Abstract
This study examined use of safety strategies, experience of violence, and perception of danger from intimate partner violence (IPV) among 197 women seeking temporary protective orders against their abusive partners/ex-partners. Latent class analysis was used to group women into classes based on their use of safety strategies. Five classes of strategy use were identified: two high-activity classes, two moderately active classes, and one low-activity class. More severe abuse, increased perception of danger, and unemployment were associated with being in the higher activity classes. More effective interventions and outreach tools are needed to help women in IPV situations.
Intimate partner violence (IPV) is a significant public health problem that represents a serious threat to women’s physical health, mental health, and safety. According to a national survey, in the United States, more than one in three women (35.6%) have experienced rape, physical violence, and/or stalking by an intimate partner in their lifetime, and roughly one in four women (24.3%) have experienced severe physical violence (defined as hit with a fist or something hard, beaten, slammed against something) by an intimate partner in their lifetime (Black et al., 2011). Often, the violence women experience is chronic in nature. In the United States, two thirds (65.5%) of women who reported being physically abused by an intimate partner and half (51.2%) of women who reported being raped by an intimate partner said they were victimized multiple times by that same partner (Langen & Innes, 1986).
Women experiencing violence in their intimate relationships use various safety strategies to protect themselves and their children (Goodkind, Sullivan, & Bybee, 2004; Goodman, Dutton, Vankos, & Weinfurt, 2005). These safety strategies can be broadly categorized as legal (e.g., obtaining a protective order), formal network (e.g., going to counseling), informal network (e.g., talking to friends and family), placating the abuser (e.g., trying not to cry during the violence), resisting the abuser (e.g., fighting back physically), and safety planning (e.g., keeping money or other valuables hidden; Goodman, Dutton, Weinfurt, & Cook, 2003). Safety planning is a term inclusive of these safety strategies. It is a practice many domestic violence service agencies engage in when working with women in IPV situations to enhance their sense of autonomy and empowerment as well as their safety (Campbell, 2001).
Previous research has shown that women use many strategies over the course of an abusive relationship (Duterte et al., 2008; Flicker et al., 2011; Logan, Shannon, Cole, & Walker, 2006; Wiist & McFarlane, 1998). Yet, despite the rich literature suggesting that women are actively using safety strategies (Bonomi, Holt, Martin, & Thompson, 2006; Duterte et al., 2008; Flicker et al., 2011; Meyer, Wagner, & Dutton, 2010), little is known about the heterogeneity or patterns of strategy use. Prior research has generally examined safety strategy use by summing the number of different strategies women report using and then comparing the strategies used with a range of demographic covariates (e.g., race/ethnicity, income, employment) and outcomes, including abuse experienced, social support, and physical and mental health, among others (Davies, Block, & Campbell, 2007; Durfee & Messing, 2012; Sabina & Tindale, 2008; Wright & Johnson, 2009). This is a variable-oriented approach, where emphasis is placed on the relationship between variables, and only statements about variables (and not individuals) can be made (e.g., race is related to contacting police; Bogat, Levendosky, & Von Eye, 2005). Although this approach enhances our understanding of the range of strategies women use to stay safe, it does not allow for the examination of combinations of strategies used and whether those combinations are associated with the range of outcomes described above.
