Patterns and Usefulness of Safety Behaviors Among Community-Based Women Survivors of Intimate Partner Violence

Abstract

Women who experience intimate partner violence (IPV) use a variety of safety strategies to reduce the frequency and severity of violence, including both informal and formal help-seeking. The purpose of this study was to identifying patterns of engagement in safety behaviors by U.S. women from outside of formal service settings, examine which factors are associate with different patterns of use, and examine the perceived usefulness of safety strategies among women who used them. Cross-sectional data from 725 women experiencing IPV were used for these analyses. A cluster analysis revealed three clusters of safety behavior use among the IPV survivors: Exploring Safety Options, Avoiding the Justice System, and Trying Everything. The trying everything cluster had high rates of use across all of the safety behaviors, they also reported the highest levels of physical, sexual, and psychological IPV. The exploring safety options cluster used the fewest safety behaviors and had the lowest level of IPV. Higher violence was related to a higher likelihood of finding safety planning helpful and a lower likelihood of finding leaving home helpful. Women who were currently living with their partner were less likely to find talking with a professional, making a safety plan, or leaving home helpful. Higher decisional conflict - uncertainty about what safety decisions would be best - was almost universally related to greater likelihood of not finding safety behaviors helpful. The study findings reinforce the importance of working with survivors to tailor safety plans with strategies that reflect their situation, and provides insights into for which tailoring of resource recommendations may be made.

Approximately 1 in 3 women in the United States are abused by an intimate partner in their lifetimes (Black et al., 2010). In addition to physical injuries that are the result of violence, many victims of intimate partner violence (IPV) report severe and chronic health and mental health issues, such as traumatic brain injury and post-traumatic stress disorder (PTSD) (Black et al., 2010; Bonomi, Holt, Martin, & Thompson, 2006; Campbell, 2002; Kwako et al., 2011; Macy, Ferron, & Crosby, 2009). For women who have survived IPV, gaining and maintaining safety is a complex undertaking. Contrary to outdated notions of helplessness, abuse survivors are involved in a complex decision making process regarding safety for themselves and their children (Davies & Lyon, 2014; Goodkind, Sullivan, & Bybee, 2004; Goodman, Dutton, Vankos, & Weinfurt, 2005). Women may feel conflicted – uncertain, unsupported, uninformed – about the best path forward (Eden et al., 2015). In particular, efforts to seek safety may be hindered by social and institutional norms that blame victims for continued abuse (Goodmark, 2011), a lack of effective and culturally relevant services (Amanor-Boadu et al., 2012; Sabina, Shamrova, Ham, & Levechenko, 2017; Reina, Lohman, & Maldonado, 2014), love for one’s partner (Messing, Mohr, & Durfee, 2015), belief that abuse is a “private matter” (Ameral, Reed, & Hines, 2017), economic uncertainly (Davies & Lyon, 2014), and fear that violent and controlling behaviors of abusive partners may escalate upon separation (Campbell et al., 2003).

Survivors may have multiple (or even competing) priorities; interventions that are adapted to their needs or service-seeking patterns are survivor-centered and have the potential to be more effective than narrowly focused interventions (Kulkarni, 2019). Given the variety of characteristics and experiences of women experiencing IPV, researchers have investigated the existence of discrete subgroups of women and their corresponding needs. In one study, survivors with severe functioning impairments were found to seek services and engage in protective actions more often than other women whereas survivors with multiple resources sought services and engaged in protective actions less often (Nurius, Macy, Nwabuzor & Holt, 2011). Outside of the US, studies have classified women receiving domestic violence shelter services in terms of service utilization (i.e., Substantial Use, Frequent Use of Welfare/Criminal Justice Service, Minimal Use; Ben-Porat, 2017) as well as service needs (i.e., High Needs, Practical Needs (daily living), Empowerment Needs (mental health), and Low Needs (Jonker, Sijbrandij, & Wolf, 2012). In both of these studies, women experiencing the most severe violence reported the greatest need for services and the highest rates of service utilization. However, beyond this, little is known about the heterogeneity of women’s help-seeking behaviors or how they strategize to meet these needs. As such, the aim of this study was to examine the patterns and correlates of help-seeking among a sample of community dwelling women in abusive intimate relationships, as well as to understand the characteristics of women who find particular safety behaviors useful.

