Literature reviews

ORIGINAL ARTICLE

An Eight State Study on the Relationships Among Domestic Violence Shelter Services and Residents’ Self-Efficacy and Hopefulness

Cris M. Sullivan1 & Tyler Virden1

Published online: 8 July 2017 # Springer Science+Business Media, LLC 2017

Abstract Survivors of intimate partner violence have myriad reasons why they turn to domestic violence shelter programs. While all are seeking immediate safety and feel a pressing need to leave their homes to obtain it, safety is rarely the only issue survivors are grappling with upon shelter entry. Other concerns are particular to each person’s history and circum- stances, and include but are not limited to employment, counseling, social support, health care, addiction recovery ser- vices, immigration help, housing, and services for their chil- dren. The current study involved secondary analysis of survey data completed at two points in time by 565 shelter residents. The original research involved 215 domestic violence pro- grams across eight states in the U.S., and surveys were com- pleted by shelter residents shortly after they arrived in shelter and again as they were close to exiting. Results confirmed that survivors had numerous needs in addition to safety when en- tering shelter. Residents’ overall rating of how helpful their stay at shelter had been for them was predicted by how much help they had received across their presenting needs, as well as how they were treated by staff. The amount of help received, as well as treatment by staff, also related to survivors’ hope- fulness at shelter exit, as well as the extent to which they felt better able to do things on their own. The study findings pro- vide further evidence that domestic violence shelter staff assist residents with a variety of complex needs in addition to safety, and that this assistance is related to positive outcomes for survivors.

Keywords Intimate partner violence . Shelter . Services .

advocate . Outcomes

Survivors of domestic violence (DV) have different reasons for using shelter-based programs, but all are in immediate danger and needing safety for themselves and often their children (Grossman and Lundy 2011; Sullivan and Gillum 2009). Some intend to leave the relationship permanently, and others are seek- ing temporary respite with the hopes they can salvage their rela- tionship if their partner is willing to change (Grossman and Lundy 2011; Sullivan 2010). Some survivors who seek shelter are being abused by ex-partners, highlighting the reality that ending the relationship does not always end the abuse (Fleury et al. 2000; Hardesty and Chung 2006). Most survivors turn to shelter programs only as a last resort, if they lack the social and economic resources to choose other options (Grossman and Lundy 2011; Panchanadeswaran and McCloskey 2007).

Survivors also enter shelter with different life experiences and needing different types of assistance (Sullivan et al. 2008; Tutty 2006). Some may need information about domestic vio- lence and safety planning, others need help with practical issues such as housing and employment, and others are seeking a combination of emotional support and practical assistance. Most survivors have multiple needs when entering shelter, and rely on staff to provide individualized services and supports to them. Because of this complexity, it is not feasible to exam- ine only one universal Boutcome^ for shelter. Some prior stud- ies of shelter, using post-only designs, have noted shelter resi- dents’ general satisfaction with programs’ Bhelpfulness,^ (e.g., Cannon and Sparks 1989; Fowler et al. 2011). Others have linked shelter stays to leaving the abusive relationship (Gondolf et al. 1992; Panchanadeswaran and McCloskey 2007) or reducing the risk of re-abuse (Bowker and Maurer 1985; Goodkind et al. 2004).

* Cris M. Sullivan [email protected]

1 Michigan State University, Psychology Building, 316 Physics Rd., E. Lansing, MI 48824-1116, USA

J Fam Viol (2017) 32:741–750 DOI 10.1007/s10896-017-9930-7

Shelter staff engage in a variety of activities to help survivors achieve the goals they have set for themselves. In brief, they engage in safety planning with survivors, provide information about their rights and options as well as about the dynamics of domestic violence, offer support and respect, and work to con- nect them with community resources (Glenn and Goodman 2015; Tutty 2006;Wettersten et al. 2004). The philosophy guid- ing advocates’ work is that abuse often diminishes peoples’ personal, interpersonal and social power, and that the way they work with survivors should result in increasing that power (Cattaneo and Goodman 2015). Engaging in empowering practice, then, involves working respectfully with survivors to help them obtain goals they find important to them and to ensure they have the knowledge, skills and self-confidence to feel control over their lives again (Davies and Lyon 2014; Goodman and Epstein 2008; McGirr and Sullivan 2016; Sullivan et al. 2008). This is expected to result in greater self- confidence as well as a more generalized sense of hopefulness.

Hopefulness is viewed as a critical factor relating to overall well-being because it fuels a person’s willingness to do what it takes to maintain or regain health and well-being (Snyder 2002). There is ample empirical support that hope is related to well- being (Gilman et al. 2012; Larson et al. 2007;Wu 2011), and the theory of change guiding advocacy with survivors purports that attaining the ultimate goal of social and emotional well-being (which includes safety) is achieved through this process of empowering practice that ideally results in greater hopefulness, access to resources, and confidence (Sullivan 2016). Studies have provided evidence for these connections, with shelter res- idents reporting staff helping them recognize their inner strengths and linking them to resources (Chanmugam 2011; Few 2005; Haj-Yahia and Cohen 2009; Wettersten et al. 2004).

Although shelters receive high ratings in general, some studies have found that the experiences of women of color are less positive than for White shelter residents (Donnelly et al. 2005; Donnelly et al. 1999; Nnawulezi and Sullivan 2014). Many shelters are staffed primarily by White women, who may be insensitive to needs and issues within cultures other than their own (Donnelly et al. 2005; Few 2005; Gillum 2008). Immigrant women face additional language, cultural, and sometimes legal barriers to accessing services (Erez, 2000; Reina et al. 2014). The complexities of the different experiences of different populations within shelter programs are far from being understood.