Two studies, that we were able to find, used cluster analysis and latent class analysis (LCA) to examine patterns of safety strategy use. In the first study, the authors used cluster analysis on data from interviews conducted with 160 abused women (Goodkind et al., 2004). Five clusters were identified: (a) high-activity cluster included women who endorsed a high level of all strategies assessed; (b) informal help-seeking cluster included women who were more likely to seek help from family and friends, while also placating the abuser; (c) going at it alone cluster included women who were the least likely to seek assistance from anyone, either formal or informal, and who were likely to use resistance strategies against the abuser; (d) placating cluster of women used several of the placating strategies and few resistance strategies; and (e) trying everything group included women who used a moderate amount of all the strategies to protect themselves (Goodkind et al., 2004). Women in the high-activity cluster experienced more physical violence, psychological violence, and injury than women in the other four clusters (Goodkind et al., 2004). In the second study, the author used data from the Violence and Threats of Violence Against Women and Men in the United States Survey (1994-1996), specifically 334 women who experienced IPV in the 12 months prior to the survey (Kaukinen, 2004). Using LCA, Kaukinen (2004) identified three classes of help seeking: (a) minimal or no help seeking, (b) family and friend help seeking, and (c) substantial help seeking (e.g., women sought help from family, friends, police, social service providers, and psychiatrists). Race was identified as a correlate of increased levels of help seeking, where White women sought more help in comparison with minority women (Kaukinen, 2004). The victim–offender relationship was another correlate; specifically, women victimized by their spouses were more likely to seek increased levels of help (Kaukinen, 2004).
We had an opportunity to add to this limited body of research by using data from a larger study of women’s IPV risks and experiences to identify patterns of strategy use using LCA, which is a person-oriented approach (Bogat et al., 2005). Unlike variable-oriented approaches, person-oriented methods presume that individuals are “unique and that their uniqueness is knowable” (Bogat et al., 2005, p. 50). The use of this method allows us to build upon the existing literature by identifying women’s patterns of safety strategy, and the negative outcomes associated with those patterns of safety strategy use. Specifically, we test the relationships among classes of safety strategies used, demographic covariates, and two outcomes: experience of violence and perception of danger from IPV. The expected heterogeneity in safety strategy use will lend support to the need for targeted and tailored interventions. Women using safety strategies are likely to be a heterogeneous population, and interventions may need to be tailored to different victim profiles (based on experience of violence, perception of danger). Person-oriented approaches, like LCA, may help public health practitioners and advocates developing and implementing interventions to understand which groups of women will be open to the different types of programs. Furthermore, a better understanding of patterns of safety strategy use may strengthen the measurement of women’s safety behaviors for future research.
To enhance our understanding of how women protect themselves from their abusive partners, more research examining patterns of safety strategy use is needed. The current study examined patterns of strategy use using 17 items from the Intimate Partner Violence Strategies Index (Goodman et al., 2003). Based on previous work (Flicker et al., 2011; Meyer et al., 2010; Wiist & McFarlane, 1998), it was hypothesized that women experiencing severe physical violence or severe sexual violence would be in higher activity safety strategy classes compared with women experiencing minor or no physical violence or no sexual violence. It was further hypothesized that women who perceived themselves to be at greater risk from IPV would be in higher activity safety strategy classes compared with women who perceived themselves to be at less risk from IPV. This relationship between perception of danger from IPV and safety strategy class has not been previously examined in the literature, but it is assumed that increased violence is highly correlated with increased perceptions of danger and therefore would operate in the same way in relationship to the safety strategy classes.
Method
Study Population
Data for this study came from a baseline interview with 197 women enrolled in the Brief Danger Assessment Prevention Intervention (BDAPI), a quasi-experimental study of women seeking temporary protective orders against their abusive male partners/ex-partners at a legal clinic run by a domestic violence service agency. The purpose of the study was to evaluate BDAPI, a standardized risk assessment and educational protocol for IPV victims. The study was conducted from 2005-2008, and interviews were carried out with eligible women who were 18 years of age or older and who had experienced some level of physical violence by a current or former intimate partner.
Recruitment and Interview Protocol
Women were recruited from a domestic violence service agency legal clinic, which helps victims of IPV obtain protective orders, peace orders, divorce decrees, custody of their children, and child support. Recruitment procedures varied depending on whether study personnel were at the legal clinics. When an interviewer was present, a woman would meet with the interviewer after she talked with clinic staff. The clinic staff informed the clients who were seeking temporary protective orders (TPO) that a university was conducting a study about women’s safety and then introduced the client to the study personnel. When an interviewer was not present, clinic staff told their clients that a university was conducting a study about women’s safety and they would ask their eligible clients whether it would be okay to provide their contact information (e.g., name, phone number, convenient and safe times to call, and other safety instructions) to the researchers. Interviews were conducted on-site, by telephone, or by mail. Questions were read aloud to participants and the interviews took between 30 and 45 min to complete. The Institutional Review Board at the authors’ institution approved all study procedures.
