Risk of Behaviors Associated with Lethal Violence and Functional Outcomes for Abused Women Who Do and Do Not Return to the Abuser Following a Community-Based Intervention

Abstract

Background: To determine the differential risk of behaviors associated with lethal violence and functioning outcomes for abused women with children who received an intervention of shelter or justice services and return to the abuser were compared with women who did not return.

Methods: The study was carried out using a matched-pair analysis of 152 women. Measures of risk of behaviors associated with lethal violence, type, and severity of abuse, and poor mental health symptoms were measured every 4 months for 24 months.

Results: Irrespective of service used, women returning to the abuser reported significantly (p<0.001) more risk of behaviors associated with lethal violence compared to women who did not return. Additionally, severity of physical abuse was significantly higher (p<0.01) for women returning to the abuser who had been in a shelter but not for women who received a protection order. Similarly, levels of depression, posttraumatic stress disorder, and anxiety were significantly higher (p<0.01) for sheltered women who returned to the abuser but not for women using a protection order who returned.

Conclusions: Risk of behaviors associated with lethal violence increases when abused women return to live with the abuser. Abused women should be informed of the heightened risk and greater probability for poor mental health.

Introduction

Among women in the United States (U.S.) and worldwide, an estimated 30% have experienced intimate partner violence.¹ Consequences of violence include acute trauma,¹ poor health,² and compromised functioning³—which is frequently associated with mental health problems of depression and posttraumatic stress disorder (PTSD),⁴ and chronic pain.⁵ The impact of abuse on women's health is extensive. Bonomi and colleagues⁶ completed telephone surveys to determine the women's abuse status and then, using the women's U.S. health plan medical records, compared the health information for those abused in the past year (n=242) with those who were never abused (n=1686). After controlling for age, they found that women reporting abuse had a greater relative risk for psychological/mental problems, musculoskeletal problems, female reproductive disorders, acute respiratory tract infections, gastroesophageal reflux disease, chest pain, abdominal pain, urinary tract infections, headaches, and contusions, abrasions, and lacerations compared with nonabused women.

When pregnant women are abused, both mother and fetus can be impacted with significantly higher rates of spontaneous abortions,⁷ preterm labor,⁸ lower birthweight infants,⁹ stillbirths,¹⁰ and slower infant growth following birth.¹¹ Additionally, abuse during pregnancy places the mother in higher danger for lethality. The results of a 10-city, case control study in the U.S., women abused during pregnancy were at a threefold risk of being murdered by an intimate partner.¹² Children living in homes with a mother who reports abuse during pregnancy show significantly more behavioral dysfunctions for up to 2 years following the pregnancy compared with youngsters living with a mother who reports abuse, but not during pregnancy.¹³

Children who witness the abuse of their mothers are at higher risk to depression,¹⁴ PTSD,¹⁵ behavioral and functional disorders,¹⁶ and academic problems.¹⁷ A meta-analysis of 118 studies concluded that children who witness domestic violence showed more psychological problems than children who do not.¹⁸ Very young children are disproportionately exposed to intimate partner violence as compared to older children,¹⁹ which increases the risk of behavioral dysfunction problems that can be internal (i.e., anxiety, depression, withdrawal) or external (i.e., aggression, hostility).²⁰ A meta-analysis conducted by Wolfe and associates²¹ indicated a positive relationship between children's exposure to violence against their mother and internalizing and externalizing behavioral problems of the child. A study of 300 youngsters of abused women who received services for abuse found boys who had seen the abuse of their mother significantly more likely to display externalizing and internalizing behaviors consistent with disturbed male youngsters under clinical care when compared to boys of abused mothers who did not see the abuse.²² A recent dyadic analyses of 300 mother child pairs found mothers who reported clinical and borderline clinical internalized problems (i.e., depression, anxiety) were 7 times more likely to have children with the same problems and mothers with borderline clinical and clinical external problems (i.e., aggression, hostility) were 4.5 times more likely to have children with the same external problems.²³ These dyadic analyses provide evidence of a direct relationship of maternal functioning on child behavioral functioning.