The evidence behind shelter effectiveness has primarily come from small studies that have either been qualitative or used post-only measures, limiting confidence in generalizabil- ity of findings. However, two larger-scale studies have been conducted that have examined the impact of shelter services on residents’ well-being and safety. Tutty’s (2006) study in- cluded 368 women using shelters in Canada, surveying resi- dents within 3 days of entering shelter and again toward the end of their stay. The most common needs survivors had at

entry were emotional support (81%) and safety (80%). Overall, residents found programs quite helpful, with the three most important services being safety, emotional support/ counseling, and housing assistance.

Tutty’s large, pre-post design also identified a number of shelter-based outcomes. As they were leaving shelter, women endorsed understanding that they deserve better (100%), feel- ing more hopeful (99%), and having more ways to keep them- selves and their children safe (97%).

A similar study conducted in Ireland, which included 201 women from 18 shelters, found similar needs and outcomes (SAFE Ireland 2009). Upon entering shelter, women’s top priority needs were: staying safe (73%), information and sup- port with housing (63%), making decisions about her life (64%), healing emotionally (63%) and understanding the im- pacts of domestic violence (60%). Out of these women, 100% had been kept safe, 92% better understood the impact of DV, 89% received support with emotional healing, 85% reported receiving information and support with housing, and 85% felt supported in making their own decisions.

The Current Study

Other than the two larger studies conducted in Canada and Ireland, our knowledge about shelter residents’ needs and out- comes has been based on small and/or qualitative studies. The current research was conducted to expand our understanding through a multi-state study in the U.S., using a similar pre-post design to the research conducted in Canada and Ireland. The study tested three hypotheses:

H1: Overall perception of shelter helpfulness will be pre- dicted by whether participants received the assistance they specifically sought, and how they were treated by staff; H2: Survivors’ sense of power at the end of their stay will be predicted by amount of help received and treatment by staff; and H3: Survivors’ hopefulness at the end of their stay will be predicted by amount of help received and treatment by staff.

Given equivocal findings across other studies with regard to outcomes relating to residents’ race and ethnicity, no hy- potheses were generated but race/ethnicity was investigated as a predictor.

Method

This study involved secondary analysis of data collected through a collaboration between the University of Connecticut

742 J Fam Viol (2017) 32:741–750

and the National Resource Center on Domestic Violence, with funding from the National Institute of Justice (Lyon et al. 2008). Domestic violence shelter programs in eight states (Connecticut, Florida, Illinois, Michigan, New Mexico, Oklahoma, Tennessee and Washington) were invited by their state coalitions to participate in the study. Human subjects ap- proval was obtained, and surveys were translated into the 10 languages identified as being the most common languages spo- ken by shelter residents in their states.

Throughout seven months of data collection, shelter staff offered the Time 1 survey to all incoming residents within approximately three days of their entering shelter. Time 2 surveys were offered as residents were getting close to leaving shelter. Participants generated a password at the beginning of each survey, and completed the surveys in private, so that surveys could be matched while anonymity was maintained. Participants were also provided with pre-addressed, stamped envelopes to send their completed surveys directly to the study’s principal investigator.

Survey Development

Survey questions were modified from prior evaluation pro- jects that had involved survivors, advocates, researchers, state coalition staff, and representatives from national domestic vi- olence resource centers, and also drew on what prior empirical studies had identified as being important shelter outcomes (Lyon and Sullivan 2007; Sullivan 1998, 2016).

Time 1 survey The initial survey included demographic ques- tions and asked about survivors’ experiences immediately up- on entry into the shelter. They were asked whether they had been to the shelter in the past and what specific needs they had with which they were hoping to receive help. A list of 37 common needs were provided that respondents could check off as needing (e.g., safety for myself, learning about my op- tions and choices) and they could write in additional needs as well. Items were generated based on the study investigators’ years of experience working with DV survivors, and had been approved by both program staff members and survivors as representing common needs of those seeking shelter (Lyon et al. 2008). The 37 items formed a highly reliable scale (BHelp Needed^), with a Cronbach’s alpha of .90.

Time 2 survey The second survey, completed toward the end of their time in shelter, asked participants to share how many days they had been in shelter and how they had been treated by staff. Specifically, seven questions referred to how staff treated them (e.g., shelter staff treated me with respect, shelter staff were caring and supportive, overall my racial/ethnic background was respected). Participants answered on a 4- point scale from 1 = strongly disagree to 4 = strongly agree, and they could also choose if an item did not apply to them

(e.g., shelter staff helped address any needs related to my disability). The items (BTreatment by Staff^) formed a highly reliable scale (Cronbach’s alpha = .93).

An almost identical list of 37 common needs were again provided to participants, but at Time 2 they were asked to indicate on a 3-point scale, for every need they identified as having, whether they had received (1) no help, (2) some help, or (3) all of the help they had needed. They indicated BI did not have this need^ if an item did not pertain to them, and scale scores were created by calculating the mean across valid re- sponses. This scale (BHelp Received^) was highly reliable, with a Cronbach’s alpha of .99. In addition to the Help Received scale, survivors were also asked to rate how helpful the shelter stay was for them overall (one item, with responses ranging from 1 = not at all helpful to 4 = very helpful).

Participants were also invited to check off which of nine common outcomes they would endorse as being true for them (e.g., I know more ways to plan for my safety, I am more hopeful about the future, I am more confident in my decision making). Response options for these items were yes/no. This scale had a Cronbach’s alpha of .89.

Program characteristics Each participating program also completed a brief survey describing their maximum shelter length of stay, how many staff they employed, the types of services they offered, and the number of beds they provided.