Measures
Safety strategies
Safety strategy use was assessed using a subset of items from the Intimate Partner Violence Strategies Index (Goodman et al., 2003). This analysis focuses on 17 specific safety strategies, which were dichotomous variables used to create the latent classes. Women reported whether they had used in the past 6 months various safety strategies from the following dimensions: safety planning (six items), placating (four items), police involvement (two items), leaving with informal network assistance (three items), and domestic violence agency outreach (two items). The internal consistency for the 17 safety strategies is acceptable (Kuder-Richardson Formula 20 [KR-20] = 0.7367).
IPV
The revised version of the Conflict Tactics Scale (CTS2-R) was used to measure violence by the women’s partners (Straus, Hamby, Boney-McCoy, & Sugarman, 1996). Women were asked to report physical violence (11-item subscale), injury (six-item subscale), and sexual violence (two-item subscale) ever experienced. If they have experienced violence or injury in their lifetime, they were then asked how often it occurred in the past 6 months with the following answer options: never happened, once, twice, 3-5 times, 6-10 times, 11-20 times, more than 20 times, and not in the past 6 months, but this has happened before. The CTS2-R is a commonly used, well-validated measure of IPV. It is highly reliable and valid (Straus et al., 1996). According to the scoring guidelines described by Straus et al. (1996) and consistent with prior research, the CTS2-R was scored based on prevalence of the abuse (Gielen, McDonnell, & O’Campo, 2002). Prevalence was the proportion of women reporting an experience of one or more acts of violence in each of the subscales in the past 6 months, and it was further classified as minor or severe.
Perception of danger from IPV
Perception of danger from IPV was measured using four items developed by D. W. Webster, P. Mahoney, and J.C. Campbell (personal communication, May 22, 2013). Women were asked to report how great the risk was that their partner, in the next year, would attempt to physically assault them, seriously physically injure them, try to kill them, or physically injure someone else whom they care about. Response categories ranged from 1-5, with 1 being low risk and 5 being high risk. A mean scale score for the four items was calculated; scores ranged from 1-5, with higher scores indicating higher perceived danger from IPV.
Sociodemographics variables
Self-reported sociodemographic variables were also included in the analyses: specifically, age, race, education, monthly income, receipt of food stamps, homeownership status, relationship to children in the household, and the woman’s relationship with the abuser.
Data Analysis
Descriptive statistics
The frequencies and percentages of categorical variables and the means and standard deviations (SDs) of continuous variables are presented. The prevalence of the individual safety strategies used in the 6 months preceding filing for a temporary protective order is examined, and the proportion of women who reported ever using one or more safety strategies in each of the dimensions is evaluated.
LCA
LCA was conducted using Mplus version 7 (Muthén & Muthén, 1998-2012). This person-based method was used to classify women into safety strategy classes based on how they answer each of the 17 dichotomous safety strategy actions (1 = yes, action used in 6 months preceding filing for a temporary protective order; 0 = no). Model building began with a two-class model and the number of classes was increased until the fit statistics indicated the model fit was no longer significant. Model fit was assessed using the Akaike information criterion (AIC), Bayesian information criterion (BIC), adjusted-BIC, Lo-Mendell-Rubin likelihood ratio test (LMR), and bootstrapped likelihood ratio test (BLRT; Nylund, Asparouhov, & Muthén, 2007). The latent class models’ entropy values were examined to evaluate how well women were classified (Clark & Muthén, 2009). Prior to determining the final number of classes, the significant models, per the fit statistics, were evaluated to verify that the suggested number of classes was theoretically meaningful and interpretable.
