Intimate partner violence (IPV) against women is a significant public health issue worldwide, with lifetime prevalence estimates of between 9% and 71% (Garcia-Moreno et al. 2005; Kishor and Johnson 2004; Johnson et al. 2008). IPV prevalence rates are generally lower in developed countries such as the United States, Australia, and Canada (Johnson et al. 2008; Statistics Canada 2015; Truman and Morgan 2014), than in developing countries, where estimates range between 24% (Serbia & Montenegro) and 71% (Ethiopia) (Demographic and Health Surveys n.d.; Garcia-Moreno et al. 2005). In Colombia, just over 37% of women aged 15 to 49 report having been physically abused by an intimate partner at least once during their lifetime (Profamilia, Ministry of Social Protection, and USAID 2010).

Following the United Nations mandate to eliminate violence against women, shelters for IPV victims have been implemented in many countries (United Nations 2009). These services are essential because they provide emergency accommodation and basic personal supplies (including food and clothes) to survivors of violence. In addition, many shelters offer an array of psychological, social, and legal services for adults, adolescents, and children (Roberts and Lewis 2000; Wathen et al. 2015) that contribute to survivors´ safety, and support women in living lives free of violence.

As many of these services are exclusively for shelter residents, the extent to which women benefit from the services depends on how long they stay at the shelter. However, the extant literature has shown that variations in the residents’ length of stay are considerable (Allen et al. 2004; Hilbert et al. 1997; Lyon et al. 2008; Perez et al. 2012a). For example, Lyon et al. (2008) reported that women could stay in shelters for anywhere between 1 and 624 days. Although the role of the length of shelter stay in levels of re-abuse and mental health has been increasingly examined over the last 20 years of research on IPV, little is known about the factors that contribute to women staying at the shelters for longer periods and thus benefitting from the services offered.

To address this gap, the current study examines demographic data, abuse history, situational needs, and contextual (shelter-specific) factors in a sample of 210 women. We reviewed the case files of all women who accessed a domestic violence shelter in the city of Bogota (Colombia) from 2010 to 2012. Results of a multivariate regression analysis show that education levels, a high level of needs, and the extent to which these needs are met by services in the shelter predict above 30% of the variance in length of shelter stay. We suggest that understanding the factors that predict length of stay can help shelters to plan programs that users will be able to complete.

In this article, we begin by reviewing the domestic violence literature on women accessing domestic violence shelters. We then focus on the empirical research that has explored the experiences of women in shelters—including the impact of their stay. We use this literature to propose factors that should be examined to understand women’s varying length of stay in shelters and develop hypotheses explaining length of stay.

Factors Related to Women Accessing DV Shelters

Previous research has examined three main types of factors related to women seeking a domestic violence shelter: life constraints and related sociodemographic characteristics, abuse history, and situational needs. Women who seek shelter often face “life constraints,” a term used to refer to those conditions that limit abused women’s ability to escape an abusive partner or become independent (Ham-Rowbottom et al. 2005). Studies have shown that shelter residents are more likely to be unemployed, have less housing stability, and lower education levels than women not seeking or residing in shelters (Galano et al. 2013; Grossman and Lundy 2011; Ham-Rowbottom et al. 2005). In terms of sociodemographic characteristics, studies have also shown the importance of age, with women accessing shelters being slightly younger (31.42 years on average) than those not requesting shelter (32.98 years on average) (George et al. 2010; Grossman and Lundy 2011). IPV survivors with children are also more likely to use shelter services than those with no children (Clevenger and Roe-Sepowitz 2009).

Furthermore, women who seek shelter are more likely to experience more controlling and violent behaviors from their partners, including physical, sexual, emotional, and economic abuse than other samples of abused women (Simmons et al. 2008). In George et al.´s (George et al. 2010) study, individuals who requested shelter were more likely to have been physically abused compared to those not requesting shelter, and twice as likely to have experienced sexual abuse. Graham-Kevan and Archer (2003) also found that women who access shelters experience more severe physical aggression than non-users, and are more likely to sustain injuries as a result of their partner’s behavior.

