Abstract
Drawing on the Women’s Health Effects Study, a community sample of women (N = 309) who recently left an abusive partner, this study examines patterns of cumulative abuse experiences over the life course, their socioeconomic correlates, and associations with a range of health outcomes. Latent class analysis identified four groups of women with differing cumulative abuse profiles: Intimate Partner Violence (IPV) Dominant, Child Abuse and IPV, All Forms, and All Forms Extreme. We find a relationship pattern between cumulative abuse and socioeconomic circumstances, and significantly worse health outcomes among women with the All Forms Extreme profile. Implications for research and practice are discussed.
Intimate partner violence (IPV), defined as a pattern of physical, sexual, and/or emotional violence by an intimate partner in the context of coercive control (Tjaden & Thoennes, 2000), involves a broad range of experiences that are best understood within a contextualized framework that takes into account the conditions of women’s lives (Ford-Gilboe et al., 2009). In this regard, a significant body of research confirms the pervasive and long-term negative health consequences of IPV (J. Campbell, 2002; Scott-Storey, 2011) with recent studies more closely examining why variation exists among women in terms of health outcomes (Ansara & Hindin, 2011). From these studies, we know that experiencing violence earlier in one’s life is a risk factor for later victimization (Briere & Jordan, 2004; R. Campbell, Greeson, & Bybee, 2008), and women rarely experience only one type of abuse (Scott-Storey, 2011; see also Cavanaugh et al., 2012; Dutton, Kaltman, Goodman, Weinfurt, & Vankos, 2005). It is now clear that a thorough understanding of women’s health must include attention to their cumulative abuse experiences. To only assess current or most recent violence misses important earlier experiences that matter for current health outcomes, and controlling for other types of abuse does not adequately account for the interactive effects of multiple forms of victimization. As R. Campbell et al. (2008) note, “How we study violence must become more congruent with how women actually experience it” (p. 196; see also Briere & Jordan, 2004; Swartout, Swartout, & White, 2011).
Until recently, most researchers focusing on violence over the life course have assumed a variable-centered approach. According to Nurius and Macy (2010), “Variable-oriented methods are based on the proposition that samples are homogeneous, or at least sufficiently similar, to allow collected research to be generalized across populations” (p. 1096). For example, analyses are conducted to predict variation in health outcomes by drawing on multiple independent variables that measure dimensions of violence at different points in both childhood and adulthood (Swartout et al., 2011). Among these studies, tremendous variation exists regarding the types of abuse examined, the measures used, and the samples studied. Nonetheless, we can conclude from this body of research that experiencing more than one type of abuse, either concurrently or at different points across the life course, is associated with poorer health (Scott-Storey, 2011), and this may operate as a dose-response effect (Schneider, Baumrind, & Kimerling, 2007; Scott-Storey, 2011).
Although important, variable-centered methods are less able to contextualize individual experiences in a way that captures the dynamics of women’s lives. For example, based on what we know, it is reasonable to expect that within a community sample, women who have recently left an abusive partner, those who are currently experiencing IPV, those whose lives have always been violence free, and those with violence experiences in the distant past will have different health profiles, especially if violence experienced in childhood is considered. A promising development in the cumulative abuse literature is the shift from a variable-centered to a person-centered approach (Nurius & Macy, 2010). In comparison with a variable-centered approach, which focuses on variation among independent variables to draw conclusions about relationships within a sample, a person-centered approach focuses on identifying subgroups of women with similar violence profiles. It is argued that this type of analysis offers a more realistic assessment of social life because it acknowledges and assesses heterogeneity in terms of the lives of different women within a sample. As Swartout et al. (2011) state, “In reality there may be different types of people within a given population” (p. 311).
Studies are beginning to use person-centered techniques to assess the health consequences of IPV (e.g., Ansara & Hindin, 2010; R. Campbell et al., 2008; Carbone-Lopez, Kruttschnitt, & Macmillan, 2006; Cavanaugh et al., 2012; Dutton et al., 2005; Lindhorst & Beadnell, 2011). Greater attention to the variation in constellations of abusive experiences among women is warranted because, first, variation exists, and, second, these differences matter in terms of health consequences (R. Campbell et al., 2008; Lindhorst & Beadnell, 2011). Using a person-centered methodology allows us to identify subgroups of women with similar cumulative profiles based on the types of abuse experienced, their cumulative nature, and relevant abuse characteristics, such as severity and duration. In this article, we contribute to this emerging body of research by using a person-centered approach to identify lifetime patterns of interpersonal violence among women in Canada who have recently left their abusive partner. This is important because misconceptions around IPV and stereotypes of “abused women” persist. The absence of accurate and contextual understandings of IPV in public discourse infiltrates policies and services, and it undermines women’s economic, social, and health recovery. Thus, in addition to considering women’s current abuse experiences in the contexts of their lifetimes, it is necessary to examine variation by economic circumstances and experiences and to document physical and mental health outcomes. Such descriptions contribute to understanding gender-based violence as a systemic problem, moving us further away from the dominant individualist discourse (Humphreys & Thiara, 2003).