Help-seeking among survivors of intimate partner violence

The only way for violence to end in an intimate relationship is for the abuser to cease being abusive. Nevertheless, women in unsafe intimate relationships employ a variety of safety strategies in an attempt to hasten or precipitate reductions in the frequency and severity of violence, including both informal and formal help-seeking. Informal strategies include actions such as seeking support from family or friends or having other people around (Goodkind et al., 2004; L. A. Goodman & Smyth, 2011; Goodman, Dutton, Weinfurt, & Cook, 2003). Survivors of IPV may seek assistance from family and friends for emotional and tangible needs (i.e., housing, transportation, childcare) with or without disclosing their victimization (Belknap, Melton, Denney, Fleury-Steiner, & Sullivan, 2009). When survivors disclose abuse, the reactions of family and friends may help survivors define their situation and build perspective around decision making (Davies & Lyon, 2014). It appears that informal supports precede formal support in most cases (Glass, Eden, Bloom, & Perrin, 2010; Goodman et al., 2003), although oscillating among choices is a normal part of the process of addressing violence and testing various strategies in an attempt to end it (Thaller, Messing, Laughon, & Campbell, 2014).

Safety plans may be developed with or without assistance from a professional, and include tactics that can be used during an incident of violence in order to avoid abuse or minimize the injurious consequences of it (Davies & Lyon, 2014). Plans may include packing a bag of necessities for a quick escape, maintaining an easy-to-access list of emergency phone numbers, creating a code to alert others of danger, changing locks, or removing weapons from the house (Goodkind et al., 2004; Messing et al., 2014). Emergency safety plans have been shown to be helpful by increasing safety behaviors and decreasing violence for some women (Goodkind et al., 2004; Goodman et al., 2003), but appear to have less utility as abuse escalates. When the abuser has inflicted minor or moderate abuse, these strategies are protective against re-abuse; however, emergency safety planning is not protective against re-abuse by perpetrators of severe violence (Goodman et al., 2003).

Over the last several decades, formal interventions targeted at helping survivors escape or avoid abuse have grown and professionalized (Messing, Ward-Lasher, Thaller, & Bagwell-Gray, 2015). Research indicates that a small proportion of abuse survivors seek formal social services, such as staying at shelter or speaking with a professional, and that service utilization increases as the severity or frequency of violence increases (Bonomi et al., 2006; Coker, Derrick, Lumpkin, Aldrich, & Oldendick, 2000; Gondolf, 1998; Macy, Nurius, Kernic, & Holt, 2005; Messing et al., 2014; West, Kantor, & Jasinski, 2005). Women with children report both delaying and seeking services in an effort to do what is best for their children; for example, services may be delayed because of a desire to keep their family together and services sought when they realize the impact of IPV on their children (Stephens & Melton, 2017). Survivors from minority cultural groups may find formal institutions particularly unhelpful because of concerns around confidentiality, a lack of cultural appropriateness in services, and language limitations (Cho, Shamrova, Han, & Levchenko, 2017; Flicker et al., 2011; Lee & Hadeed, 2009).

Criminal justice remedies for IPV have been embraced through law and policy as the primary strategy for intervening with offenders of IPV, despite mixed evidence as to their effectiveness (Belfrage et al., 2012; Campbell et al., 2003; Cho & Wilke, 2010; Felson & Paré, 2005; Hirschel, 2008). One meta-analysis indicated that arrest produces a small but significant reduction in the likelihood of re-offending and a small non-significant reduction in re-arrest (Maxwell, Garner, & Fagan, 2001). In cases of mandatory arrest, victims may be arrested themselves or coerced to obtain additional formal services, such as a no-contact order, as a result of their partner’s arrest (Davies & Lyon, 2014). There is a belief that immigrant and minority racial and/or sexual status women may be less likely to seek formal help for fear of bias from police officers and the criminal justice system (Anyikwa, 2015; Cho, Shamrova, Han, & Levchenko, 2017; Edwards et al., 2015).

Most women in violent relationships eventually leave their partners, but, for many, this process requires persistence and fortitude in the face of complex life issues, such as financial self-sufficiency, adequate housing, child custody, and continued safety (Cattaneo & Goodman, 2005; Messing, O’Sullivan, Cavanaugh, Webster, & Campbell, 2017). Many women may be in a process of mental planning or emotional dissociation from their partners that may not be evident to those around them (Bermea, Khaw, Hardesty, Rosenbloom, & Salerno, 2017). Moreover, they may be weighing the advantages of accessing a certain safety behavior with its associated risks, or determining whether their needs can be met through formal services (Ben-Porat, 2017). Aggregate research has identified correlates of women’s formal help-seeking behavior, such as severe violence, PTSD, employment, older age, having children, and attempting to separate from one’s partner (Ben-Porat, 2017; Goodkind et al., 2004; Nurius, Macy, Nwabuzor, & Holt, 2011). However, women’s needs depend upon a unique configuration of the type of violence they are experiencing, individual and interpersonal characteristics, as well as broader sociocultural factors.

Previous research has examined patterns of service utilization and needs but not patterns of safety strategies or help seeking among US women, nor have prior studies examined these patterns among women who were not sampled within service settings. Thus, this study adds to the current literature by identifying patterns of safety behaviors by US women sampled from outside of service settings, as well as correlates of women’s safety and the perceived usefulness of safety strategies among those who used them.