Finally, women who access shelters have a multitude of situational needs (Allen et al. 2004; Lyon et al. 2008). In Allen et al.´s (Allen et al. 2004) study, the majority of women indicated their need to work on obtaining material goods (86%), address health related issues (77%) and increase their level of social support (77%), but more than half also mentioned the need for educational, employment, and legal services. For about two thirds of the women with children, childcare and other issues related to their children were also important. In a larger study based on a sample of 3410 shelter residents in the U. S., the majority of women reported having safety (85%), housing (83%), and information needs (80%) (Lyon et al. 2008). Among women with children, needs related to children’s safety, psychological well being, education and health were prioritized (Lyon et al. 2008). Lastly, in a recent study in Canada, most women reported entering the shelter to obtain emotional support and counseling (81%). Consistent with other studies, the needs of about half of the women with children were related to understanding how the abuse had affected children, childcare, and counseling for their children (Tutty 2015). The literature has also suggested that the extent of the women’s needs may vary and constitute distinct profiles (Jonker et al. 2012), and that the clustering of needs can result in more difficulties when it comes to leaving an abusive partner (Krishnan et al. 2004).

The Experience and Impact of Staying in DV Shelters

Evidence suggests that shelters provide immediate safety for IPV victims whose lives are at risk (Roberts and Lewis 2000; Tutty et al. 1999; Tutty 2015; Wathen et al. 2015). For many women, they represent a crucial resource for leaving an abuser (Zosky 2011). Shelters have also been seen by women’s advocates as a safe place from which to initiate or develop a process of empowerment (Kasturirangan 2008). Studies have found that shelters help users gain information about violence, increase their self-esteem and sense of control and self-efficacy, improve their quality of life, and establish links with other survivors of violence and external social services (Bennett et al. 2004; Few 2005; Tutty et al. 1999).

To respond to their clients´ needs, shelters usually provide a variety of services, including crisis intervention, individual and group counseling, legal advocacy, networking support, vocational training, job placement, and referral services for adult clients (Roberts and Lewis 2000; Wathen et al. 2015). Though less often, they also provide child advocacy, screening, and counseling, in addition to childcare and educational services for children (see Poole et al. 2008, for a review of these programs). Some shelters have even implemented parent training programs (Keeshin et al. 2015; Poole et al. 2008). However, the relationship between the length of shelter stay and the extent to which residents´ needs are met by services provided by or through the shelter has not yet been explored.

Research has shown that the longer women remain in shelters, the more likely they are to benefit from their stay. In the study conducted by Hilbert et al. (1997), staying in a shelter longer was identified as one of the strongest predictors of gaining independence from abusers. In their study, the longer women stayed in the shelter, the less likely they were to return to their abusers following their leaving the shelter. Length of shelter stay has also been associated with lower levels of re-abuse at six months after leaving. In a recent longitudinal study, Perez et al. (2012b) found that the longer women were able to stay in a shelter, the lower the severity of re-abuse experienced over the 6-month period following their exit from the shelter. Based on their findings, the authors recommended that greater flexibility around length of shelter stay could benefit women survivors of IPV.

Length of stay is also related to improved mental health outcomes. Orava, McLeod and Sharpe (Orava et al. 1996) found that a longer stay in a shelter was positively correlated with increased self-esteem and negatively correlated with depression scores. The study by Itzhaky and Ben Porat (Itzhaky and Ben-Porat 2005) showed that self-esteem and sense of personal empowerment were slightly but significantly increased after three months in a domestic violence shelter. The authors suggested that a longer shelter stay could be required to promote more important improvements in these areas. Nevertheless, the literature regarding the effect of shelter stay on mental health outcomes has shown some mixed results. Based on a sample of 150 first-time shelter users, the findings by McFarlane et al. (2014) indicated that women’s mental health functioning and resilience increased at four months as abuse decreased and regardless of the time spent at the shelter. The authors explained their findings by arguing that any contact with the shelter and acknowledgement of the abuse may empower women and thus foster better functioning in mental health dimensions such as depression, anxiety, and somatization.

Just as a longer shelter stay can be an opportunity for residents to use many of the available resources, a shorter shelter stay is a major obstacle for delivering programs to adults and children (Poole et al. 2008). Given that most shelters offer accommodation for up to three months, intervention programs are commonly adapted to shelter conditions of a limited length of stay, recurrent times of crises, and participants´ multiple needs. In the last 10 years, some of these treatment programs, such as HOPE, SUPPORT, and Motivational Interviewing have shown promising results that increase victims´ safety and motivation for leaving abusive relationships, reducing PTSD symptoms, or fostering better relationships between mothers and children (Johnson et al. 2011; Keeshin et al. 2015; McDonald et al. 2006; Rasmussen, Rasmussen et al. 2008). However, between a third and half of the residents leave the shelter before completing the programs, missing important components of the interventions, and making impact evaluation more challenging (Johnson et al. 2011; McNamara et al. 2008; Rasmussen et al. 2008).