Latent class probabilities, or the probability of latent class membership, were reported. The class probabilities indicate the proportion of the population that is in a particular class (Clark & Muthén, 2009). The conditional probabilities, or the probability of endorsing a particular safety strategy given a woman’s class membership, were also reported (Clark & Muthén, 2009). Finally, posterior probabilities, or the assigning of likely class membership, were determined for each woman. Women were assigned to a class based on their highest probability of being in that given class, though it was often the case that women were partial members of more than one class (Clark & Muthén, 2009).
Exploration of safety strategy classes
To evaluate whether LCA-determined safety strategy classes differed by age, race, education, income, employment status, relationship to children in the household, and relationship with the abuser, ANOVA and Pearson’s chi-square tests were used. Other variables examined were severity of abuse experienced and perception of danger from IPV. For the between-class comparisons, class membership was treated as an exact observed variable (Clark & Muthén, 2009). These analyses were carried out using Stata 11.2 Statistical Software (StataCorp, 2009).
Results
Sample Characteristics
The majority of women were between the ages of 18 and 29 years (44%), were Black or African American (77%), and had attended some college or vocational school (54%; Table 1). About 56% of women had a monthly income of US$1,200 or less, 29% received federal income support in the form of food stamps, and 48% rented their homes. About half of the women had one or more children with their abusive partner (51%) and 61% of women said their abuser was an ex-partner, either boyfriend, husband, or common-law. When asked about their risk for future violence over the next year, on average, women scored 2.8 (SD = 1.29) on a mean scale of 1-5, where higher scores indicate higher perceived risk of future violence. When asked about violence experienced in the 6 months prior to filing for a temporary protective order, 78% of women reported experiencing severe physical violence, 19% reported experiencing severe sexual violence, and 77% reported experiencing a severe injury from IPV (Table 1).
Distribution of Sociodemographic and Abuse Characteristics for a Sample of 197 Women in IPV Situations Seeking Temporary Protective Orders.
Note. IPV = intimate partner violence; GED = General Educational Development tests.
Other includes White, non-Hispanic (n = 32), Hispanic (n = 8), Other (n = 6).
Other includes living in a shelter (n = 6) or staying with someone else (n = 47).
Other includes homemaker (n = 15), disabled (n = 13), unemployed (n = 28), other (n = 8).
Perception of danger mean scale score ranges from 1-5, where 1 is low risk and 5 is high risk.
Safety Strategy Use
Table 2 presents the frequency distributions for the five safety strategy dimensions, the 17 individual safety strategies, and the mean number of strategies used in each of the dimensions in the 6 months preceding filing for a temporary protective order. Women reported using on average 8.2 (SD = 3.3) of the 17 safety strategies. A majority of the women reported using strategies from the “leaving with informal network assistance dimension” (90%) and “placating dimension” (89%). On average, women used two (SD = 0.99) of the three strategies in the “leaving with informal network assistance dimension” and 2.6 (SD = 1.3) of the four strategies in the “placating dimension.” Women also frequently used strategies from the “safety planning dimension” (84%), with women reportedly using on average 2.5 (SD = 1.8) of the six strategies. Similarly, women also used strategies from the “legal involvement dimension” frequently (80%). On average, women used 0.89 (SD = 0.53) of the two strategies. The most infrequently used safety strategy dimension was “domestic violence agency advocacy” (23%). Women only used, on average, 0.26 (SD = 0.5) of the two strategies.
Reported Use of Safety Strategies in the 6 Months Preceding Filing for a Temporary Protective Order by a Sample of 197 Women in IPV Situations Seeking Temporary Protective Orders.
Note. IPV = intimate partner violence.
Missing one woman’s response.
Twenty-one women recoded as “no” because they reported never living with their partner.
Missing two responses.
Missing seven women’s responses.