Profiles of Interpersonal Violence Across Different Sample Populations
To date, some of our knowledge about patterns of interpersonal violence is based on a handful of studies drawn from community, shelter, and population-based samples of women who have experienced, or are at high risk of experiencing, interpersonal violence (Ansara & Hindin, 2011; Dutton et al., 2005; Glass et al., 2009; Lindhorst & Beadnell, 2011). Despite the fact that these studies vary considerably in terms of methodology, especially regarding sample characteristics and measures, we see real differences among survivors of IPV based on patterns of partners’ abusive behaviors (Glass et al., 2009) and patterns of exposure to different types of, and severity of, IPV (Ansara & Hindin, 2011; Dutton et al., 2005; Lindhorst & Beadnell, 2011). Findings suggest that differences associated with race/ethnicity (Dutton et al., 2005; Glass et al., 2009), education (Lindhorst & Beadnell, 2011), and employment (Dutton et al., 2005; see also MacMillan & Gartner, 1999) are associated with different profiles of abuse. More consistent findings relate to health outcomes. Indeed, the highest levels of depression (Ansara & Hindin, 2011; Dutton et al., 2005; Lindhorst & Beadnell, 2011), anxiety (Ansara & Hindin, 2011; Lindhorst & Beadnell, 2011), and posttraumatic stress disorder (PTSD; Dutton et al., 2005) have been found among those with more “severe” profiles, with those having moderate profiles showing poorer health than the low exposure groups (Ansara & Hindin, 2011; Dutton et al., 2005; Lindhorst & Beadnell, 2011).
A second group of studies extends this work by including lifetime indicators of violence, thereby contributing to our understanding of women’s cumulative abuse experiences and the effects on their health. Pimlott-Kubiak and Cortina (2003) utilize the U.S. National Violence Against Women Survey to examine lifetime “aggression exposure histories” and their relationships to health outcomes among 12,273 women and men. In addition to those with no reports of victimization, seven subgroups emerged with differing lifetime profiles: (a) Child Physical Abuse, (b) Adult Physical and Emotional Abuse, (c) Severe Sexual Violence, (d) Multiple Victimization without Sexual Violence, (e) Adult Stalking, (f) Multiple Victimization with Sexual Violence, and (g) Minimal Abuse History (Pimlott-Kubiak & Cortina, 2003). In the examination of health outcomes associated with the cumulative abuse profiles, each group was compared with those reporting no victimization, controlling for covariates. For all groups, compared with those with no abuse, significantly higher levels of depression and physical health (except for Child Physical Abuse Subgroup) were found, with the most elevated levels noted among the multiple victimization profiles (Pimlott-Kubiak & Cortina, 2003).
In comparison, R. Campbell et al. (2008) examined cumulative abuse profiles among a much smaller sample of mostly African American women (n = 298) who were veterans or reservists in the United States military. Cumulative abuse measures assessed childhood sexual abuse, adult sexual assault, IPV, and workplace sexual harassment. Four subgroups were identified: (a) Low All, (b) High All, (c) Sexual Revictimization With Harassment, and (d) IPV With Harassment (R. Campbell et al., 2008). No differences were found among the groups of women by age, race/ethnicity, education, or employment status; however, significantly lower household income was noted for women with “High All” cumulative experiences. Three of the key findings were as follows: (a) those with “High All” profiles had the highest level of PTSD; (b) experiences of multiple types of violence undermined health, although sexual violence had a particularly negative effect; and (c) PTSD symptomatology mediated the relationship between violence and physical health (R. Campbell et al., 2008).