Similarly, the experience of using a shelter is not always a positive one. Although many women who have accessed shelters indicate that these are very helpful and emotionally supportive environments (Tutty et al. 1999), some shelter users experience frustration with overly restrictive rules (Glenn and Goodman 2015) and find that the shelter environment can be controlling, coercive and isolating, and the relationships with other residents can be difficult and unpleasant (Haj-Yahia and Chaya Cohen 2009). Studies have also documented some tensions between shelter residents and staff regarding the latter’s controlling behavior, lack of support, and negative perceptions of users´ mothering practices (Glenn and Goodman 2015; Haj-Yahia and Chaya Cohen 2009; Krane and Davies 2007; Peled and Dekel 2010; Tutty 2015). Whether women find the shelter environment nurturing or difficult may influence the length of their stay.

Research Question and Hypotheses

The reasons behind some victims using shelter services for a longer period of time than others remain unclear. Understanding these factors can help promote a longer stay among victims, which has been shown to have positive effects on their health and safety. Since there is no literature specifically studying factors that influence length of stay, we explore whether: 1) factors that influence women’s decision to enter a shelter might also influence the length of their shelter stay, and 2) factors that influence women’s experiences once in the shelter itself might influence their length of stay. Based on the factors identified in the literature, this study examines the life constraints, abuse history, situational needs, and contextual factors (access to services and difficulties) that predicted women’s longer stay in a shelter in Bogotá, Colombia. We propose the following hypotheses:

  • Hypothesis 1 (H1): Women with more life constraints will stay at the shelter longer.

  • Hypothesis 2 (H2): Women who in their history of abuse have experienced multiple forms of controlling and violent behaviors before entering the shelter will stay at the shelter longer.

  • Hypothesis 3 (H3): Women with more needs will stay longer than women with fewer needs.

  • Hypothesis 4a (H4a): The more services women use related to their self-reported needs, the longer they will stay in the shelter.

  • Hypothesis 4b (H4b): Women who do not report difficulties during their stay in the shelter will stay longer than women who do.

Method

The Setting

Colombia is the fourth largest Latin American country, with an estimated population of over 48 million inhabitants (Departamento Administrativo Nacional de Estadística (DANE) 2015). Its capital city, Bogota, has an estimated population of nearly eight million (DANE 2015). According to the Demographic and Health Survey (Profamilia et al. 2010), 37.4% of women aged 14 to 49 in Colombia has experienced physical abuse at some point in their life at the hands of an intimate partner, 7% has been threatened with a weapon, 3.1% has been attacked with a weapon, and 9.7% has been raped. The availability of legal, social, and psychological services for survivors of IPV in Bogota is limited.

The shelter at which the study was conducted was established in Bogota in 1995 by a non-profit organization. It was for many years the only available shelter for abused women in the city. This shelter met the description of an urban house structure as described by Gengler (2012). It was used for the sole purpose of a domestic violence shelter, and housed 12 adult women with up to four children each. Each family unit had access to a private space, and shared communal areas such as the living room, dining room, bathrooms, kitchen, and outdoor play area with other shelter residents. The shelter’s façade blended into a residential neighborhood, with easy access to public transportation, schools, and commercial areas. After 18 years of operation, in 2012, the shelter was closed indefinitely for lack of funding.

The main purpose of the shelter was to offer women who experienced partner violence a safe place to stay for up to three months (only in exceptional cases could they further their stay). In addition to accommodation, food, and personal effects as required, the shelter offered individual therapy, socio-legal services, occupational therapy, mother-child workshops, and access to health care. It also offered educational support for children. As indicated in its contract with the City, this shelter was available for low-income women only. Furthermore, due to lack of specialized staff and facilities, the shelter was not able to accept women with disabilities or substance abuse problems.