LCA
Five latent classes of safety strategy use were identified. Several fit indices informed model selection and they are reported in Table 3. Results showed that the BLRT was significant for the two-class, three-class, four-class, and five-class models, but not for the six-class model. The LMR likelihood ratio test was significant for the two-class and five-class models. The BIC indicated a two-class model, while the AIC and adjusted-BIC indicated a five-class model. Based on the fit statistics and the substantive interpretation of the model, a five-class model was selected. The entropy for the five-class model was 0.901, suggesting an acceptable level of classification accuracy (Clark & Muthén, 2009).
Fit Statistics for Latent Class Analysis Models.
Note. Latent class analysis carried out using 17 dichotomous safety strategy actions: 1 = yes, action used in the preceding 6 months; 0 = not used. AIC = Akaike information criterion; BIC = Bayesian information criterion; LRT = likelihood ratio test.
Figure 1 displays the item profile plot reporting conditional probabilities of having used the different strategies by safety strategy class. Class 1 (high activity, “doing everything”) was characterized by high probabilities of using all of the safety strategies and comprised 13.7% of the sample. Class 2 (medium activity, “safety planning”) was characterized by moderate probabilities of using the safety strategies, though a greater proportion of women in this class endorsed using the strategies in the “safety planning dimension.” This class comprised 11.8% of the sample. Class 3 (high activity, “police avoidant”) was characterized by high probabilities of using all the safety strategies, except the strategies from the “police involvement dimension,” and comprised 11% of the sample. Class 4 (moderate activity, “placating”) was characterized by moderate probabilities of using the safety strategies, though a greater proportion of women in this class endorsed using the strategies from the “placating dimension.” This was the largest class, and it comprised 52.9% of the sample. Class 5 (low activity, “police involved”) was characterized by low probabilities of using the safety strategies, though a large proportion of women in this class endorsed using the strategy of call the police from the “police involved dimension.”

Probability of engaging in 17 safety strategies given latent class among 197 women in IPV situations in Baltimore, MD.
Associations Between Latent Classes of Safety Strategy Use and Participants’ Demographics
Table 4 summarizes the associations between the demographic covariates and class membership. Significant differences in age, race, employment status, and relationship status with abusive partner were observed (non-significant differences for sociodemographics were excluded from the table). A greater proportion of women in the “safety planning” class were older compared with women in the “police involved” class; for example, 70% of women in the “safety planning” class were 30 years of age and over compared with 31% of women in the “police involved” class. A greater proportion of women in the “placating” class identified as White/Caucasian or Other (30%) compared with women in the “police involved” class (9%). A pattern of statistically significant differences was observed regarding women’s employment status. A greater proportion of women in the “doing everything” class were not employed (i.e., categorized as other, meaning they were unemployed, disabled, or homemakers; 53%) compared with women in the “placating” class (27%) and “police involved” class (9%); and compared with the “police involved” class, a greater proportion of women in the “safety planning” class were not employed (42%). Significant differences regarding women’s relationships with their abusers also emerged. A greater proportion of women in the “safety planning” class said their partner was either their ex-partner or they were estranged from each other (96%) compared with women in the “placating” class (73%) and “police involved” class (68%).
Distribution of Sociodemographic and Abuse Characteristics by Latent Class Membership Among a Sample of 197 Women in IPV Situations Seeking Temporary Protective Orders.
Note. IPV = intimate partner violence.
Difference between Class 1 and Class 2, p < .05.
Difference between Class 1 and Class 4, p < .05.
Difference between Class 1 and Class 5, p < .05.
Difference between Class 2 and Class 3, p < .05.
Difference between Class 2 and Class 4, p < .05.
Difference between Class 2 and Class 5, p < .05.
Difference between Class 3 and Class 5, p < .05.
Difference between Class 4 and Class 5, p < .05.