Most recently, Cavanaugh et al. (2012) drew on a sample of 1,424 female nurses, almost three quarters of whom were Caucasian, to investigate lifetime patterns of child abuse, IPV, and workplace violence (WPV) on mental health outcomes. Four groups of women were identified based on patterns of lifetime violence: (a) Low All, (b) IPV, (c) WPV, and (d) Childhood Abuse. Women in the “IPV” group were more likely to be older, mothers, and single compared with those in the “Low All” group. Those in the “Low All” group were more likely to be White than those in the “IPV” and “WPV” groups. Women in the “IPV” and “Childhood Abuse” groups were more likely to screen positive for depression, whereas those with the “IPV” profiles were also more likely to screen positive for PTSD than those with the “Low All” profile. Economic circumstances were not examined (Cavanaugh et al., 2012).
In summary, the following conclusions can be drawn from studies that have examined the cumulative effects of violence. Even though the findings are based on very different samples both in size and composition (e.g., national survey and selective occupation-based samples), subgroups of women with some commonalities have been identified. In addition to women with minimal and/or no exposure are those with “context specific violence,” for example, “Workplace Violence” (Cavanaugh et al., 2012), “IPV” and “Childhood Abuse” (Cavanaugh et al., 2012; Pimlott-Kubiak & Cortina, 2003), and “Sexual Abuse and Assault” and “Stalking” (Pimlott-Kubiak & Cortina, 2003). There is also evidence of “violence across contexts” (Cavanaugh et al., 2012) in two studies (R. Campbell et al., 2008; Pimlott-Kubiak & Cortina, 2003), for example, women with “High All” profiles and combination profiles, such as “Sexual Revictimization With Harassment” (R. Campbell et al., 2008).
This research adds to IPV studies by more accurately capturing the lifetime exposure of abuse among different groups of women. The method of person-centered analysis tends to group women based on indicators of severity and abuse types. Although different health outcomes have been assessed in various studies, they generally tend to differ across groups as expected, with worse health associated with more extreme violence profiles. Context-specific violence may also have unique health consequences, although more research is needed to establish consistent patterns. Finally, associations have been found between social and economic factors and degrees of cumulative abuse, although, again, more evidence is needed before conclusions can confidently be drawn.
In this article, we draw on data from a community sample of 309 Canadian women who participated in Wave 1 of the longitudinal Women’s Health Effects Study (WHES) (see Ford-Gilboe et al., 2009) to describe their cumulative abuse profiles and to examine whether health, social, and economic conditions vary across these profiles. We employ a person-centered approach, using Latent Class analysis (LCA), to identify subgroups of women based on their exposure to interpersonal violence throughout their lives. We add to and extend the literature by including types, severity, and duration of childhood abuse and IPV. Specifically, our measures assess violence in childhood (sexual, physical, and emotional), sexual assault during adulthood, physical and non-physical IPV severity from their most recent partner, duration (childhood and adulthood) of victimization, and lifetime number of abusive partners. Our nuanced snapshot of cumulative abuse profiles for this community sample of women who have recently left their abusive partner provides detailed information to augment our understanding of the relationships among their cumulative abuse patterns, socioeconomic circumstances, and health.
Method
Women who had separated from their abusive partner in the preceding 3 years were recruited from three Canadian provinces using advertisements placed in community settings (e.g., libraries, community centers), service agencies (e.g., shelters, health clinics), and through media coverage. Women contacted the research team by telephone or email and were screened for eligibility. Exposure to IPV was confirmed using a modified, four-item version of the Abuse Assessment Screen (AAS; Parker & McFarlane, 1991). Eligible women were provided with a verbal description of the study and invited to participate.
The mean age of participants was 39.4 years (SD = 9.80, range = 19-63), and women had completed an average of 13.4 years of education (SD = 2.60, range = 6-22 years). Although almost half (45%, n = 139) of the participants were employed, the vast majority (90%) reported having difficulty living on their current income. Annual income ranged from CAD$0 to CAD$95,000 per year, with a mean income of CAD$20,391 (SD = 17,145, median = CAD$15,684). Slightly more than half (56.3%, n = 174) were parenting at least one child below the age of 18 years. Although the majority of women were Caucasian, 16.8% (n = 51) identified themselves as being members of a visible minority group and 7.4% (n = 23) identified as Aboriginal.
Data were collected using structured interviews focusing on resources, service use, and demographic characteristics; in-depth abuse histories and health assessments were conducted by Registered Nurses. A detailed safety protocol was used in all interactions. Women were offered a participation fee of CAD$30 for each session completed and reimbursed for childcare and transportation costs. Ethics approval was obtained from the Research Ethics Board at each study site prior to recruitment, and written consent was obtained from each woman prior to data collection.