Data Collection

We reviewed the case files of all the women who accessed the shelter between January 2010 and June 2012. A total of 219 files were reviewed, but nine were excluded from the sample because of missing data. We collected information from the forms used by the shelter from intake point until discharge. Files included sociodemographic information, assessment of socio-emotional functioning, history of abuse, self-reported needs at entry point and during shelter stay, progress notes written by professional staff, resident complaints or comments regarding services, and sometimes follow-up records. The information was de-identified and entered into an SPSS database. The project was approved by the University Institutional Review Board, and the non-profit organization’s review board.

Measures

Dependent Variable

Length of Stay

This variable was calculated by using the number of days that each woman stayed in the shelter. We obtained the entry and exit dates for each woman and her children from the shelter records. The time between these dates determined length of stay, measured in number of days.

Independent Variables

Life Constraints

The shelter did not collect information about housing difficulties and income. All the women who received shelter services were on a low income, as this was one of the shelter’s acceptance conditions. We focused our analysis of life constraints based on two variables: education (0 = did not complete high school, 1 = completed high school or higher) and employment status (0 = unemployed, 1 = employed).

History of Abuse

Information was collected from the files regarding 1) physical abuse, 2) sexual assault, 3) threats of harm with a weapon, and 4) emotional/verbal abuse. We computed the variable Various types of abuse by adding the number of types of abuse that each woman experienced (i.e., from 0 to 4).

Situational Needs

Information was collected in relation to the needs women reported during their interactions with professionals at entry point or during their stay at the shelter. Shelter residents were asked an open-ended question about their needs at entry point but could report other needs during their stay. These were recorded on the same form and included physical health, psychological support, social support, employment, legal support, and support with children. These variables were recorded dichotomously (0 = no, 1 = yes). In order to test our hypothesis, we computed the variable Level of Needs, defined as the total number of needs, by adding these six variables, which gave us a scale representing how many needs the shelter should attempt to meet.

Contextual (Shelter-Specific) Factors

Information was recorded regarding the services used by adults and children during the stay in the shelter such as socio legal services, health services, psychotherapy, parent-children workshops, and occupational therapy. These services were optional for all shelter residents. Socio legal services involved interactions where professionals worked to create links between the person and agencies, including access to the few free legal-aid services available in the city. Health services consisted of obtaining medical appointments for users and sometimes accompanying users (adults or children) to their appointments. Parent-children workshops were spaces where mothers and children interacted in age-appropriate supervised activities, occurring regularly once a week. Occupational therapy consisted of identifying women’s strengths and difficulties when trying to obtain employment, working on interview skills, and connecting women to potential employers.

In order to test our hypothesis, we created a set of variables to indicate whether women used services at the shelter that met their needs. For example, if a woman reported needs related to employment issues and attended occupational therapy, we set the variable Employment Match to 1, and zero otherwise. We then created a variable (Percentage of needs met) to record the extent to which the number of needs were met by the number of services offered by the shelter, calculated as the number of services matched divided by the number of self-reported needs.

Information was also recorded regarding Difficulties in adapting to the shelter: we recorded whether users or professionals reported difficulties in terms of adapting to the shelter (e.g., inconformity with rules) or whether women experienced any problems with other shelter residents. This variable was recorded dichotomously (0 = no problems reported, 1 = at least one complaint from shelter residents or staff).

Controls

We controlled for several individual demographic factors: age (in years), cohabitation with the abusive partner (0 = no, 1 = yes), number of children, and the number of children who entered the shelter with their mothers. In Colombia, race and ethnicity are not recorded unless the person is indigenous.

Data Analysis

The data was analyzed using the Statistical Package SPSS WINDOWS 22. Descriptive and bivariate analyses were run, followed by a series of hierarchical linear regressions. Data were entered in six stages to determine how demographic data, abuse history, situational needs, and contextual (shelter-specific) factors predicted abused women using shelter services for a longer period of time.

Results

Shelter Residents

A total of 210 women and 363 children used the domestic violence shelter between 2010 and 2012. On average, adult residents were 29.56 years old (SD = 8.824, range = 17–63). Two age groups constituted the majority of the sample: 17 to 25 years old (38.3%) and 26 to 35 years old (40.3%). Children who entered the shelter were aged 0 to 16, but 45% were under 5. Of the total sample, 86.5% of women accessed the shelter with their children. Women had from one to up to seven children with them at the shelter. Less than a third of the women in the sample had obtained a high school degree (23.3%; n = 49) but more than half the sample of women (62.9%; n = 132) was employed at shelter entry. Most women (83.8%; n = 176) cohabited with the aggressor, either in common-law relationships or they were married. The average length of cohabitation was of 3.97 years (SD = 3.99, range = 0–25).