Analyses

Because participants were nested within states, preliminary analyses examined if participants differed on treatment by staff and help-received based upon state. No differences were found for treatment by staff, and the only difference found for help received was participants from Tennessee reporting re- ceiving more help than participants from New Mexico. A Bonferroni correction, dividing the normally used alpha level of .05 by the number of analyses conducted, was used to control for multiple analyses. Because only a single difference between states was found, state was not controlled for in the regression analyses.

To examine if racial/ethnic differences existed in the total number of needs reported by survivors, an ANOVAwas con- ducted comparing the three racial/ethnic groups with a large enough number of participants to run meaningful analyses. Because of the small number of participants in the Asian/ Pacific Islander, Native American, Multiracial, and Other cat- egories, analyses included only: 1)White (n = 336), 2) African American/Black (n = 69), and 3) Hispanic/Latina (n = 57). A Bonferroni correction was again applied.

An ANOVA assessed race/ethnic differences for overall help received and, because survivors may differ on help re- ceived for specific needs, multiple ANOVAs were conducted to examine individual items with a Bonferroni correction

J Fam Viol (2017) 32:741–750 743

applied to the alpha level. ANOVAs were also used to deter- mine if there were race/ethnic differences on treatment by staff and overall ratings of staff helpfulness. Multiple ANOVAs were conducted to examine if race/ethnic differences existed on the individual staff treatment items, and a Bonferroni cor- rection was applied. Because there were no race/ethnic differ- ences found for treatment by staff and help received, race/ ethnicity was not controlled for in the regression models.

To examine the relationships among help-received and staff treatment on survivors’ ratings of overall help received, doing more on their own, and being more hopeful about their future, hierarchical regression models were conducted. Prior stay in shelter and length of stay in shelter were controlled for because it was hypothesized that these variables could impact reported help-received and treatment by staff. To examine the relationship among help-received and staff treatment with sur- vivors’ ratings of overall help, a hierarchical multiple regres- sion was utilized. Test for multicollinearity revealed no con- cerns. Finally, hierarchical logistic regressions were utilized to examine the relationship among help-received and staff treat- ment with survivors’ outcomes of doing more on their own and being more hopeful about their future. This analysis was chosen over an OLS model because logistic regressions pro- duce a more accurate estimate of probability for the dependent variable (Pohlman and Leitner 2003).

Results

Participating Programs

Two hundred fifteen programs across the eight states partici- pated in the study (81% participation rate). They averaged 16.5 full-time equivalent (FTE) staff and 15 volunteers, and average capacity was 25 beds (range = 4–102). Median length of stay limit was 60 days, although 18% reported 30-day limits. Some shelters allowed residents to stay for up to two years, and others stated that they had no pre-determined max- imum length of stay.

Participants

Findings are based on the 565 participants for whom matched surveys over time were available.1 Of these participants, 61% were White, 12% were African American/Black, 10% were Hispanic/Latina, 4% were Native American, 2% were Asian/ Pacific Islander, 4% were Multiracial, and 7% identified as other. A majority of the participants (69%) were between the

ages of 25 and 49, while 21% were under 25 and 9% were 50 years old or older. Educational attainment was quite vari- able, with 29% having less than a high school education and 28% having a high school diploma or GED. Thirty percent had taken some college courses and 13% had a college degree or higher. Only two study participants were male. Length of shelter stay ranged from 1 to 624 days (M = 29.83; sd = 46.06). Most surveys were completed in English (96%); 4% were completed in Spanish, and a small number of surveys were completed in Korean, Russian, or Vietnamese (<1% each).

Descriptives Regarding Services Needed, Help Received, and Treatment by Staff

Participants reported having a wide range of needs when first entering shelter. Of the 37 areas listed for them to choose from (Help Needed), respondents endorsed between 1 and 37 needs (M = 15.54; sd = 7.42). The types of needs that respondents endorsed were categorized into Safety (8 items), Information (3 items), Self-Care and Connections (6 items), Community Resources (14 items), and Children (5 items), and are present- ed in Table 1. Responses are presented within each category by most highly endorsed to least highly endorsed.

As shown in Table 1, the most highly endorsed need in this sample was safety (85.2%), which is not surprising given the purpose of domestic violence shelters. Other highly endorsed needs spanned the categories of Information (e.g., learning about my options and choices: 78.5%), Self-Care and Connections (e.g., emotional support: 75.5%), and Community Resources (e.g., finding affordable housing: 77.9%). We then examined the item asking survivors to rate how helpful the shelter stay was for them overall (from not at all helpful to very helpful). Most (93%) found the shelter stay to have been helpful (16%) or very helpful (77%), while very few reported the shelter stay to have been only a little (6%) or not at all helpful (1%).

Participants reported very favorable treatment by staff overall, with over 90% agreeing or strongly agreeing with each of the seven statements provided (see Table 2). Survivors reported strongest agreement that their sexual ori- entation, religious, and racial backgrounds were respected (98%, 97%, and 97%, respectively), and reported the weakest, but still high, agreement to their disability (92%) and age- related issues (93%) being addressed.

Examining Racial/Ethnic Differences on Needs, Help Received, and Staff Treatment

Analyses examining whether race/ethnic differences existed on overall number of needs reported or help received revealed no significant differences. Chi square analyses examining racial/ethnic differences by individual needs revealed four

1 1,881 shelter residents completed Time 1 surveys only, and 964 completed Time 2 surveys only. Some residents did not both enter and exit shelter during the sevenmonths of data collection andwere therefore unable to complete both surveys. Others may not have provided identification numbers on each survey allowing them to be matched.