Measurement of Cumulative Abuse Profiles
The variables selected for inclusion into the LCAs were chosen based on theoretical and substantive rationale, as well as availability within our data set. Our goal was to identify indicators of lifetime abuse that captured dominant types of interpersonal violence experienced beginning in childhood, and we also included measures of severity and duration, to better capture the complexities that exist across women’s lives.
Childhood Abuse
Severity of child abuse
We used three of the Childhood Trauma Questionnaire (CTQ) scales to assess severity of physical, sexual, and emotional abuse in childhood. The CTQ was created to capture histories of childhood abuse and neglect (Bernstein & Fink, 1998). Scores on each scale range from 5 to 25, which represent the sum of five items, each measured on a 5-point Likert-type scale (from “never true” to “very often true”). Therefore, higher scores reflect more severe abuse, with the following interpretations recommended: 5-8, “none–minimal”; 9-12, “low–moderate”; 13-15, “moderate–severe”; and 16+, “severe–extreme.”
Duration of child abuse
Number of years abused as a child is a derived continuous variable ranging from 0 to 16 years based on women’s self-report to questions that assessed whether she believed that she was abused as a child up until the age of 16 (no = 0, yes = 1), how old she was when the abuse started, and how long the abuse lasted (in years).
Adult Sexual Assault
Exposure to adult sexual assault was assessed with a one-item variable (no = 0, yes = 1), based on the question, “Not including your relationship with your index partner, have you ever been sexually assaulted as an adult? By sexual assault, I mean did you ever have sexual contact against your will?”
Severity of Former Partner Violence
Power and control
How women experienced violence by their former partner was assessed with the Women’s Experiences of Battering Scale (Smith, Earp, & DeVellis, 1995). Smith, Edwards, and DeVellis (1998) describe the scale as measuring “the psychological experience of being battered,” which refers to “experiences [of] vulnerability, loss of power and control, and entrapment as a consequence of the other member’s exercise of power through the patterned use of physical, sexual, psychological and/or moral force” (Smith, Thornton, DeVellis, Earp, & Coker, 2002, p. 1210). The 10-item summated rating scale asked participants to rate the extent of their agreement with 10 items describing abuse experienced from their ex-partner while in a relationship with him on a 6-point Likert-type scale, ranging from “agree strongly” to “disagree strongly.” Sample items include the following: “I feel ashamed of the things he does to me” and “I hide the truth from others because I am afraid not to.” Items were reverse-scored and summed to produce a scale ranging from 10 to 60. Internal consistency was .82. Higher scores indicate greater severity of abuse.
Physical and non-physical intimate partner violence
The severity and magnitude of physical and non-physical abuse was assessed with the 30-item Index of Spouse Abuse (ISA) (Hudson & McIntosh, 1981). The items capture frequency of abusive acts, ranging from 0 (“never”) to 4 (“very frequently”). The “physical abuse” total score was derived from 11 items related to physical or sexual abuse that were weighted for severity, with the summed weighted scores ranging from 0 to 100. Similarly, the “non-physical abuse” total score was derived from 19 items, also weighted for severity and summed, ranging from 0 to 100 (Hudson & McIntosh, 1981). Validity and internal consistency is excellent across a variety of studies of abused and non-abused women, as well as other subpopulations (J. Campbell et al., 1999; Cook, Conrad, Bender, & Kaslow, 2003; Woods, 2000). Internal consistencies of the “physical” and “non-physical” abuse items in this study were .84 and .83, respectively.
Duration of Abuse From Most Recent Partner
Duration of abuse is a derived variable based on questions that asked the participant when the abuse from her former partner began, how long it lasted, and whether it was still ongoing. Responses range from 0.25 to 37 years.
Number of Abusive Partners
In addition to length of abuse from her most recent partner, we included a measure of the number of abusive partners in the woman’s lifetime, based on the question, “Besides your relationship with (former partner), how many other abusive relationships with partners have you had?” Because this item did not include the most recent partner, the total was increased by 1. Responses range from 1 to 19; however, given that few women reported more than five partners, responses from 5 to 19 were collapsed into the Category 5 or more.
Socioeconomic Factors
Education
Education is a continuous measure of number of years completed; it ranges from 6 to 22 years.
Employment status
Employment status is a single indicator (not employed = 0; employed = 1) collapsed from a variable that asks participants about their current work status. The employed category includes women working in both full-time and part-time jobs.