METHODS

Overview

Analyses were conducted using baseline data from a multisite randomized controlled trial (ClinicalTrials.gov # NCT01312103). Detailed study methods are published elsewhere (Eden et al., 2015). Briefly, the study enrolled English- or Spanish-speaking adult women in four states (Arizona, Maryland, Missouri, and Oregon). Women were recruited through advertisements (e.g., posted online, in community centers), screened and enrolled into the study via telephone by a study research assistant. Women were eligible if they reported currently being in an unsafe or abusive relationship and access to a safe computer to complete intervention measures. Women were randomized to receive an intervention safety-decision aid website (consisting of an interactive risk assessment, priority setting activity, and personalized safety plan) or a standard informational website related to safety planning. Following randomization, women were directed to the study website and data collection platform to complete baseline measures before being taken to their respective study intervention website. All data were collected between March 2011 and May 2013. Study protocols were approved by IRBs at Johns Hopkins University, Arizona State University, University of Missouri and Oregon Health Sciences University.

Measures

Demographics

Demographics included standard sociodemographic variables well as some particular to the IPV context drawn from our previous research. Standard socio-demographic variables included age; race (White, Black, Asian, Native American, Pacific Islander, multi-racial, and other); educational status (no high school diploma or GED, high school diploma or GED; some college; associate’s degree or vocational graduate, 4 year college degree/Bachelor’s degree, post-baccalaureate/master’s degree/Ph.D.), and number of children under 18 living at home with the participant. The IPV descriptive variables included the gender of their abusive partner (male, female), and whether they currently live with their abusive partner.

Safety-Related Behaviors

Safety-related behavior use was measured using a subset of items from a larger 35 item instrument adapted from those used in previous studies of help seeking behavior among women in abusive relationships (Goodman et al., 2003; Sullivan & Bybee, 1999). This includes a checklist of safety strategies (e.g. asking for help, removing gun from home, hiding important papers) and use of community resources, including criminal justice, health, and social resources (Parker, McFarlane, Soeken, Silva, & Reel, 1999). In the analysis section, more will be explained about how items were chosen for inclusion in clusters.

Decisional Conflict

In this study, Decisional Conflict Scale DCS scores pre-intervention were used to understand the relationship of women’s help seeking choices to their decision about their path forward for safety. The DCS includes 16 items and includes four subscales: uncertain, uninformed, unsupported, and unclear about priorities (e.g. feels clearer about personal values regarding risks and benefits of safety decision; “I am clear about which benefits (of the relationship) matter most to me.”). Questions were scored as 0 (yes), 2 (unsure), and 4 (no). The mean DCS score was multiplied by 25 to create an overall conflict score, ranging from 0 (no decisional conflict) to 100 (extremely high decisional conflict). The DCS has demonstrated effectiveness in discriminating between people who make decisions and those who delay making decisions in both English and Spanish speaking samples (Cronbach’s alpha = 0.72–0.92) (O’Connor, 1993; O’Connor, 1995). The Cronbach’s alpha for our sample was .78.

Intimate Partner Violence

The Severity of Violence Against Women Scale (SVAWS) was used to assess the severity of IPV. The SVAWS measures frequency of 46 violence behaviors rated on a scale ranging from 1 (never) to 4 (many times). Subscales include non-physical (19 items, alpha=.92), physical (21 items, alpha=.94), and sexual (6 items, alpha=.91) violence and are computed by summing the items (Marshall, 1992). The alpha for the full scale in our sample was .96

Mental Health

Symptoms of depression and post-traumatic stress were measured using the Center for Epidemiologic Studies Depression Scale, Revised (CESD-R) and PTSD Checklist (PCL), Civilian Version respectively. Both measures have been used in diverse community-based samples and have demonstrated good reliability and correlation with clinical diagnostic criteria (Blanchard, Jones-Alexander, Buckley, & Forneris, 1996; Radloff, 1977). The internal consistency reliability for the PCL in our sample was alpha=.92, and for the CESD alpha=.95.

Substance Use

Alcohol use was measured using the Alcohol Use Disorder Identification Test (AUDIT). This screening tool includes 10 items related to hazardous drinking behavior, including excessive drinking, alcohol dependence symptoms (impaired control and morning drinking), and alcohol-related problems. A cut off score of 3 was used to indicate a positive screen for alcohol use disorder; sensitivity of 99% and a specificity of 74% (Bohn, Babor, & Kranzler, 1995; Reinert & Allen, 2007). Past year drug use was measured using the 10-item version of the Drug Abuse Screening Tool (DAST). The measure provides a total score (range 0–10) by summing item responses. A cut score of 3 was used to indicate a positive screen for drug abuse or dependence; sensitivity=.70 and specificity=.80 (Skinner, 1982; Maisto et al., 2000).