Associations Between Latent Classes of Safety Strategy Use and Participants’ Experience of Violence
When comparing women’s experiences of physical violence, significant differences were observed (Table 4). A greater proportion of women in the “doing everything” class (87%) and women in the “police avoidant” class (95%) had a history of severe physical violence compared with women in the “police involved” class (59%). Patterns also emerged regarding women’s experiences of sexual violence. For example, compared with women in the “police involved” class (0%), a greater proportion of women in the “doing everything” class (33%), the “police avoidant” class (35%), and the “placating” class (19%) experienced severe sexual violence. Similarly, a greater proportion of women in the “doing everything” class (33%) and “police avoidant” class (35%) experienced severe sexual violence compared with women in the “safety planning” class (4%). Moreover, differences were observed when examining women’s experiences of injury from IPV. Compared with women in the “police involved” class (50%), a greater proportion of women in the “doing everything” class (90%), “police avoidant” class (80%), and the “placating” class (78%) experienced a severe injury from IPV.
Associations Between Latent Classes of Safety Strategy Use and Participants’ Perception of Danger From IPV
Significant differences were observed when comparing women’s perception of danger from IPV (Table 4). Women in the “doing everything” class (M = 3.3; SD = 1.2) had a higher mean score assessing their perception of danger from IPV compared with women in the “placating” class (M = 2.7; SD = 1.3) and “police involved” class (M = 2.2; SD = 1.2).
Discussion
There is a large body of research describing safety strategy use among women in IPV situations. These studies are largely descriptive and explore various factors, including sociodemographics and IPV victimization and severity, which may predict use of those safety strategies, either individually or in conceptual groupings that were defined a priori (Bonomi et al., 2006; Brabeck & Guzman, 2008; Cattaneo, Bell, Goodman, & Dutton, 2007; Duterte et al., 2008; Flicker et al., 2011; Fugate, Landis, Riordan, Naureckas, & Engel, 2005; Hutchison & Hirschel, 1998; Lipsky, Caetano, Field, & Larkin, 2006; Logan et al., 2006; West, Kantor, & Jasinski, 1998; Wiist & McFarlane, 1998). In contrast, this study is only the second study, to our knowledge, that uses statistical modeling, specifically LCA, which is a data-driven approach, to determine classes of safety strategy use based on women’s 6-month history of help-seeking behavior (Goodkind et al., 2004; Kaukinen, 2004). This classification of women into subgroups provides support for the hypothesis that divergent patterns of safety strategy use exist, and that they are associated with differences in experiences of IPV and perception of danger from IPV. In the following sections, a summary of each of the classes is provided, followed by a discussion of implications for future research and intervention.
The current study identified five classes of safety strategy use among a sample of 197 women seeking temporary protective orders against their abusive male partners/ex-partners. The use of LCA enabled us to group women based on their responses to 17 safety strategy items, which is a more informative approach than only examining the total number of strategies used. There were two high-activity classes of women who reported using most of the safety strategies (“doing everything” and “police avoidant”). These classes included women who were the most active safety strategy users. The major difference between the two classes was their use of police involved safety strategies. The “doing everything” class had the highest proportion of women endorse calling the police and filing or trying to file criminal charges strategies, while the “police avoidant” class had the lowest proportion of women of all the classes endorse these strategies. Although these data are cross-sectional, it should be noted that both high-activity classes had strong relationships with severity of violence experienced and perception of danger from IPV. These relationships were present with and without police involvement, which leads to the question of whether police involvement reduces women’s risk of IPV. Previous research on the helpfulness of police in reducing women’s risk of future violence, as reported by the women, is inconsistent with some women reporting the police as helpful and others saying the police made the situation worse (Davies et al., 2007; O’Campo, McDonnell, Gielen, Burke, & Chen, 2002). A qualitative exploration of women’s safety strategy use is needed to examine why these relationships exist while also helping to contextualize women’s help-seeking behavior. Longitudinal studies are also necessary to study the trajectory of women’s safety strategy use.