Although the health impact of partner violence is extensive and crosses generations, evidence-based interventions are few, and long-term outcomes of lethality risk and functioning scant. Two major community-based interventions for abused women are protection orders, obtained through nongovernment, and government justice agencies, and designated safe shelters for abused women, also known as domestic violence shelters. The United States Health and Human Services website on safety planning specifies both obtaining a protection order and safe shelter as part of safety planning.²⁴ While many abused women use safe shelter and protection orders, there is little long-term (i.e., multiyear) cohort evidence to inform abused women, public health practitioners, and policy makers as to interventions effectiveness.

The World Health Organization (WHO) guidelines for policy and clinical practice for partner violence call for evidence-based interventions that are culturally sensitive and effective for low, medium, and high resource countries.²⁵ The Institute of Medicine²⁶ calls for evidence for strategies to decrease violence against women and children. The recent U.S. Preventive Services Task Force revised their recommendations, now a B recommendation that supports all health care providers screen all women between the ages of 14 and 46 for intimate partner violence and provide or refer to services that offer support to abused women.²⁷

Implied in offering referrals is that health care providers have knowledge about the availability of the resource and expected outcomes. For informed decision making by abused women and evidence-based referrals from practitioners who screen and advise abused women, outcomes are needed following the use of community interventions, such as safe shelters and protection orders for women who return and do not return to the abuser. Although we did not identify a published report on differential outcomes, reports exist on characteristics of women who leave an abusive relationship, as well as some information on the frequency with which women return to the abuser, and abuse following return to the abuser.

Predictors of leaving/returning to an abusive relationship and revictimization

Women who experience partner violence frequently report leaving and returning to the abusive relationship multiple times.²⁸ Predictors of women likely to leave an abusive relationship include young age, history of leaving the relationship, having a protection order, an abuse-related health visit,²⁹ higher education,³⁰ and higher use of community resources.³¹

Abused women report going back to the abuser due to abuser's harassment, feelings of guilt, attachment to abuser, and encouragement by relatives.³² In addition, women who are more forgiving to their abuser are more likely to report intention to return to them.³³ Bonomi and colleagues³⁴ used constructivist grounded theory to understand the underlying coercive dynamics of perpetrator appeals to the victim to recant her report of abuse to authorities. Using audio-recorded telephone calls between domestic violence perpetrators and victims to answer questions about how and why victims arrive at their decision to recant and/or refuse prosecution efforts, 17 abusers and their victims were interviewed. Results showed that a victim's recantation intention was foremost influenced by the perpetrator's appeals to the victim's sympathy through descriptions of his suffering from mental and physical problems, intolerable jail conditions, and life without her. Goodman³⁵ discusses similar factors, including fear.

Regarding further abuse when women return to the abuser after using a shelter, 43% of 141 women exiting a shelter reported reabuse 6 months later.³⁶ Other researchers report that 3 years post shelter, 81% of 124 sheltered women did not return to the abuser; however, 19% of women reported being assaulted and 22% being harassed by the abuser.³⁷ In the same study,³⁷ women who worked with advocates after leaving the shelter experienced less violence over time, reported higher quality of life and social support, and had less difficulty obtaining community resources. More than twice as many women receiving advocacy services experienced no violence across the 2 years post intervention period compared with women not receiving the advocacy services.

Regarding protection order applicants, a review of 15 empirical studies of outcomes concluded that there is evidence that receiving a protection order reduces the victim's risk for further violence.³⁸ Holt and colleagues³⁹ compared victims of intimate partner violence who reported the abuse to the police and obtained protection orders with women who reported to the police only. Findings indicated that women who had protection orders at baseline were significantly less likely than those who did not to be contacted by the abuser, to experience injury or weapons threats, and to receive abuse-related medical care between the intake and 1 year later. Holt and colleagues⁴⁰ also found permanent protection orders were associated with an 80% reduction in reports of physical violence during the 12 months after the victim made a first report of intimate partner violence to the police, compared with temporary orders, which made no difference in the risk of repeat violence. Fatalities did not correlate with protection order status and the researchers caution that unmeasured characteristics of women who obtain protection orders may be responsible for the observed effect, rather than the orders themselves.^39,40 A study of 150 women who qualified for a protection order found significantly lower levels of violence over 18 months, regardless of whether the women actually received the order.⁴¹ Outcomes specific to whether or not the women were living with the abuser were not found for protection order recipients.

Study purpose

The purpose of this study is to describe the differential outcomes in danger for lethality and abuse severity and mental health (e.g., PTSD symptoms, anxiety, somatization, and depression), as a function of whether or not an abused woman with children returns to live with the abuser within 24 months of using safe shelter or protection orders for the first time. The functioning of the children in each group was also measured.