744 J Fam Viol (2017) 32:741–750

significant differences: health related issues for my children, χ2(1) = 17.70, p < .001, Φ = .20, counseling for my child, χ2(1) = 16.19, p < .001, Φ = .19, my abuse related injuries, χ2(1) = 15.55, p < .001, Φ = .19, and immigration

issues,χ2(1) = 54.07, p < .001, Φ = .34. To better understand these differences, follow-up chi square analyses examined where the racial/ethnic differences existed, and a restricted alpha level of p < .005 was used due to conducting multiple

Table 1 Extent of help needed and received Help Received by Time 2

Needed at Time 1

Bno^ help

Bsome^ help

Ball^ help

Safety

Safety for myself (n = 559) 85.2% 1.4% 6.4% 92.3%

Safety planning (n = 559) 66.0% 3.9% 15.9% 80.2%

Safety for my child (n = 559) 48.8% 2.2% 7.9% 89.9%

Leaving my relationship (n = 557) 48.7% 8.3% 19.6% 72.1%

Protective/restraining order (n = 558) 26.7% 7.6% 14.3% 78.2%

My abuse related injuries (n = 559) 21.1% 10.3% 13.8% 75.9%

My abuser’s arrest (n = 558) 14.7% 23.1% 9.6% 67.3%

My own arrest (n = 559) 3.0% 12.5% 12.5% 75.0%

Information

Learning about options and choices (n = 559) 78.5% 4.0% 26.1% 70.0%

Understanding domestic violence (n = 559) 70.5% 3.6% 19.1% 77.3%

Budgeting and handling my money (n = 559) 47.4% 21.0% 18.0% 61.0%

Self-care and Connections

Emotional support (n = 559) 75.5% 10.1% 19.3% 70.6%

Pay attention to my wants and needs (n = 558) 74.7% 3.1% 26.6% 70.2%

Handling stress (n = 559) 72.3% 15.6% 24.9% 59.5%

Counseling for myself (n = 559) 71.6% 15.7% 16.3% 68.0%

Support from other survivors (n = 559) 51.5% 6.8% 17.9% 75.3%

Reconnecting with my community (n = 559) 32.7% 13.3% 25.2% 61.5%

Community Resources

Finding affordable housing (n = 556) 77.9% 19.7% 24.5% 55.8%

Connect with others who can help me (n = 559) 71.4% 9.2% 20.8% 70.0%

Job or job training (n = 558) 47.8% 26.6% 21.3% 52.1%

Transportation (n = 559) 47.0% 12.3% 23.6% 64.2%

Health issues for myself (n = 559) 43.5% 14.5% 22.6% 62.9%

Education for myself (n = 559) 39.7% 24.5% 29.8% 45.7%

Other government benefits (n = 558) 34.9% 21.2% 17.4% 61.4%

TANF/welfare (n = 559) 25.6% 18.2% 14.1% 67.7%

Child care (n = 555) 23.6% 19.8% 21.0% 59.3%

Divorce-related issues (n = 557) 23.2% 16.1% 20.7% 63.2%

Custody or visitation issues (n = 559) 21.1% 19.2% 20.5% 60.3%

Child Protection Services (n = 559) 19.0% 13.2% 20.6% 66.2%

Child Welfare System (n = 557) 14.0% 17.3% 23.1% 59.6%

Immigration issues (n = 559) 8.2% 14.8% 18.5% 66.7%

Child(ren)

Pay attention to child’s wants and needs (n = 559) 47.6% 5.2% 18.6% 76.2%

Respond to my children when they are upset or causing trouble (n = 557)

34.5% 13.7% 20.1% 66.2%

Counseling for my child (n = 558) 29.9% 13.6% 22.0% 64.4%

Child’s education (n = 559) 27.4% 6.7% 14.3% 79.0%

Child’s health issues (n = 559) 16.6% 8.8% 24.6% 66.7%

J Fam Viol (2017) 32:741–750 745

analyses. Results indicated that need for assistance with health issues for their children χ2(1) = 16.90, p < .001, Φ = .21 and child counseling χ2(1) = 16.13, p < .001, Φ = .21, were higher among Hispanic/Latina (36%; 53%) survivors when com- pared to White (14%; 26%) survivors. Additionally, the need for assistance with abuse related injuries was endorsed far more by African American survivors (33%), χ2(1) = 11.73, p = .001, Φ = .17, compared to White survivors (16%). Finally, the need for assistance with immigration issues was endorsed far less by African American/Black (10%), χ2(1) = 8.45, p = .004, Φ = .26, and White (3%), χ2(1) = 56.58, p < .001, Φ = .38 survivors compared to Hispanic/Latina survivors (31%).

ANOVAs examining if race/ethnic differences existed in treatment by staff (both overall and by individual item) were found to be non-significant, and post hoc analyses revealed no individual group differences.

Did Help Received and Treatment by Staff Predict How Helpful Survivors Found Shelter?

A hierarchical multiple regression was conducted to examine the study’s first hypothesis, that overall helpfulness ratings would be predicted by extent of Help Received as well as Treatment by Staff. We first controlled for whether the partic- ipant had stayed at the shelter in the past as well as the number of days they had stayed at the shelter (as these may have also contributed to perception of overall helpfulness). The overall model was significant, F(4, 433) = 103.91, p < .001, R2 = .490, with Help Received (p < .001) and Treatment by Staff (p < .001) being significant positive predictors of overall helpfulness. Average rating of Help Received was the stron- gest predictor, explaining 36.3% of the variance, and Treatment by Staff added 12.2% unique explained variance (see Table 3).