Most women (87.1%; n = 183) reported at least one variable contemplated in life constraints (low level of education or unemployment) and 40.5% reported two. Of the total sample, 87.6% of users reported having experienced physical assault, 22.3% indicated having been threatened with a weapon, 85.2% experienced verbal or emotional abuse, and 13.9% had been sexually assaulted. Table 1 provides a summary of the characteristics of shelter residents and their experiences of abuse.

Table 1 Characteristics of shelter residents
Full size table

Women in the shelter reported having numerous needs, among which the most frequent were employment needs, social support needs, and needs related to children. Figure 1 presents all the needs reported by users. During their stay at the shelter, women used different services: 91.9% attended individual psychotherapy, 90% attended socio legal services, 64.8% attended occupational therapy, 27.1% attended parent-children workshops, and 26.3% attended health services. In terms of adapting to the shelter, 19.6% of the shelter residents reported some kind of difficulty during their stay at the shelter.

Fig. 1
figure 1

Shelter users´ self-reported needs

Full size image

The average length of shelter stay was 39.66 days (SD = 34.036; range = 1–141). One fifth of the women had left the shelter within a week, and half the women in the sample had left the shelter within a month.

Analyses

Correlation analyses (see Table 2) show that multiple variables relate significantly to the length of stay. From the demographic factors, age was positively related to length of stay (r(196) = .216, p < .01) as was the number of children (r(207) = .242, p < .01) but, surprisingly, not the number of children in the shelter. Among life constraint factors, the correlation between education and length of stay was significant, (r(207) = .242, p < .01) but employment was not. Experiencing various types of abuse was not significantly associated with length of shelter stay. The correlation between level of needs and length of shelter stay was statistically significant in the expected direction (r(183) = .363, p < .001). The proportion of needs met by the shelter services was also significantly and positively correlated with length of stay (r(182) = .484, p < .01). The correlation between difficulties in adapting to the shelter and relationships in the shelter was negative but not significantly related to length of stay.

Table 2 Means, standard deviations, and correlations
Full size table

Multivariate Linear Regressions

In order to provide empirical tests for our hypotheses and understand the combination of factors that affect women’s length of stay in the shelter, we used multivariate hierarchical linear regression analyses. In these, we controlled for age, cohabitation, number of children, and number of children entering the shelter in all models. Results are reported in Table 3.

Table 3 Hierarchical linear regression analysis for variables predicting length of stay
Full size table

Model 1 included only the control variables; it showed that only age was significantly and positively related to length of stay (β = .231, p < .05). Employment status, cohabitation, number of children, and the number of children that entered the shelter were not significantly related to length of stay.

  • Hypothesis 1. Model 2 included the variables of life constraints and constituted a test for Hypothesis 1. Of the two life constraint variables (education and employment), only education was significantly related to an increased length of stay (β = .165, p < .05); however, the effect was positive (i.e., more educated women stay longer at the shelter), when we were expecting a negative relationship. Hence, our Hypothesis 1 was not supported.

  • Hypothesis 2. Model 3 tested Hypothesis 2, which predicted that women who had experienced more types of abuse would stay longer at the shelter than women who experienced fewer types of abuse. The results did not support this relationship.

  • Hypothesis 3. Model 4 tested Hypothesis 3, which predicted that women with more needs would stay at the shelter longer than women with fewer needs. The results supported this association as the women’s levels of need were positively and significantly related to the length of shelter stay (β = .294, p < .01).

  • Hypotheses 4a–b. Model 5 tested Hypotheses 4a and 4b. Hypothesis 4a stated that the more needs are met through the use of specific services at the shelter, the more likely women are to stay longer. We found support for this hypothesis as the percentage of needs met by services was significantly and positively related to length of stay (β = .516, p < .01). Hypothesis 4b, which stated that women who do not report difficulties during their stay in the shelter would stay longer, was not supported.

Finally, Model 6 combined our different hypotheses together in one model and identified the best set of variables that predict length of shelter stay. Model 8 explains 30.5% of the variance in women’s length of stay at the shelter, which is explained mainly by their education levels (β = .183, p < .05) (Hypothesis 1), their level of needs (β = .262, p < .01) (Hypothesis 3), and the meeting of these needs through specific services (β = .399, p < .01) (Hypothesis 4a).