Personal income
Personal income is a self-report measure based on participants’ estimation of their total personal income from all sources before taxes for the current calendar year. It is a continuous variable ranging from CAD$0 to CAD$95,000.
Social assistance/disability insurance
Economic circumstances were also captured by including a measure that assessed whether the participants were currently receiving social assistance or disability insurance (no = 0; yes = 1).
Health Outcomes
To capture a range of health outcomes, we included measures of “chronic pain,” “depressive symptoms,” “PTSD symptoms,” and “overall physical health.”
Chronic pain
The measure of “chronic pain” is based on a single item from the Chronic Pain Grade Scale (Von Korff, Ormel, Keefe, & Dworkin, 1992). Respondents were asked how many days in the last 6 months they were kept from their usual activities (work, school, or housework) because of pain. Responses vary from 0 to 180 days. This indicator was selected because it captures the degree to which pain interferes with women’s lives, rather than simply the intensity of pain.
Depressive symptoms
We used the Centre for Epidemiologic Studies Depression Scale (CESD), the most commonly used measure of depressive symptoms in studies of IPV, with well-established reliability and validity. It is a 20-item summated rating scale capturing depressive symptomatology in the past week (Radloff, 1977). Responses vary from 0 (“rarely” or “none of the time”) to 3 (“most of the time”); the summed score ranges from 0 to 60. The CESD has established reliability and validity with the general population and with abused women. In this study, Cronbach’s alpha was .93.
PTSD symptoms
The Davidson Trauma Scale (DTS) first asks participants to identify whether or not they have experienced a traumatic event (Davidson, 1996; Davidson et al., 1997). Affirmative responses are followed by a 17-item self-report summated rating scale that assesses the frequency and severity of each symptom in the previous week; the scale ranges from 0 to 136. The DTS captures symptoms of hyper-vigilance, intrusion, and avoidance/numbing. The measure has established reliability and validity in a variety of trauma populations; in this study, Cronbach’s alpha was .95.
Physical health
Physical health was measured using the well-established Physical Health Component Score on the Short-Form Health Survey Version 2 (SF-12v2; Ware, Kosinki, Turner-Bowker, & Gandek, 2002). The SF-12v2, which incorporates improvements in Version 2 of the SF-36 Health Survey, is a shorter measure of health status designed to better suit longer surveys, while maintaining reliability and validity (Ware, Kosinski, & Keller, 1996). The Physical Health Component Score incorporates four dimensions of health functioning: physical functioning, bodily pain, general health, and physical role performance. Each item is measured using a Likert-type scale to assess perceived health over the previous 4 weeks. Scores range from 14.43 to 68.71.
Analytic Design
LCA, conducted with Latent Gold software (Vermunt & Magidson, 2005), was used to identify subgroups of women with different abuse profiles. Because we are interested in associations that result from an unobserved, latent source of variation, traditional regression models, which include parameters that describe associations between observed variables, are not appropriate. In this analysis, the observed abuse variables cannot measure women’s abuse histories, directly or without error. LCA makes it possible to detect unobservable characteristics and statistically test a latent structure that explains the data (Vermunt & Magidson, 2005). Because the latent variable in LCA is categorical, LCA differs from more traditional latent variable approaches, such as factor analysis, that are based on continuous latent variables. In addition, factor analysis, regression analysis, and structural equation modeling (SEM) take a variable-centered approach; in other words, the focus is on relationships among variables with the goal of predicting outcomes. Some research questions, such as ours, require a person-centered approach, in which the focus is on relationships among individuals, with the goal of grouping “individuals into categories of individuals who are similar to each other and different from individuals in other categories” (Muthén & Muthén, 2000, p. 882).
Thus, LCA uses maximum-likelihood methods to classify individuals into mutually exclusive and exhaustive latent classes based on differences in a set of indicators (McCutcheon, 1987). Indicators of mixed-scale types may be included. Cases are assigned to the class for which the posterior probability is highest; that is, cases are classified into clusters based upon membership probabilities estimated directly from the model (Dayton, 1998; Vermunt & Magidson, 2005). Because of its ability to detect associations among variables due to an unmeasured, latent source of variation, using LCA enables us to identify groups of individuals who differ in their experience of abuse over the life course.
We first estimated a one-cluster model, which tested the assumption that all respondents share similar abuse histories, compared with two-, three-, and four-cluster models. Various diagnostics, including the Bayes’ Information Criterion (BIC) and log-likelihood statistic (LL), were used to determine the optimal number of clusters. Based on these statistics, we concluded that a four-cluster model best fits the data. Classification statistics from the four-cluster model indicated that 96% of the cases were classified correctly.