Did Help Received and Treatment by Staff Predict Survivors’ Sense of Power?

An explicit goal of DVadvocates is to engage in empowering practice with survivors so that they will regain their sense of

personal power. The outcome most closely indicative of this change was BI feel I can do more things on my own.^ To examine whether a survivor’s endorsement that they could do more things on their own was predicted by Help Received as well as Treatment by Staff (hypothesis 2), we conducted a hierarchical logistic regression using prior stay at the shelter, number of days stayed at the shelter, Help Received, and Treatment by Staff as independent variables (see Table 4). The overall model was significant, χ2(3, N = 418) = 58.30, p < .001, R2 = .130, with average rating of Help Received (p < .001) and Treatment by Staff (p = .013) being signif icant posi t ive predictors . Specifically, for every point increase on Help Received, participants were five times more likely to believe they could do things on their own. For every one point increase on Treatment by Staff, participants were two times more likely to feel like they could do things on their own. Help Received was the strongest predictor, explaining 11.8% of the variance, and Treatment by Staff added 1.2% unique explained variance.

Did Help Received and Treatment by Staff Predict Survivors’ Hopefulness?

A further goal of DVadvocates is to provide survivors with the type of support needed that will raise their sense of hope for the future. We again employed hierarchical logistic regression to examine whether increased hopefulness was predicted by prior stay at the shelter, days in shelter, Help Received, and Treatment by Staff (hypothesis 3; see Table 5). The overall model was significant, χ2(3, N = 425) = 90.78, p < .001, R2 = .192, with average rating of Help Received (p < .001) and Treatment by Staff (p < .001) being significant positive predictors. Specifically, for every one point increase in extent of Help Received, participants were eight times more likely to feel more hopeful about the future, and with a one point in- crease in Treatment by Staff, participants were three times more likely to feel more hopeful about the future. Help Received was the strongest predictor, explaining 16.7% of the variance, and Treatment by Staff added 2.5% unique ex- plained variance.

Table 2 Treatment by staff (reported at time 2)

I feel shelter staff… Strongly Disagree Disagree Agree Strongly Agee Mean (SD)

Treated me with respect (N = 556) 1.4% 3.1% 20.0% 75.5% 3.70 (.60)

Were supportive (N = 551) 1.1% 3.4% 19.8% 75.7% 3.70 (.59)

Respected my religious views (N = 444) 2.5% 0.9% 21. 1% 75.5% 3.70 (.62)

Respected my sexual orientation (N = 430) 1.4% 0.5% 18.1% 80.0% 3.77 (.52)

Respected my racial background (N = 489) 2.0% 1.4% 19.2% 77.3% 3.72 (.60)

Addressed my disability needs (N = 284) 3.2% 4.9% 20.4% 71.5% 3.60 (.73)

Addressed my age related needs (N = 310) 2.9% 3.9% 22.9% 70.3% 3.61 (.70)

746 J Fam Viol (2017) 32:741–750

Discussion

The shelter residents across the eight states participating in the study reported having a variety of needs in addition to safety that they were hoping to receive help with during their shelter stay. Needs ranged from emotional support to practical assis- tance across a variety of areas, and overall, survivors felt that they received a great deal of help and that they were well- treated by staff. As hypothesized, survivors’ outcomes and overall helpfulness ratings were predicted both by what they received from the program and how they were treated by staff.

In both models, help received accounted for most of the var- iance, while treatment by staff accounted for less than 3%. However, treatment by staff was not inconsequential, and they did together lead to increased hope and survivors’ increased confidence in their abilities. As each of these has been found in prior studies to positively relate to well-being (see Sullivan 2016 for a review), these are important findings corroborating the importance of DV shelters to those seeking their services.

Interestingly, shelter outcomes were not predicted by whether the survivor had been in the shelter in the past, nor by their length of stay. Shelter stays were quite variable, rang- ing from 1 to 624 days, and averaging one month in duration (sd = 46.06). The fact that length of stay was not a predictor of overall shelter helpfulness nor of any of the shelter outcomes suggests that shelter staff likely were indeed responding to survivors’ individual needs and goals by allowing them to stay in shelter as long as they found necessary to reach their indi- vidual goals. Some survivors are only in need of brief assis- tance and may only stay in shelter for a few days. Others require more assistance and could stay a year or more. There is no ideal length of time that would work for all survivors, given their individualized circumstances, and these results suggest that shelter staff members are working to allow survi- vors to determine their ideal length of shelter stay.

Few differences in initial needs at shelter entry were found by race or ethnicity. Latinas weremore likely than either Black or White survivors to have immigration needs, and they were more likely than White survivors to have health-related or counseling needs for their children. Unfortunately, no addi- tional data about these issues were collected in the surveys so it is unclear why these differences may have existed. The finding that Black survivors were more likely than White

Table 4 Hierarchical logistic regression predicting participants’ beliefs in their ability to do things on their own (N = 418)

B SEB Odds ratio 95% CI R2 ΔR2

Step 1: χ2(1, N = 418) = .35, p = .554 .001

Not stayed before vs. stayed before −.23 .39 0.79 0.37–1.69

Step 2: χ2(2, N = 418) = 1.17, p = .557 .003 .002

Not stayed before vs. stayed before −.24 .39 0.79 0.37–1.68

Length of stay .00 .00 0.61 0.99–1.02

Step 3: χ2(3, N = 418) = 52.30, p < .001 .118 .115

Not stayed before vs. stayed before .03 .43 1.03 0.44–2.42

Length of stay .00 .00 1.00 0.99–1.01

Help received*** 2.19 .33 8.92 4.70–16.95

Step 4: χ2(3, N = 418) = 58.30, p < .001 .130 .012

Not stayed before vs. stayed before .09 .46 1.10 0.45–2.69

Length of stay .00 .00 1.00 0.99–1.02

Help received*** 1.67 .38 5.33 2.53–11.24

Treatment by staff* .78 .32 2.17 1.18–4.00

The Cox and Snell pseudo-R2 was reported in this analysis

***p < .001, *p < .05

Table 3 Hierarchical regression predicting overall rating of shelter experience (N = 437)