Additional Analyses

In the additional analyses reported in Table 4, we examined which specific needs (Model A1), and the meeting of which needs (Model A2) would predict a longer stay in the shelter. In Model A1, we included needs for social support, needs related to children, employment needs, legal needs and emotional support needs. We found that, of the needs reported by women, only legal needs were significantly and positively related to an increased length of stay (β = .209, p < .05). In Model A2, we examined whether meeting specific needs would be related to the length of stay. We found that meeting legal needs (β = .180, p < .05), needs related to children (β = .289, p < .05) and needs related to employment (β = .246, p < .05) were significantly and positively related to an increased length of stay.

Table 4 Additional Analyses. Hierarchical Linear Regression Analysis for Variables Predicting Length of Stay
Full size table

Robustness Checks

Due to missing data points, we lost many cases in all of the models and especially in Model 6. We conducted a t-test to check whether women for whom we had missing values for Model 6 differed significantly in terms of length of stay from the rest of the population. The t-test was significant, meaning that women for whom we had missing values tended to have shorter stays at the shelter than those for whom we had complete data. In order to address this issue, we employed three analytical strategies. First, instead of using listwise deletion of cases based on missing values, we used the option to replace the missing value with the mean of the variable, or to delete the missing values pairwise. Using these two options, the results presented in Model 6 were supported. Second, we removed the variable controlling for the number of children that accessed the shelter because it was one of the variables with most missing data. Removing this variable increased the number of valid cases to 134 and our pattern of results was confirmed. When using this strategy, an additional variable (Difficulties in the shelter) became significant (and negative, as expected). Third and finally, we used a multiple imputation procedure in which we generated five datasets with different randomly imputed values for each missing case and variable, and then ran the estimation on each dataset and averaged the parameter estimates. This method of handling missing data is considered robust (Allison 2012) and our pattern of results was replicated.

Additionally, because approximately 20% of shelter users left the shelter within 7 days, our results may be biased towards considering the use of shelter services as more important (i.e., women who stay only for a few days would not have had the time to use the services). To ensure that our results were not affected by very short stays, we replicated our models after removing all cases with a length of stay inferior or equal to 7 days. The results obtained in these robustness checks replicate the pattern of results obtained in the main dataset, indicating that women who stayed for short periods of time are not influenced by factors that are different from those that influence the women who stayed for longer periods of time.

Discussion

This study examined life constraints, abuse history, situational needs and contextual (shelter-specific) factors that predict whether abused women use shelter services for longer periods of time. Results indicated that having a higher level of education, more needs, and having a higher number of those needs met by the shelter’s services were associated with a longer stay in the shelter. These three variables explained about 30% of the variance of women’s shelter stay. Conversely, we did not find that employment status, history of abuse, and having difficulties in the shelter were related to length of stay.

The current study adds to the literature by identifying demographic, situational, and contextual factors that influence the length of shelter stay. We based our hypotheses and empirical analysis on the assumption that factors that have been shown in the literature to predict the use of shelters by women victims of abuse would also predict their length of stay in the shelter. Interestingly, our results show that while some factors that are important when requesting a place in a shelter predict the length of stay in the shelter (such as education), others do not (such as the type of abuse experienced).

Our results did not provide support for the hypothesis that life constraints (unemployment, less education) would increase women’s stay in the shelter. In contrast to findings examining women who received shelter services (Grossman and Lundy 2011), women who were more highly educated tended to stay longer. It is possible that life constraints do not influence length of stay in the expected direction because not all women who access shelters strive to be independent from abusers. Previous research has shown that women who access shelters have different goals when using the shelter and different intentions regarding staying with or leaving the abusive partner, such as respite and transition (Krishnan et al. 2004; Tutty 2015). Another possibility is that life constraints do not influence length of stay in the expected direction because shelters are considered a valuable option and not a last resort. Also, in contrast to literature showing that women who experience more types of abuse are more likely to access shelters, our findings showed that they do not tend to stay for longer (or shorter) durations at the shelter than women who experience fewer types of abuse. Most of the sample had experienced physical and emotional abuse, and experiencing various types of controlling behaviors was not related to staying longer in the shelter. However, this result is limited, as we did not examine the severity of the abuse.