There were two moderately active classes that included women who used the safety strategies at varying frequencies (i.e., “placating” and “safety planning”). Although very similar to each other in their use of safety strategies, the main difference between the classes was their use of private safety strategies. As the class name suggests, the “placating” class used more placating strategies while the “safety planning” class used more safety planning strategies. In comparison with the women in the two high-activity classes, the women in the moderately active classes used fewer strategies. Moreover, there was one low-activity class, which included women who reported using few safety strategies, except many of the women called the police.
As posited, divergent patterns of safety strategy use among women were identified. Four of the classes accounted for a small proportion of the sample (proportions in these classes range from 11-13%) and their use of the safety strategies ranged from using few of the strategies (low-activity, “police involved” class) to using nearly all of the strategies (high-activity, “trying everything” class). One class, in particular, the medium activity, “placating” class, accounted for more than half of the study population (53%).
It was hypothesized that women experiencing severe physical violence or severe sexual violence would be in the higher activity safety strategies classes compared with women who experienced minor or no physical violence or no sexual violence. The results partially supported this hypothesis in that a greater proportion of women from the two high-activity classes, “doing everything” and “police avoidant,” experienced severe physical violence compared with women in the low-activity “police involved” class. Similarly, compared with women in the low-activity “police involved” class, a greater proportion of women in the high-activity classes, “doing everything” and “police avoidant,” and moderate activity class, “placating,” experienced severe sexual violence. A related relationship was observed when comparing the two high-activity classes with a moderately active class. A greater proportion of women in the two high-activity classes, compared with the moderately active “safety planning” class, experienced severe sexual violence. It was also hypothesized that women who perceived themselves to be at greater risk from IPV would be in higher activity safety strategy classes compared with women who perceived themselves to be at less risk from IPV. This hypothesis was partially supported whereby women in the high-activity “doing everything” class perceived themselves on average to be in greater danger from IPV compared with the less active “placating” and “police involved” classes.
Hypotheses regarding the associations between sociodemographic characteristics and class membership were not proposed, yet significant differences between classes and employment, in particular, were observed. Women who were not employed full-time or part-time were generally in the more active safety strategy classes. For instance, a greater proportion of women in the “doing everything” class were not employed compared with the moderately active “placating” class and the comparatively inactive “police involved” class. Previous research looking at the relationship between employment and safety strategy use is mixed. For example, in one study, among a sample of women who had called the police, it was reported that unemployed women compared with employed women were more likely to sign warrants (Hutchison & Hirschel, 1998), while in another study of women recruited from health centers, being employed or being a homemaker, versus being unemployed, predicted pursuing a protective order (Sabina & Tindale, 2008).
By focusing on patterns of safety strategy use, this study sheds light on the heterogeneity among women and how they protect themselves from risk of future violence. Women are likely using a variety of strategies because no single strategy has been identified that ends the abuse; furthermore, a strategy or set of strategies that may work for one woman may not work for another. The five distinct classes identified in this study are consistent with those identified by Goodkind et al. (2004). Using cluster analysis to group women based on their pattern of safety strategies, five clusters using 28 safety strategies were identified (Goodkind et al., 2004). However, because women were clustered using more strategies and strategies that differed from those examined in this study, a direct comparison is difficult to make. The clusters described included a high-activity cluster (i.e., women endorsed a high level of all strategies), informal help-seeking cluster (i.e., women were more likely to seek help from family and friends and least likely to use placating), going at it alone cluster (i.e., women were not likely to seek help from formal or informal sources), placating cluster (i.e., women used placating strategies and few resistance strategies), and trying everything cluster (i.e., women did not use any of the strategies frequently or intensely) (Goodkind et al., 2004). There is an overlap in the classes identified in this study and Goodkind et al.’s (2004) clusters.