B SE β R2 ΔR2

Step 1: F(1, 436) = 1.28, p = .258 .003

Not stayed before vs. stayed before −.08 .07 −.05 Step 2: F(2, 435) = 1.02, p = .362 .005 .002

Not stayed before vs. stayed before −.09 .07 −.06 Length of stay .01 .01 .04

Step 3: F(3, 434) = 84.07, p < .001 .368 .363

Not stayed before vs. stayed before −.02 .06 −.01 Length of stay .00 .00 .03

Help received*** .79 .05 .60

Step 4: F(4, 433) = 103.91, p < .001 .490 .122

Not stayed before vs. stayed before .00 .05 −.00 Length of stay .00 .00 .04

Help received*** .47 .06 .36

Treatment by staff*** .50 .05 .43

***p < .001

J Fam Viol (2017) 32:741–750 747

women to endorse needing help with domestic violence inju- ries may be an indication that they see shelter programs as more of a Blast resort^ than doWhite survivors, turning to this resource only when the violence is especially severe. While this interpretation is conjecture, and speaks to the need for additional studies with qualitative components, prior studies have noted the reluctance of Black survivors to use shelter programs (Flicker et al. 2011; Gillum 2008).

No racial/ethnic differences were found in survivors’ rat- ings of their treatment by staff nor their shelter outcomes. Ratings were equally high across groups, and survivors of color felt as well-treated by staff as did their White counter- parts. This finding aligns with that of Nnawulezi and Sullivan (2014) as well as Few (2005), but differs from that of Donnelly et al. (1999). Given that each shelter program is unique, and some are more culturally responsive than are others, it is not surprising that there would be differences in experience across shelter programs. Within this fairly large sample spanning eight states, however, no racial or ethnic differences were found across either shelter experiences or outcomes achieved.

Findings should be viewed in light of study limitations. This was a self-report, pre-post design of shelter residents that included no comparison group of survivors who wanted shel- ter but were not able to receive it. Therefore, causality can not be determined from these data. The results are also based on survivors’ perceptions of how much their outcomes were achieved due to services received. Survivors who were less satisfied with services or who were exited from shelter by staff (e.g., for rule violations) may have been less likely to complete an exit survey and may therefore be underrepresented in this sample. Some participants may have also been more inclined

toward positive responses since they were receiving free ser- vices being provided by individuals they liked and respected. However, the fact that respondents did vary their responses to items throughout the surveys gives more credence to the like- lihood that they were providing accurate responses to the questions. Residents’ positive perceptions about the help they received, and their satisfaction with their experience overall, corresponds with findings from a number of prior studies of shelter residents (e.g., Fowler et al. 2011; Goodkind et al. 2004; SAFE Ireland 2009; Sullivan et al. 2008; Tutty 2006).

Analyses were limited by the type of data gathered in the original study. In order to keep the survey brief, many con- structs were measured by only one item rather than by a stan- dardized scale (e.g., hope), and the measure of staff treatment was created specifically for this study. Further, almost all of the study participants were women, and over half (61%) were White. It is unknown how representative these experiences and outcomes are to male shelter residents or individuals from different racial or ethnic groups. More studies, with more di- verse samples, are needed to better understand who needs what from shelter programs, how different people are treated within programs, and whether some individuals attain greater outcomes than others.

In addition to these limitations, the study had a number of noteworthy strengths. This is one of very few studies that included surveys collected at two points in the shelter stay so that change across time could be examined by participant. This was also the largest shelter study conducted to date, in- volving 565 survivors across eight diverse states. Across the eight states, 215 programs participated (81% response rate), and significant efforts were made to ensure that participation in the study was facilitated. Surveys were translated into 10

Table 5 Hierarchical logistic regression predicting increased hope for the future (N = 425)

B SEB Odds ratio 95% CI R2 ΔR2

Step 1: χ2(1, N = 425) = .01, p = .940 .000

Not stayed before vs. stayed before .03 .40 1.03 0.47–2.24

Step 2: χ2(2, N = 425) = .79, p = .673 .002 .002

Not stayed before vs. stayed before .02 .40 1.02 0.47–2.22

Length of stay .00 .00 1.00 0.93–1.02

Step 3: χ2(3, N = 425) = 77.67, p < .001 .167 .165

Not stayed before vs. stayed before .42 .48 1.52 0.60–3.88

Length of stay .00 .00 1.00 0.92–1.01

Help received*** 2.75 .36 15.60 7.69–31.67

Step 4: χ2(3, N = 425) = 90.78, p < .001 .192 .025

Not stayed before vs. stayed before .56 .52 1.76 0.63–4.87

Length of stay .00 .00 1.00 0.99–1.02

Help received*** 2.08 .40 8.00 3.67–17.46

Treatment by staff*** 1.16 .32 3.17 1.70–5.93

The Cox and Snell pseudo-R2 was reported in this analysis

***p < .001

748 J Fam Viol (2017) 32:741–750

languages, and the National Domestic Violence Hotline of- fered to assist with additional languages or if completing the survey verbally would be easier for someone than completing it in writing. The original research team is to be applauded for how diligently they considered issues of accessibility and protecting anonymity in designing this ambitious study.