Similarly, most of our models found that age plays an important role when seeking help from a shelter, even though the effect was not significant in the final model. However, in contrast to the findings of studies focusing on people requesting or receiving shelter services (George et al. 2010; Grossman and Lundy 2011), we found that women who stay longer at the shelter are older than those staying for shorter periods. Two of the factors that may explain this finding are as follows: a) it is possible that women have a clearer disposition to leave the abusive partner as they get older, and b) it is possible that women get to better understand the system as they get older and use more of the available services in order to gain independence.

By contrast, we found that some of the factors that predict the likelihood of using a shelter also predict length of stay. Shelter residents with more needs stayed in the shelter for significantly longer periods. Needs included physical health, psychological support, social support, employment, legal support, and support with children. These results add empirical evidence to the literature showing that clusters of needs will influence women’s likelihood of becoming independent from their abuser (Krishnan et al. 2004). We also identified factors specific to the shelter that influence women’s length of stay at the shelter. Specifically, the percentage of needs met by services was significantly and positively related to length of stay. This finding is consistent with previous research findings showing that women who experience IPV have many needs (Allen et al. 2004; Lyon et al. 2008), and provides further evidence for the need to offer comprehensive services in shelters.

The importance of legal needs among shelter residents has been identified in previous studies (Allen et al. 2004). In our study, having legal needs was the only separate need that predicted a longer length of stay. We explain this finding by suggesting that legal needs are related to women’s intentions to separate from their abuser. Staying longer at the shelter could be an opportunity to understand legal proceedings and obtain legal advice. Furthermore, the use of socio legal services, employment services, and parent-children services by women who had expressed those needs was significantly associated with their length of stay in the shelter. As in the case of legal needs, results that highlight the importance of socio legal and employment services might indicate women’s intentions to become independent from their abusers. They are also consistent with previous research showing that children’s needs are key in women’s decision to access a shelter (Clevenger and Roe-Sepowitz 2009; Oths and Roberston 2007; Sabina et al. 2014). We interpret our findings as showing that when women receive shelter services, they take the opportunity to address their concerns about their children. This interpretation is consistent with studies that have documented that some women who experience IPV feel that the abuse has had negative effects on their parenting, and perceive that their relationship with their children has suffered because they have less energy or feel anger towards their children or themselves (Levendosky et al. 2000). Staying in the shelter and participating in parent-children workshops would provide a means to repairing the relationship.

Finally, our study contributes to the literature by offering a way to reconcile inconsistent findings regarding the importance of shelter stay. Recently, findings by McFarlane et al. (2014) questioned the importance of the length of shelter stay by showing reduced measures of re-abuse and increased measures of women’s mental health functioning after four months of shelter stay regardless of the time spent at the shelter. The authors explained their findings by arguing that contact with the shelter might have empowered women and thus improved their mental health. While this explanation seems reasonable, we suggest that differences between residents may be found if, in addition to length of shelter stay, the study had included information about the use of different support services during the stay.

Limitations of the Study

The results of this study have to be critically considered in light of the following limitations. First, our findings regarding the relevance of situational and contextual factors may apply only to social contexts where there are few service choices available for women who experience IPV (as in most developing countries), and collectivistic cultural values prioritize the needs of family members over individual needs. Nevertheless, we point out that victims´ multitude of needs and the key role of children in women’s decision to access help were first identified in research conducted in other, more individualistic, cultural contexts such as the United States and Canada. It is therefore possible that our findings are not unique to the case of developing and collectivistic countries. There are also a number of limitations in terms of the generalizability of the results given that, although we considered all women who received shelter services in the city of Bogotá from 2010 to 2012, the conditions of the shelter contract with the City limited services to women who had reported their experiences of partner violence to the police and had a low level of income.

A second limitation is that the data for this study relied on an archival review of case files. We collected data from institutional forms completed by service providers at the shelter that had not been validated or tested for reliability. As information in the file was commonly collected by several professionals using semi-structured interviews, there was some variance in the amount of detail included in the file. Similarly, the shelter determined the ways in which data was collected as part of their services, and we had no control over what was left in and what was left out. For example, examining the severity of the abuse and housing needs is paramount for women who experience IPV, but these issues were not reported in the files. It is possible that professionals did not ask these questions because they assumed that all women were severely abused and had housing needs, otherwise they would not have requested a shelter stay. It is also possible that, as social services that respond to housing needs for the population do not exist in Colombia, women only reported the needs which they expected the shelter could help them with.