Results
Table 1 presents results from the LCA that identified four groups of women with differential experiences of lifetime abuse and shows conditional probabilities/means indicating the prevalence, or level, of each indicator within each latent class. Our LCA includes a control for age in the first survey cycle, as age may condition the likelihood of experiencing several of the chosen indicators. Table 1 includes the average age of women in each subgroup.
Percentages and Means of the Profile Variables.
Note. IPV = intimate partner violence.
Four groups of women with different profiles of violence exposure and severity across the life course were identified: Profile 1, “IPV Dominant”; Profile 2, “Child Abuse & IPV”; Profile 3, “All Forms”; and Profile 4, “All Forms Extreme.” Profile 1, “IPV Dominant,” consists of 87 women for whom IPV is their dominant exposure to violence. Most have little experience of child abuse. In adulthood, 22% report experiencing sexual assault (sexual contact against their will, not including most recent partner). Average scores on measures of partner violence severity reflect coercive control and moderate levels of physical and non-physical IPV from their former partner. Women with this profile had been with their most recent abusive partner approximately 9 years and reported one to two abusive partners in their lifetime. The three remaining groups include women whose abuse profiles began in childhood.
Profile 2, “Child Abuse and IPV,” includes, on average, the youngest women in our sample. Their childhoods are marked by abuse, with moderate levels of emotional and sexual abuse and low levels of physical abuse. On average, the duration of child abuse is 7.6 years. Sixteen percent report sexual assault in adulthood. The women in this subgroup report having had only one relationship with an abusive partner; however, this was long-lasting, averaging almost 11 years. Indicators of IPV reveal mean levels of non-physical and physical violence that are somewhat higher than the women in Profile 1.
The 90 women in Profile 3, “All Forms,” report severe emotional and sexual child abuse and moderate physical child abuse, compared with the other groups. Fifty-five percent experienced sexual assault in adulthood. Average levels of IPV severity from their most recent partner are the lowest compared with the three other profiles. Similarly, the IPV duration is the shortest, on average just over 5 years. Furthermore, the women in this group reported having the most abusive partners, on average just over 3.
Those in the subgroup Profile 4, “All Forms Extreme,” report the highest mean scores across the four groups in all indicators except for number of abusive partners and number of years abused. Severe levels on each dimension of child abuse are revealed, and over 60% of women in this group report experiencing sexual assault in adulthood. With respect to the level of partner abuse, reported scores are the highest among the four profiles. Finally, women in this group had, on average, just under three abusive partners over their lifetime, with the length of abuse from the most recent partner lasting 9.5 years, on average.
Finally, in the last row of Table 1, we see the covariate results for age. On average, the youngest women in the sample are those in the subgroups of “Child Abuse and IPV” (M = 36.43) and “IPV Dominant” (M = 38.78) and report having one, or slightly more, abusive partner. In comparison, the women in Profiles 3 and 4 are slightly older (M = 41.14 and 40.31, respectively) and report, on average, three abusive partners.
Table 2 reveals the means of the adult socioeconomic indicators by the cumulative abuse patterns. The four profiles are ordered along a continuum of cumulative abuse with the fourth being the most extreme. While there are no differences among the four groups of women with respect to education, significant differences emerge for all the remaining socioeconomic status indicators. Generally, there is a relationship pattern between cumulative abuse and socioeconomic status; as abuse profiles become more extreme, current socioeconomic resources decline. Specifically, 41% of the women with the “IPV Dominant” profile are not employed, compared with 71% of those with the “All Forms Extreme” profile (p < .05). The average personal income for women with the “IPV Dominant” profile is CAD$25,610.67, significantly higher than the income for women in the “All Forms” (CAD$18,191.32) and “All Forms Extreme” (CAD$15,278.33) groups (p < .05). Similarly, the percentage of women in each group who were receiving social assistance or disability benefits is significantly lower among the “IPV Dominant” group (24%), compared with women in the “All Forms” (44%) and “All Forms Extreme” groups (58%; p < .05).
Mean and Percentage Differences of Economic Status Indicators by Cumulative Abuse Profiles.
Note. IPV = intimate partner violence.
Significantly different than Profile 1 : IPV Dominant, p < .05.
Significantly different than Profile 3 : All Forms, p < .05.
Significantly different than Profile 4: All Forms Extreme, p < .05.