Policy and Practice Implications

Policy makers and funders are increasingly looking for evi- dence to guide their decisions, and this study adds to the ev- idence base in ways that smaller or more qualitative studies can not. It included hundreds of survivors across eight diverse states, reporting at two different points in time about their needs, experiences and outcomes. Findings confirmed that survivors have many needs besides safety when they enter shelter, and further established that shelter stays lead to reports of greater hopefulness and self-efficacy. These results provide additional empirical support for the theory of change guiding DV programs (Sullivan 2016), and highlight the importance of examining how advocates do their work and treat survivors in addition to their providing actual tangible help. This focus on the importance of the alliance between survivor and advo- cate has received scant but growing attention. A recent study found, for example, that a strong alliance between the two facilitated survivors’ positive mental health through the mech- anism of helping them regain a sense of power and control regarding their safety (Goodman et al. 2016). Policies are needed that provide more resources to programs so that they can address survivors’ complex needs in empathic ways.

Study findings have policy implications beyond funding shelter programs as well. While residents reported receiving a great deal of help from staff in areas related to safety and information, there was more variability in help received with regard to community resources that are scarce in many com- munities. It is informative that, of those survivors needing help finding housing, getting a job, finding and paying for child care, managing custody or visitation, wanting additional edu- cation, or needing government benefits, approximately 20% of the sample received no help that they needed. Affordable housing is becoming increasingly scarce (National Low Income Housing Coalition 2016) and poverty remains a seri- ous problem in this country (Iceland 2013). DV program staff can only do so much when faced with limited community resources; without significant structural and systems changes, too many survivors will continue facing significant obstacles to their safety and well-being.

With regard to practice implications, it is noteworthy that survivors’ increased hopefulness and confidence in their abil- ities were predicted not just by how helpful staff were but how they were treated while in shelter. This suggests that staff need training and ongoing supports to encourage them to behave in

supportive and empowering ways with shelter residents, and to avoid burnout and/or punitive or judgmental behaviors.

In conclusion, the findings from this study indicate that DV shelters provide far more than immediate safety for survivors and their children. Survivors come to shelters with a variety of critical needs, and staff work creatively to meet those needs within an empowering framework. At the end of their shelter stay, survivors not only reported having more strategies to pro- tect themselves and their children, but they felt more confident in their abilities, knew more about domestic violence dynamics as well as their options, and felt more hopeful. These findings suggest that DV shelter programs are serving a critical role in helping survivors recover from abuse and gain the information, skills and confidence needed to rebuild their lives.

References

Bowker, L. H., & Maurer, L. (1985). The importance of sheltering in the lives of battered women. Response to the Victimization of Women & Children, 8, 2–8.

Cannon, J., & Sparks, J. (1989). Shelters - an alternative to violence: A psychosocial case study. Journal of Community Psychology, 17, 203–213.

Cattaneo, L. B., & Goodman, L. A. (2015). What is empowerment any- way? A model for domestic violence practice, research, and evalu- ation. Psychology of Violence, 5, 84–94.

Chanmugam, A. (2011). Perspectives on US domestic violence emergen- cy shelters: What do young adolescent residents and their mothers say? Child Care in Practice, 17, 393–415.

Davies, J., & Lyon, E. (2014). Domestic violence advocacy: Complex lives/difficult choices. Thousand Oaks: Sage.

Donnelly, D. A., Cook, K. J., & Wilson, L. A. (1999). Provision and exclusion: The dual face of services to battered women in three Deep South states. Violence Against Women, 5, 710–741.

Donnelly, D. A., Cook, K. J., Ausdale, D., & Foley, L. (2005). White privilege, color blindness, and services to battered women. Violence Against Women, 11, 6–37.

Erez, E. (2000). Immigration, culture conflict and domestic violence/ woman battering. Crime Prevention & Community Safety, 2, 27–36.

Few, A. L. (2005). The voices of black and white rural battered women in domestic violence shelters. Family Relations, 54, 488–500.

Fleury, R. E., Sullivan, C. M., & Bybee, D. I. (2000). When ending the relationship doesn’t end the violence: Women’s experiences of vio- lence by former partners. Violence Against Women, 6, 1363–1383.

Flicker, S. M., Cerulli, C., Zhao, X., Tang, W., Watts, A., Xia, Y., & Talbot, N. L. (2011). Concomitant forms of abuse and help- seeking behavior among White, African American, and Latina women who experience intimate partner violence. Violence Against Women, 17, 1067–1085.

Fowler, D. N., Faulker, M., Learman, J., & Runnels, R. (2011). The influence of spirituality on service utilization and satisfaction for women residing in a domestic violence shelter. Violence Against Women, 17, 1244–1259.

Gillum, T. L. (2008). Community response and needs of African American female survivors of domestic violence. Journal of Interpersonal Violence, 23, 39–57.

Gilman, R., Schumm, J. A., & Chard, K. (2012). Hope as a change mechanism in the treatment of posttraumatic stress disorder.

J Fam Viol (2017) 32:741–750 749

Psychological Trauma: Theory, Research, Practice, and Policy, 4, 270–277.

Glenn, C., & Goodman, L. (2015). Living with and within the rules of domestic violence shelters: A qualitative exploration of residents’ experiences. Violence Against Women, 21, 1481–1506.

Gondolf, E. W., Fisher, E., & McFerron, J. R. (1992). The helpseeking behavior of battered women; An analysis of 6000 shelter interviews. In E. Viano (Ed.), The Victimology Handbook: Research Findings, Treatment, and Public Policy (pp. 113–127). New York: Garland Publishing.