Analysis

Hierarchical cluster analysis with Ward’s Method was used to explore groups of women with different patterns of safety behaviors. We began by looking at the percentage of participants who used each safety behavior. Safety behaviors specific to children were omitted from the analysis as they were not applicable to all participants. Behaviors that less than 10% or more than 90% of participants used were excluded or combined. The following variables were combined: 1) Staying in a shelter or with family or friends, 2) obtaining a restraining order, pressing criminal charges, and using legal assistance, 3) Talking with a boss was combined with talking with friends or family (Table 1 presents the final set of items). The agglomeration schedule, dendrogram, and interpretability of solutions were used to determine the number of clusters to retain. Next, we used K-means cluster analysis to finalize the classification of individuals into clusters and discriminant analysis to further validate classification. Chi-square tests and ANOVAS were used to examine differences between the safety clusters. Finally, logistic regressions were used to examine the characteristics that were uniquely associated with finding individual safety variables useful among those who used that specific behaviors. All predictors were entered into the model simultaneously.

Table 1:

Proportion of women who have tried safety behaviors by cluster.

	Exploring Safety Options	Avoiding the Justice System	Trying Everything
	(n=190)	(n=251)	(n=273)
Have you made sure there were other people around so that your partner would not hurt you physically, sexually or emotionally	.19	.74	.77
Have you talked with family, friends, neighbors, or your boss about your partner hurting you physically, sexually, or emotionally	.41	.78	.83
Talked with a professional (such as a victim advocate, hotline worker, clergy, doctor/nurse, counselor, caseworker) about your partner hurting you physically, sexually, or emotionally	.32	.37	.68
Have you developed a safety plan (e.g. made escape plan; hid valuables, money, keys; packed a suitcase including emergency supplies; created codewords to use with others when in danger, removed weapons, changed locks, etc)	.73	.98	.97
Have you stayed with family or friends in order to avoid your partner hurting you physically, sexually, or emotionally	.14	.67	.82
Have you left home and stayed anywhere else (such as a hotel, your car) to get away from your partner	.06	.65	.90
Have you called the police or had police come out to you when your partner was hurting you or threatening to hurt you physically, sexually, or emotionally	.23	.04	.93
Engaged the legal system, such as filing for a restraining order, pressing criminal charges, or seeking legal assistance (for example legal aid, talked to a lawyer) for protection from your abusive partner?	.14	.11	.67

RESULTS

A total of 725 women completed baseline measures. Women ranged in age from 18–66 (M=33.41, SD=10.65). The majority of women were White (63.76%), followed by Black (24.89%), multi-racial (5.09%), Asian (3.49%), Native American (1.60%), Pacific Islander (0.29%), and other (0.87%). Eleven percent (11.34%) identified as Hispanic. Most (78.8%) had completed at least some college. More than half (57.74%) currently lived with their abusive partner. Eleven percent of women had a female abusive partner. Less than half (43.78%) of participants had one or more child living at home under the age of 18.

Safety Behavior Clusters

While there is no definitive evidence on the sample size necessary for a cluster analysis, one commonly used rule of thumb is 2^m, where m is the number of items (Formann, 1984). We used 8 items in our cluster analysis thus our sample size of 725 was well over 2⁸=256. The agglomeration schedule and dendrogram for the hierarchical cluster analysis with Ward’s Method suggested that we consider solutions for 3–4 clusters of survivors with different patterns of safety behaviors. After looking at the interpretability of the solutions, we chose to retain the three-cluster solution. We then used the cluster centers from the hierarchical cluster analysis in a K-means cluster analysis to finalize the classification of individuals into clusters. The proportion of women trying each safety behavior for the three clusters is presented in Table 1. A discriminant analysis indicated that 95.1% of cases were classified correctly providing further validation for the classification.

We named cluster 1 “Exploring Safety Options.” Women in this cluster primarily limited their safety behaviors to developing a safety plan (73% of those in this cluster had done so) and talking with people for support, including 41% who talked with friends/family/boss and 32% who talked with a professional. Women in this cluster engaged in the fewest safety behaviors. Indeed, even the safety behaviors that distinguished this group were utilized in proportions lower than those in the other clusters. The majority of women in cluster 2 tried most of the safety behaviors except for calling the police and using the legal system. Fewer than half (37%) of women in this cluster spoke with a professional, but this remained higher than the proportion of women speaking to a professional in cluster 1. Because of the low proportions of justice system involvement (4% criminal justice involvement, 11% legal system involvement) and the relatively high proportions of other types of safety behaviors in this cluster, we named it “Avoiding the Justice System.” Finally, cluster 3 participants were the most active with regard to engaging in safety behaviors, with the lowest proportion in any of the eight categories being 67% (legal system involvement) and several safety behaviors tried by 90% or more of the participants (criminal justice, staying somewhere else, developing a safety plan). As such, this cluster was named “Trying Everything.”