This study’s findings should be viewed in light of several limitations. The cross-sectional study design limits the ability to make causal interpretations from the data. A longitudinal study design is needed to determine causality. All measures were self-report, which may have resulted in response bias. In addition, all women were recruited while they were seeking temporary protective orders against their abusive male partners/ex-partners. As a result, the significant findings obtained are only generalizable to a similar population of help-seeking women in urban settings. In addition, this group of women represents an important subsample of the women using safety strategies, but the results presented here may overestimate the amount of help-seeking women reported. A question that deserves further examination is how to study safety strategy use among women who do not use formal sources of help (i.e., police, medical professionals, lawyers, domestic violence service agencies). Next, although the sample size in this study may be considered small, it is an acceptable number for performing LCA. In fact, LCA has been conducted using similar or even smaller sample sizes (Bair-Merritt, Ghazarian, Burrell, & Duggan, 2012; Golder, Connell, & Sullivan, 2012; Vaughn, DeLisi, Beaver, & Howard, 2009; Yampolskaya, Greenbaum, & Berson, 2009). Finally, class membership was treated as an exact observed variable, and according to Clark and Muthén (2009), there are two important limitations to this approach. First, estimates may be distorted because women are placed in their most likely latent class, ignoring partial class membership. Second, the standard errors are likely incorrect because the uncertainty of the classification has not been accounted for given that class membership is treated as an observed variable (Clark & Muthén, 2009).
Despite the limitations, the results from this study help to understand the patterns of safety strategy use among women who are actively help seeking. Results suggest that important differences exist among women in this high-risk population with regard to safety strategy use. Women are using a variety of safety strategies to protect themselves, which illustrates the point that there is not one particular strategy that works for all women. These findings can inform the development of intervention strategies targeting the particular needs of different subgroups of women so that their risk of future violence can be reduced or prevented. Specific considerations related to targeted interventions should be made for each safety strategy class, including how to reach the different classes of women and connect them to services. For example, women expected to be in one of the two medium activity safety strategy classes (“placating” and “safety planning”) highlight the fact that many women do not seek help from formal sources. Based on this finding, programs or interventions may need to conduct active and targeted outreach to reach these women. Potential outreach locations may include the workplace, clinical setting, or through general community postings (e.g., coffee shops, listservs, online women’s forums). For instance, for women who are employed, employee assistance programs (EAP) may be a valuable resource. Generally, EAPs provide screening, assessments, and referrals for brief intervention or counseling for a range of issues employees may encounter, including IPV (The Employee Assistance Trade Association, 2013). Clinicians are another valuable resource, regardless of women’s employment. In 2011, Department of Health and Human Services Secretary Kathleen Sebelius issued new preventive health services guidelines, which included screening and counseling for interpersonal and domestic violence. The U.S. Preventive Services Task Force released a statement recommending clinicians screen women of childbearing age for IPV, and provide or refer women who screen positive to intervention services (Moyer, 2013). With these requirements, clinicians are in a position to identify women in IPV situations and provide counseling and/or referrals to existing services in the community (Illangasekare & Gielen, 2013). Furthermore, because many women in IPV situations never access formal resources (Goodman et al., 2003), two other ways in which women can be reached are online (if a safe computer is available) through postings on various webpages or listservs and through flyers posted throughout the community. The online posts or flyers would need to contain information that directs women to local resources in the community, including domestic violence service providers. From there, the next step in helping women find safety is to engage them in the safety planning process, where a range of strategies should be discussed. Safety planning must be tailored to meet the needs and priorities identified by the women. In doing so, women may be more likely to implement their safety plans and therefore reduce their risk of future violence.
The purpose of the current study was to use LCA to examine the heterogeneity of women’s safety strategy use and to better understand the relationships between strategy use and experience of violence and perception of danger from IPV. Results suggest that important differences exist among women with regard to their strategy use, experience of violence, and perception of danger. These findings can be used to inform the design and implementation of interventions targeting the specific needs of the different classes of women so that their risk of future violence can be reduced.
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: This work was supported by the T32 Ruth L. Kirschstein National Service Award Institutional Training Grant, National Institutes of Child Health and Development.