Goodkind, J., Sullivan, C. M., & Bybee, D. I. (2004). A contextual anal- ysis of battered women’s safety planning. Violence Against Women, 10, 514–533.

Goodman, L. A., & Epstein, D. (2008). Listening to battered women: A survivor-centered approach to advocacy, mental health, and justice. Washington, DC: American Psychological Association.

Goodman, L. A., Fauci, J. E., Sullivan, C. M., DiGiovanni, C. D., & Wilson, J. M. (2016). Pathways to empowerment and mental health for intimate partner violence survivors: The role of the alliance with advocates. American Journal of Orthopsychiatry, 86, 286–296.

Grossman, S. F., & Lundy, M. (2011). Characteristics of women who do and do not receive onsite shelter services from domestic violence programs. Violence Against Women, 17, 1024–1045.

Haj-Yahia, M. M., & Cohen, H. C. (2009). On the lived experience of battered women residing in shelter. Journal of Family Violence, 29, 95–109.

Hardesty, J. L., & Chung, G. H. (2006). Intimate partner violence, paren- tal divorce, and child custody: Directions for intervention and future research. Family Relations, 55, 200–210.

Iceland, J. (2013). Poverty in America: A handbook. Berkeley: University of California Press.

Ireland, S. A. F. E. (2009). Safety and change: A national study of support needs and outcomes for women accessing refuge provision in Ireland. Westmeath: SAFE Ireland.

Larson, J. H., Farrag, M., Jamil, H., Kafaji, T., Abdulkhaleq, H., & Hammad, A. (2007). Hope and fostering the wellbeing of refugees from Iraq. Ethnicity & Disease, 17(Supp 3), 83–84.

Lyon, E., & Sullivan, C. M. (2007). Outcome evaluation strategies for domestic violence service programs receiving FVPSA funding: A practical guide. Harrisburg: National Resource Center on Domestic Violence.

Lyon, E., Lane, S. & Menard, A. (2008). Meeting survivors needs: A multi-state study of domestic violence shelter experiences. Final re- port prepared for National Institute of Justice Grant #2007-IJ-CX- K022. https://www.ncjrs.gov/pdffiles1/nij/grants/225025.pdf. Accessed 25 Aug 2016.

McGirr, S., & Sullivan, C. M. (2016). Critical consciousness-raising as an element of empowering practice with survivors of domestic vio- lence. Journal of Social Service Research. Open access. doi:10. 1080/01488376.2016.1212777.

National Low Income Housing Coalition. (2016). Out of Reach 2016: No refuge for low income renters. http://nlihc.org/sites/default/files/oor/ OOR_2016.pdf. Accessed 12 Feb 2017.

Nnawulezi, N., & Sullivan, C. M. (2014). Oppression within safe spaces: Exploring the presence of racial microaggressions with- in domestic violence shelters. Journal of Black Psychology, 40, 563–591.

Panchanadeswaran, S., &McCloskey, L. A. (2007). Predicting the timing of women’s departure from abusive relationships. Journal of Interpersonal Violence, 22, 50–65.

Pohlman, J. T., & Leitner, D. W. (2003). A comparison of ordinary least squares and logistic regression. The Ohio Journal of Science., 103, 118–125.

Reina, A. S., Lohman, B. J., &Maldonado, M. M. (2014). He said they’d deport me: Factors influencing domestic violence help-seeking prac- tices among Latina immigrants. Journal of Interpersonal Violence, 29, 593–615.

Snyder, C. R. (2002). Hope theory: Rainbows in the mind. Psychological Inquiry, 13, 249–275.

Sullivan, C. M. (1998). Outcome evaluation strategies for domestic vio- lence programs: A practical guide. Harrisburg: Pennsylvania Coalition Against Domestic Violence.

Sullivan, C. M. (2010). Victim services for domestic violence. In M. P. Koss, J. W. White, & A. E. Kazdin (Eds.), Violence against women and children: Navigating solutions. Volume 2 (pp. 183– 197). Washington, DC: American Psychological Association.

Sullivan, C. M. (2016, revised). Domestic Violence Shelter Services: A Review of the Empirical Evidence. Harrisburg: National Resource Center on Domestic Violence. http://www.dvevidenceproject.org. Accessed 25 Aug 2106.

Sullivan, C. M., & Gillum, T. L. (2009). Evaluating community-based services. In E. Stark & E. Buzawa (Eds.), Violence against women in families and relationships. Volume 1: Victimization and the commu- nity response (pp. 55–71). Santa Barbara: Praeger.

Sullivan, C. M., Baptista, I., O’Halloran, S., Okroj, L., Morton, S., & Stewart, C. S. (2008). Evaluating the effectiveness of women’s ref- uges: A multi-country approach to model development. International Journal of Comparative and Applied Criminal Justice, 32, 291–308.

Tutty, L. M. (2006). Effective practices in sheltering women leaving vio- lence in intimate relationships. Toronto: YWCA Canada.

Wettersten, K. B., Rudolph, S. E., Faul, K., Gallagher, K., Trangsrud, H. B., Adams, K., Graham, S., & Terrance, C. (2004). Freedom through self-sufficiency: A qualitative examination of the impact of domestic violence on the working lives of women in shelter. Journal of Counseling Psychology, 51, 447–462.

Wu, H. (2011). The protective effects of resilience and hope on quality of life of the families coping with the criminal traumatisation of one of its members. Journal of Clinical Nursing, 20, 1906–1915.

750 J Fam Viol (2017) 32:741–750