Developing a safety plan was the safety behavior with the highest utilization across all clusters (73–98%), but a safety plan can have many different components (e.g., save money to be able to stay in hotel if violence increases, talk with an advocate on the domestic violence hotline, ask for time off from work to move to new apartment, file for a protective order). Therefore, we tested the differences between the clusters in the number of strategies included in the safety plan to determine if the pattern was consistent with the interpretation of cluster 1 as women who have tried the fewest safety behaviors followed by cluster 2 and then cluster 3. Cluster 1 had significantly fewer strategies in their safety plan than the other two clusters and cluster 2 had significantly fewer strategies than cluster 3 (cluster 1 – M=1.03, SD=1.14, cluster 2 – M=2.39, SD=1.57, and cluster 3 – M=2.97, SD=1.85). On average, the safety planning done by cluster 1 was minimal, with approximately one safety planning behavior tried and 95% of the women in this cluster engaging in 3 or less safety planning behaviors. In contrast, women in cluster 2 engaged in 2 safety planning behaviors on average with 95% of women engaging in 5 or less safety planning behaviors, and those in cluster 3 averaged 3 safety behaviors with 95% of women engaging in 7 or less safety planning behaviors.

Table 2 describes similarities and differences across clusters on socio-demographic and relationship characteristics, substance use, mental health, decisional conflict, and experience of violence. Across clusters, participants evidenced some similarities. Specifically, women in all of the clusters were of similar ages and racial/ethnic categories. Given the widely accepted belief that there are differences in utilization by race and ethnicity we explored our data further by combining the minority racial categories and comparing these non-White individuals to White individuals, this comparison did not reach statistical significance either. Participants were also similarly likely to be employed and living with a male abusive partner. Equivalent proportions of participants across all clusters screened positive for alcohol abuse. Finally, on the decisional conflict measure, women across clusters were alike in their feelings of being informed, certain about their path, and supported. With regard to differences across clusters, women in the Exploring Safety Options cluster reported the lowest levels of non-physical, physical and sexual violence perpetrated by their partners. Women in the Avoiding the Justice System cluster were least likely to have children; women in this cluster had similar DAST scores, symptoms of PTSD, and values clarity as those in the Exploring Safety Options cluster. The partners of participants in the Avoiding the Justice System cluster perpetrated more non-physical, physical and sexual violence than the partners of women in the Exploring Safety Options cluster. While women in the Avoiding the Justice System cluster reported less non-physical and physical violence than those in the Trying Everything cluster, they reported experiencing similar levels of sexual violence as women in the Trying Everything cluster. Women in the Trying Everything cluster were more likely than women in the other 2 clusters to have children and screen positive for drug abuse. Women in this cluster reported the highest levels of non-physical and physical violence as well as the highest levels of PTSD symptoms. These women also reported lower decisional conflict (i.e., more likely to make decisions) around values clarity than women in the other 2 clusters.

Table 2.

Demographic, relationship, and decision making characteristics by cluster.

	Total	Exploring Safety Options (n=190)	Avoiding the Justice System (n=251)	Trying Everything (n=273)	p- value
Age	16.41(10.65)	33.18(11.52)	33.52(10.76)	33.60(10.04)	.913
Race					.094
White	63.76%	64.41%	64.29%	63.22%
Black	24.89%	24.29%	20.59%	28.35%
Other	11.35%	11.30%	15.13%	8.43%
Hispanic	11.34%	14.29%	10.76%	9.93%	.325
Currently live with partner	57.74%	55.85%	63.31%	54.44%	.099
Male partner	89.04%	86.70%	90.00%	90.48%	.396
Have children under 18 in the home	43.78%	44.44%	33.86%	53.11%	<.001
Currently employed		41.58%	49.60%	40.07%	.070
+AUDIT screen	28.39%	27.51%	29.88%	27.94%	.833
+ DAST screen	21.77%	17.11%	19.35%	27.24%	.019
SVAWS
Non-physical abuse	46.68(12.29)	41.05(11.42) a,b	45.41(11.01) a,c	51.65(11.90) b,c	<.001
Physical abuse	40.97(14.12)	35.48(12.50) a,b	39.65(13.11) a,c	45.57(14.35) b,c	<.001
Sexual abuse	10.73(4.99)	9.52(4.34) a,b	11.00(5.06) a	11.20(5.16) b	.001
Symptoms of PTSD	48.56(14.11)	45.45(14.65) a	46.92(13.28) b	51.97(13.63) a,b	<.001
Decisional conflict
Uninformed	19.62(17.67)	21.05(20.22)	18.45(16.80)	19.71(16.57)	.308
Uncertain	37.03(32.79)	35.39(33.24)	36.89(32.62)	38.00(32.66)	.702
Unclear	29.13(30.07)	30.60(31.89)	32.07(30.53) a	25.00(28.25) a	.019
Unsupported	40.79(32.73)	39.15(33.49)	42.97(32.64)	39.58(32.50)	.380