Table 3 assesses the relationships among the cumulative abuse profiles and various health outcomes. For all four health indicators (chronic pain, depressive symptoms, PTSD symptoms, physical health), women with the “All Forms Extreme” profile have significantly worse health, compared with each of the other groups of women.
Mean Differences of Health Outcomes by Cumulative Abuse Profiles.
Note. IPV = intimate partner violence; PTSD = posttraumatic stress disorder.
Significantly different than Profile 1: IPV Dominant, p < .05.
Significantly different than Profile 2: Child Abuse and IPV, p < .05.
Significantly different than Profile 3: All Forms, p < .05.
Significantly different than Profile 4 All Forms Extreme, p < .05.
Discussion
This study is based on a community sample of women who have recently left their abusive partners. The challenges that women experience throughout the process of leaving, including continued, escalated violence; poor health; systemic barriers; and economic difficulties are now well established (Davies, Ford-Gilboe, & Hammerton, 2008; Ford-Gilboe, Wuest, & Merritt-Gray, 2005; Humphreys & Thiara, 2003). However, research examining the effects of violence on women’s lives is still often limited to their most recent partner experience, implying that this is their only experience of interpersonal abuse. The results of this study highlight the merits of utilizing the person-centered approach to examine cumulative abuse because doing so identifies lifetime abuse profiles within our sample, thereby revealing distinct groups of women based on exposure to violence. This gives us a more nuanced and comprehensive picture by uncovering patterns of abuse the women have experienced (R. Campbell et al., 2008), examining the association between patterns of abuse and socioeconomic circumstances, and assessing whether the identified patterns are predictive of poor health.
In this way, stereotypes that serve either to homogenize or to marginalize survivors of abuse are challenged, and patterns of lifetime violence exposure associated with increased likelihood of health problems are exposed. For example, the stereotypes of the heroine escaping a single abusive partner who has disrupted an otherwise idyllic life, and the woman whose life is fraught from birth by multiple forms of violence and multiple abusive partners until she knows no other possibilities, reflect real-life experiences. But as stereotypes are in danger of doing, they draw attention away from violence as systemic and gender-based. In other words, stereotypes frame violence against women as an individual issue when, in reality, all women, by virtue of being female, are at risk of IPV. However, the magnitude of this risk varies depending upon socioeconomic status, and socioeconomic status can constrain or enable the leaving process (Barrett & St. Pierre, 2011). All of these factors are relevant for understanding health consequences of IPV. Demonstrating the associations between cumulative abuse profiles, socioeconomic status indicators, and health outcomes, therefore, helps to clarify these systemic and gendered links.
Our findings draw attention to the considerable variation in abuse histories among this community sample of women who have recently left an abusive partner. The relatively small percentage of women in the “IPV Dominant” profile (28%) is inconsistent with the dominant societal and media discourses that continue to position IPV as an isolated event in women’s lives, rather than part of a pattern of gender-based victimization. Similarly, R. Campbell et al. (2008) find that only 26% of their primarily African American veteran sample report IPV-dominant abuse experiences. In both studies, the remaining 70% or so of respondents experienced cumulative abuse across three patterns, reflecting the fact that “violence tends to co-occur” and considerable variation exists among these co-occurrences (Cavanaugh et al., 2012, p. 1).
Second, when compared with the other three groups identified, we find that the women with the “All Forms” profile had the lowest IPV scores. However, when other forms of abuse are considered, these women have experienced considerable cumulative lifetime abuse. R. Campbell et al. (2008) similarly found that the abuse profile of “Sexual Revictimization With Harassment” was characterized by low levels of IPV and high levels of other types of abuse (child sexual abuse, adult sexual abuse, and sexual harassment). Together, these findings suggest a limitation of using only IPV scores in any analysis as it may underestimate women’s exposure to violence, weakening its predictive power.
