease the risk of femicide once a history of intimate partner violence or controlling behavior has been uncovered.