Notes: Superscript letters indicate statistically significant differences between cluster means within a continuous dependent variable. Any two means with the same superscript are statistically different from one another. AUDIT: Alcohol Use Disorder Identification Tool. DAST: Drug Abuse Screening Tool; PTSD. Post-Traumatic Stress Disorder PTSD ranges from 16-80 with higher scores indicating greater PTSD. SVAW: psychological abuse ranges from 19-76 with higher scores indicating more abuse. SVAW: physical abuse ranges from 21-84 with higher scores indicating more abuse. SVAW: sexual abuse ranges from 6-24 with higher scores indicating more abuse. All decisional conflict scales range from 0 (no decisional conflict) to 100 (extremely high decisional conflict).

Helpfulness of Safety Behaviors

It is important that to understand both the patterns of safety behaviors that survivors use as well as whether they report finding the safety behaviors that they tried as helpful. Using ANCOVA, we examined if the clusters were different in the percent of safety behaviors that they tried and found helpful controlling for the number of safety behaviors that they tried. Women in the Exploring Safety Options cluster found fewer of the safety behaviors that they tried helpful (M=48.60%, SE=1.95) than women in the Avoiding the Justice System cluster (M=57.00%, SE=1.65, p=.001) and the Trying Everything cluster (M=58.01%, SE=1.67, p<.001). The Avoiding the Justice System cluster and the Trying Everything cluster were not statistically significant from one another in the percentage of safety behaviors they found helpful.

In order to examine what factors might be helpful in tailoring recommended safety behaviors for women experiencing IPV, we examined if sample characteristics (whether participants currently live with their partner, have children, are pregnant, are employed; the severity of violence and PTSD symptoms; abuse of alcohol and illegal drugs; and decisional conflict) were related to the perceived helpfulness of individual safety behaviors. In the analyses, for each safety behavior, we included only those women who reported using that behavior and we did not differentiate this by cluster. All predictors were entered simultaneously so the odds ratios reflect the unique relationship between a predictor and the outcome controlling for the other predictors in the model. The results can be found in Table 3. The characteristics examined did not significantly predict the helpfulness of the safety behavior of having other people around to avoid abuse (40.1% of women found this helpful overall) nor did these characteristics predict the helpfulness of engaging the legal system (to obtain a restraining order, for example; 16.4% of women found this helpful overall). Decisional conflict predicted the helpfulness of the most safety behaviors (n=5), with higher scores on decisional conflict being associated with significantly lower likelihood of finding the following safety behaviors helpful: talking with friends or family about the abuse; developing a safety plan; staying with friends, family, or in a shelter; leaving home; and using the criminal justice system. Living with an abusive partner predicted the helpfulness of three safety behaviors, women who currently lived with their abusive partner were less likely to find talking with a professional, developing a safety plan, and staying somewhere else helpful. Higher level of violence was associated with a significantly increased odds of finding a safety plan helpful, but decreased odds of finding leaving home helpful. Women with higher levels of PTSD symptoms were significantly less likely to find staying somewhere else helpful and, finally, women who screened positive for drug use were more likely to find leaving home helpful than those who did not screen positive for drug use.

Table 3:

Results from logistic regression models predicting whether women reported specific safety behaviors as helpful.

	Had other people around (n=423)		Talked with friends/family (n=487)		Talked with professional (n=324)		Safety Plan (n=633)		Stayed somewhere else (n=410)		Left home (n=413)		Police (n=309)		Legal (n=229)
	OR	p	OR	p	OR	p	OR	p	OR	p	OR	p	OR	p	OR	p
Live with partner	.91	.651	.89	.562	.48	.013	.52	.000	.59	.027	.73	.164	.67	.132	.64	.118
Have children	1.35	.171	1.26	.234	1.19	.528	.98	.920	1.01	.976	.79	.296	1.48	.133	1.76	.055
Pregnant	.75	.329	.89	.631	.77	.406	.83	.390	.82	.512	.58	.060	.63	.194	.75	.409
Employed	1.03	.890	.87	.488	1.13	.663	.85	.356	.85	.483	1.18	.452	.95	.852	.67	.171
SVAWS	.99	.101	.99	.059	1.00	.449	1.01	.000	.99	.077	.98	.001	.99	.272	.99	.051
Symptoms of PTSD	.99	.453	.99	.080	.99	.501	1.00	.482	.98	.034	.99	.197	1.00	.934	.99	.474
+ Audit screen	1.01	.957	1.19	.451	.96	.903	.75	.145	.91	.693	.72	.156	.83	.562	.84	.592
+ DAST screen	.87	.598	.88	.591	.82	.541	1.20	.374	.72	.211	1.71	.038	1.17	.599	.57	.115
Decisional conflict	.99	.524	.97	.000	.99	.143	.99	.004	.98	.005	.98	.001	.97	.001	.99	.457