To our knowledge, our analysis is the first to include the number of lifetime abusive partners as a cumulative abuse measure. The variation found across the three groups, with younger women reporting fewer partners on average than their older counterparts, helps to make visible the systematization of abuse, because it suggests that women do face difficulties disentangling from the IPV in their lives (see also Humphreys & Thiara, 2003). When we examine the associations between the profiles and the socioeconomic indicators, the relationship between economic vulnerability and greater cumulative abuse becomes clearer, reinforcing knowledge that some women are lacking both non-abusive men in their worlds and the economic resources necessary to get away from the IPV in their lives (Jewkes, 2002). Indeed, we see patterned differences among the profiles in most of the socioeconomic indicators, with higher cumulative abuse generally being associated with lower resources (with significant differences between least and most cumulative abuse profiles). Yet, across the subgroups, no group is distinguished by high average levels of economic resources. In other words, economic disadvantage may be worse among those with more cumulative abuse, but it is not experienced only by women with the most cumulative abuse (see also R. Campbell et al., 2008). Further research is needed to disentangle the causal relationships among economic disadvantage, cumulative abuse, and health. By illustrating that patterns of abuse experiences do occur, and connecting these patterns to socioeconomic circumstances, our results help to move public and professional understanding beyond stereotypes and begin to unravel the interactions among abuse experiences across the life course, suggesting circumstances that increase risk of revictimization. The complex relationship between abuse and poverty must be analyzed beyond the stereotype that arises from understanding that economically disadvantaged women are more vulnerable to abuse, while overlooking the fact that abuse makes women economically vulnerable (see also Davies et al., 2008).
Our research adds to the limited number of studies that have used the person-centered approach to examine the health effects of cumulative abuse. As noted earlier, this research builds upon a large body of variable-centered research that supports the conclusion that prior violence increases risk of revictimization. By assessing lifetime interpersonal violence, we are able to improve our understanding of its health effects (Briere & Jordan, 2004; Nurius & Macy, 2010; Scott-Storey, 2011). Across all physical and mental health outcomes assessed, we find significantly worse health among the women with the “All Forms Extreme” profile compared with each of the other three groups. Despite methodological differences, our results are consistent with the conclusions drawn by Pimlott-Kubiak and Cortina (2003) and R. Campbell et al. (2008). Together with our finding that women with the “All Forms Extreme” profile also have significantly lower average incomes, higher rates of non-employment, and greater reliance on social assistance/disability than those with the “IPV Dominant” profile, we see a picture that connects economic disadvantage, extreme and multiple victimization, and poor health in multiple reinforcing ways (Briere & Jordan, 2004). This fits with research from the general population; for example, Honkalampi et al. (2005) find that childhood adversities increase depression in adulthood among women, and recovery is made more difficult by current stress and poor financial circumstances.
When interpreting our results, the following limitations should be considered. Our findings are at a bivariate, thus descriptive, level. They are only suggestive of relationships among cumulative abuse patterns, socioeconomic status, and health outcomes. Future research is needed to establish these connections with more sophisticated multivariate models. Second, these associations cannot speak to causality. All data were collected during the first wave of data collection. Third, the sexual assault variable included in our cumulative abuse profiles is a one-item indicator of sexual assault in adulthood; hence, it misses variability in sexual assault experiences. For example, it does not capture variation in the number of sexual assaults experienced or address the fact that the identities of the perpetrators vary; therefore, compared with the other indicators of abuse included, it is less informative. Finally, we cannot generalize the four cumulative abuse patterns identified here to all survivors of IPV. It is important, therefore, that these profiles not be reified. Their relevance to policy is in shifting the framework that views abuse as a discrete event that happens without context, toward a more contextualized view of diversity among women’s cumulative abuse experiences and associated health outcomes.
Nevertheless, our results do convey the importance of inquiring about women’s violence profiles, particularly for those in need of health care. Current health reflects an accumulation of experiences that must be incorporated into recovery plans and health care interventions (J. Campbell, 2002). Health must be understood as a process, as well as static; it is evolving and, therefore, has consequences over a lifetime (George, 2003). Social policies that recognize diversity among women in terms of cumulative abuse profiles refocus our attention on reducing interpersonal violence in childhood and adolescence as well as in adulthood. Furthermore, incorporating awareness of lifetime abuse experiences when creating interventions—for example, advocacy intervention—would mean designing empowerment training that is sensitive to the impact of, and variations among, women’s abuse histories. This would improve recognition and understanding of the diverse needs that exist within any population of survivors. Being sensitive to abuse histories and simultaneously targeting economic, social, and health improvements will likely have the biggest health benefit (currently and across generations) because empowering women in multiple domains, as others have also argued, reduces the likelihood of further victimization across the life course, individually and collectively.
Footnotes
Declaration of Conflicting Interests
The author(s) declared no potential conflicts of interest with respect to the research, authorship, and/or publication of this article.
Funding
The author(s) disclosed receipt of the following financial support for the research, authorship, and/or publication of this article: The authors acknowledge funding from the Canadian Institutes of Health Research (CIHR) New Emerging Team Grant #106054 and the Institute of Gender and Health Operating Grant #15156.