Third, the study examined information from one DV shelter with a specific average length of stay. Nevertheless, as it was the only shelter available for the whole city of Bogota from 2010 to 2012, the sample contains all women who accessed a shelter in this city between these dates. It is also the first study to identify the needs of women who access a domestic violence shelter in Colombia.

Avenues for Future Research

To our knowledge, this study is the first to explicitly attempt to identify the factors that affect women’s length of stay in shelters. Given the importance of understanding these factors in terms of resource allocation for services from the perspective of shelters and in terms of increased well being for users, we believe that there is a great need for additional research to both replicate our findings and extend our understanding of the factors that influence length of stay.

First, future research should attempt to replicate our study in different contexts in order to capture additional factors that may impact women’s length of stay. We were not able to collect information regarding the severity of the abuse, the women’s expectations of services, readiness for change, and perceptions of risk. The relationship between some of these factors and the women becoming independent from their abusers is complex. For example, some studies have found that women who are abused only verbally or emotionally are more likely to become independent (Hilbert et al. 1997). Likewise, previous research has identified that women can predict future victimization with relative accuracy, depending on whether they intend to leave their abuser or continue the relationship (Harding and Helweg-Larsen 2009). This in turn might be related to their length of stay in the shelter. This information needs to be collected at various points in time during the stay, as perception of risk can also change as a result of participating in interventions focusing on risk and safety evaluation.

Second, future research should also evaluate or include the evaluations of shelter services and interventions, the extent to which length of stay matters for outcomes, and the extent to which benefits continue after their stay. By looking at the relationship between length of stay and outcomes, researchers could offer suggestions regarding the duration of intervention programs, help improve program completion, and better predict the impact of services.

Implications for Practice

Our study found that a number of demographic, situational and contextual factors influence women’s length of stay in a domestic violence shelter. These findings have several implications. First, identifying the different characteristics between women who leave shortly after they enter a shelter and those who stay longer may help practitioners predict length of stay and thus improve shelters´ ability to plan and implement programs. There is a loss of resources every time goals are set for months of psychosocial interventions, which are not achieved as the women leave after only a week of their stay. Similarly, the effectiveness of some interventions can be affected when programs are condensed to adapt to short stays and rates of attrition. Being able to predict who is more likely to stay longer can help practitioners set realistic goals for each client and offer programs that are more likely to be completed.

Second, this study provides some insights about the services that could be offered in shelters. Our findings support practitioners´ calls for offering women comprehensive services in shelters by showing that matching women’s needs with services will increase their stay. However, this may be unrealistic. Shelters are commonly under-funded (Roberts and Lewis 2000), and services are often too few (Wathen et al. 2015). Our findings suggest that, among services, offering legal, employment, and mother-child programs is key. It does not claim that these services are more important than others, such as counseling or health care. Rather, it suggests that some women might see shelters as much more than a refuge. They will stay longer in order to take advantage of the opportunity to explore and understand the legal issues regarding their situation, improve their chances of employment, and work on their relationship with their children. The evidence provided by this study supports previous claims that shelters are safe places in which women become empowered (Kasturirangan 2008).

Third, the study highlights the importance of giving women residents the possibility of choosing the services they use when they enter a shelter. Although many shelters identify the individual needs of their clients, they also direct victims into appropriate services based on ideas on what is best for them (Chantler 2006). As in any intervention program, motivation is key. While all the services offered in a shelter may potentially benefit a population with multiple needs, women residents have explicit or unexpressed priorities and limited time; as such, they should be able to control those choices. To facilitate survivors´ control over their choices and actions is a primary goal of trauma-informed services (Elliot et al. 2005). In addition, in social contexts where legal and employment services are mostly unavailable or inaccessible, as in the Colombian case, women might stay longer not only because they need these services but because they provide a unique opportunity to access them. The same might be true for mother-child programs, which have been positively evaluated (Keeshin et al. 2015) but are seldom offered in shelters (Wathen et al. 2015). Consequently, it is important to both increase the extent to which individual needs are matched to services in shelters, and to make sure that women are freely engaging in the programs of their choice.

Understanding the factors that predict women staying longer in a shelter is a step towards improving our responses to domestic violence. Future research should further explore the relevance of social and cultural contexts in understanding women’s use of this essential resource.