Note: Only participants who reported using a strategy were include in the analyses for that strategy. All predictors were entered into a model simultaneously. AUDIT: Alcohol Use Disorder Identification Tool; DAST: Drug Abuse Screening Tool; PTSD: Post-Traumatic Stress Disorder SVAWS: Severity of Violence Against Women Scales.

DISCUSSION

These findings advance current knowledge of safety behavior engagement among community-based survivors of intimate partner violence. The three cluster patterns, Exploring Safety Options, Avoiding the Justice System, Trying Everything, represent groups of women who seek both informal help and formal services. The Exploring Safety Options Cluster is distinguished by talking to other people, both family/friends and professionals, and developing a safety plan. However, about one-quarter (23%) of women in this cluster also used the criminal justice system, indicating a combination of formal and informal safety behaviors. Our findings are consistent with previous research (i.e., Ben-Porat, 2017) that found that 499 abused women recruited from a shelter program were classified into three safety behavior groups: minimal use, social and criminal justice services, and all services. The Minimal Use and All Services clusters from the previous study align with the Exploring Safety Options and Trying Everything clusters in this study. The remaining cluster, Avoiding the Justice System in the current study and Criminal Justice Services identified by Ben-Poret (2017) do not reflect one another. This may be due to differences in measurement as we included informal services and safety planning as help-seeking, creating a broader definition of safety behaviors. It may also reflect the different contexts in which the studies were conducted (US vs. Israel) or the different sampling strategies used (survivors recruited from services vs. community-based recruitment).

The majority of this community-based sample of abused women reported seeking support from formal services and the distinctions between clusters appear to be based more upon the specific type of services that women choose to use. The Avoiding the Justice System cluster exemplifies this, with both high levels of formal and informal services, but a conspicuous lack of justice system engagement. We found no differences in patterns of service use by race, ethnicity, or partner gender. While it is generally believed that women from minority racial, ethnic, cultural, and/or sexual status may be less likely to seek help from formal services such as law enforcement, there is some evidence that this might not be true for police services (Anyikwa, 2015; Cho et al., 2017; Edwards et al., 2015). A mixed methods study with 110 Black survivors found that they are highly likely to rely on informal supports such as family and members of their faith communities (Anyikwa, 2015). Black survivors have also been shown to use higher rates of informal supports when compared to White survivors (Cho et al., 2017), they did not compare formal service use. Anyikwa also found that Black individuals where highly likely to use emergency services like the calling the police (66.4%). High levels of both informal support and emergency formal help seeking could be explained by reports of greater severity of IPV among Black survivors than White survivors (Cho et al., 2017). A systematic review of IPV research among people who are LGB indicated that similar to heterosexual survivors they also report high levels of informal support seeking (Edwards et al., 2015), with, 60% of LBG survivors indicating that they reported IPV to the police. While we found no differences in patterns of support seeking by race, ethnicity or sexual minority status, these women may differ from the other clusters in unmeasured structural variables that make police intervention more or less likely (e.g., low/high density housing, low/high socioeconomic status). It is also possible that if help seeking resources were considered on a more micro level differences difference between sociodemographic characteristics would become more apparent. It may also be the case that minority racial/ethnic survivors place a high value on keeping their partner from entanglements with the justice system. More research on this topic is needed.

The women in the Avoiding the Justice System cluster were the least likely of the three clusters to have children (33.86%). Previous research has found that women often choose to both delay and seek services in order to do what is best for their children (Stephens & Melton, 2017). Women with children were more likely to use the justice system in this sample, perhaps child custody plays a role in these findings. Women in the Avoiding the Justice System Cluster reported less non-physical and physical violence than the Trying Everything cluster, but the amount of sexual violence that their partners perpetrated was the same as the Trying Everything cluster (and higher than the Exploring Safety Options cluster). A review of the IPV literature found that women are less likely to report intimate partner sexual violence to the police than other forms of IPV (Bagwell-Gray, Messing, & Baldwin-White, 2015). Further, sexual violence is less likely to be reported to police by bystanders due to its intimate nature; as such, the relatively high levels of sexual violence reported by women in Avoiding the Justice System cluster may be related to the low levels of justice system involvement.

Women in the Trying Everything Cluster had high levels of both social service and justice system involvement as well as the highest levels of physical violence in this sample. This is consistent with previous research indicating that service utilization increases as the severity or frequency of physical violence increases (Bonomi et al., 2006; Coker et al., 2000; Gondolf, 1998; Macy et al., 2005; Messing et al., 2014; West et al., 2005). Women in this cluster were most likely to have children, which may indicate that their high levels of service use is associated with a desire to protect their children. However, these women also face high levels of barriers, including screening positive for drug abuse and high levels of symptoms consistent with PTSD. Consistent with this finding, previous research has indicated that both PTSD symptoms and substance use are associated with IPV (Coker et al., 2002; Dutton et al., 2006; Larson et al., 2005; Machtinger, Wilson, Haberer, & Weiss, 2012; Woods, 2005). Although these health issues may make service access and utilization more difficult, this did not appear to be the case for women in this sample. Staying somewhere else or leaving home were less helpful for women with symptoms of PTSD or substance abuse issues, perhaps indicating that these women have more difficulty separating from a partner who may control their access to drugs and/or alcohol.

Although severity of violence was associated with higher service use, overall severity of violence was not predictive of the helpfulness of most formal services, including use of the justice system. A higher score on the SVAWS (non-physical, physical, and sexual violence) was associated with decreased helpfulness of the safety behavior of leaving home. This could reflect the pervasive, controlling and dynamic nature of IPV. Partners who perpetrate violence are likely to continue to do so; safety behaviors, such as trying to leave home, may trigger more severe violence, resulting in the survivor deciding to remain in the relationship for her and her children’s safety. Women reporting more severe violence also reported increased helpfulness of safety planning. Safety plans include emergency strategies that may reflect their utility in escaping severe violence while it is occurring. Regardless of cluster, women who were currently living with their partner were less likely to find talking with a professional, making a safety plan, or leaving home and staying with a friend or family member or at a shelter helpful. This may be due to the difficulty of separating or finding distance from an abusive partner when your family, social network and financial situation is entwined.

We found decisional conflict to be an important, and widely overlooked, factor in working with women to promote safety. Indeed, the overall decisional conflict score was one of the most robust predictors of the helpfulness of a safety behaviors with higher levels of decisional conflict associated with reports of safety behaviors being less useful for five of the eight categories examined including: talking to a friend or family member; safety planning; leaving home; staying somewhere else; and calling the police. This may indicate that women who are feeling uncertain are in a process of test different strategies for safety, during this process they may to continue to experience a higher degree of decisional conflict if they are not finding the strategies they are trying helpful. Women in the Trying Everything Cluster also reported the lowest levels of decisional conflict with regard to clarity of values. This may indicate that eventually as women try more and more options they are able to focus on helpful strategies and get greater clarity about what is most important to them. The only factor that was related to whether the women found involving the police helpful was having high decisional conflict, perhaps indicating that a survivor is more likely find the involvement of the police helpful if she is uncertain or unclear about what steps to take next to increase her safety.

Limitations

The study is cross-sectional analysis of survivors use and the helpfulness of safety behaviors and is therefore limited in its ability to explain potential changes in participant safety-seeking behavior over time. We recognize that women’s lives and relationships are complex and dynamic, and capturing all the context in which decisions are made regarding service use and usefulness is unlikely using survey methods. While our sample provides new information regarding women who are not currently in or recruited from a shelter or service seeking setting, it is still limited in geographic, racial, and ethnic diversity. Our sample is approximately one-quarter Black, largely coinciding with our Maryland recruitment site. While we did not find racial or ethnic difference in cluster membership in our sample the experiences of minority groups are likely not fully represented, but may contribute to variations in help seeking behavior.

Conclusions

The study findings reinforce the importance of service providers and advocates to work with survivors to tailor safety plans with strategies that reflect their situation, including a survivor’s certainty, clarity, and support for making safety decisions. A risk assessment tool, such as the Danger Assessment, helps women to understand the severity and lethality of the violence in the relationship and is critical information survivors need to plan for safely leaving the relationship or staying in the relationship. Further, tailoring safety behaviors to women’s priorities and risk for repeat, severe violence will likely result in women utilizing the behaviors and finding them helpful for her safety. In response to these findings and others, the team developed myPlan app, the first, to our knowledge, web-based safety decision and planning app for survivors and service providers that includes a validated risk assessment (the Danger Assessment, www.dangerassessment.org), reviews safety priorities, and provides a tailored safety action plan that is linked to diverse services. myPlan app is accessible online or downloadable for free on iOS and Android platforms (www.myplanapp.org). There are apps that provide emergency solutions in times of crisis or resources for survivors, but myPlan app is intended to promote a survivor’s decisional clarity about safety as she assesses her situation and how to best plan and take action for their unique safety needs, including links to services.