Methods

Design

A matched-pair analysis is followed, with repeated measures every 4 months for 24 months.

Setting

The study is occurring in a large urban metropolis in the United States with a population exceeding 4 million. Five shelters designated for abused women with a collective bed capacity of 400 for women and their children serve the population, along with a central District Attorney's Office for processing of protection orders. The shelters offer standard services of safe shelter, counseling, and advocacy. Services of the District Attorney's Office are legal processing for a protection order.

Population and sample size determination

Our eligible population is English- and Spanish-speaking abused women seeking a shelter for abused women or justice services, specifically a protection order, for the first time that had never used either shelter services or applied for a protection order in the past. Additionally, qualifying women had at least one child between the ages of 18 months and 15 years who lived with the mother at least 50% of the time. Sample size was determined for two independent samples (i.e., sheltered women and protection order applicants) with a conservative effect size of 0.40, a power of 0.90, and alpha of 0.05, 135 women were needed in each group. Allowing for attrition, we set the sample at 150 women and 150 children in each group, for a total of 300 women and 300 children. If the woman had more than one child between the ages of 18 months and 15 years, each child was given a number according to birth order and a die was rolled to select one child at random to be followed in the study.

Procedures

Following internal review board approval, recruitment began at five local shelters for abused women and the District Attorney's (DA) Office. Trained researchers who speak both English and Spanish approached all women entering the shelter or applying for a protection order and established eligibility criteria. If the woman was eligible, she was taken to a private room and invited to participate in the study. After signing informed consent and completion of the 60-minute interview, all 300 women were offered $30 cash. The interview could be completed in 60 minutes in both languages. The researchers recorded the woman's responses. Recruitment, eligibility screening, and entry into the study continued daily for 13 months at the 5 shelters and Monday through Friday at the DA's Office. Over the 13 months, 330 women met eligibility criteria. A total of 19 eligible women refused to participate, usually due to a lack of time, and 11 women were missed. A total of 300 women (e.g., 150 at the shelters and 150 at the DA's office) met eligibility criteria and agreed to participate in this study.

Participants

Participants were recruited over a 13-month period, and a total of 330 women met eligibility criteria. A total of 19 eligible women refused to participate, usually due to a lack of time, and 11 women were missed (i.e., left the shelter or DA's Office before the study could be explained). A total of 300 women (e.g., 150 at the shelters and 150 at the DA's office) met eligibility criteria and agreed to participate in this study. A total of 17 women dropped out of the study due to lost contact with the researchers.

Of the 300 mother–child dyads who were recruited in the overarching study, a total of 76 women reported that they had returned to their abuser at some point in the first 2 years of this study—52 women (35%) in the shelter group and 24 women (16%) in the protection order group. To investigate potential differences in outcomes as a function of whether or not a woman returns to the abuser, a case-control, matched-pair sample was created. Women who reported returning to their abuser were matched by intervention type (e.g., protection order [PO] or shelter), age, ethnicity, immigrant status, education, and child age and gender with women who did not report return to the abuser. Intervention group, child gender and age, ethnicity, and immigrant status prioritized matching. These demographic factors were chosen as matching characteristics due to found group differences reported elsewhere in order to control for any potential demographic differences across groups. Participants were matched as closely as possible with the given data set and following the match process, the matched pairs were tested for differences in key demographics, yielding no significant differences across demographics of those who did and did not return to the abuser.

The final matched sample consisted of 152 mother–child dyads (n=104 [shelter]; n=48 [PO]), with half of the women in each group reporting that they had returned to the abuse. The demographics, by intervention group, are presented in Table 1.

Table 1.

Descriptives of Final Sample by Intervention Group

	Shelter		DA
	n	%	n	%
Returned to abuser
No	52	50.0	24	50.0
Yes	52	50.0	24	50.0
Immigrant status
U.S. born	56	53.8	34	70.8
Immigrant	48	46.2	14	29.2
Child age
1.5–5 years	59	56.7	26	54.2
6–18 years	45	43.3	22	45.8
Child gender
Boy	55	52.9	26	54.2
Girl	49	47.1	22	45.8
Education
Less than high school	47	45.2	8	16.7
High school/GED	19	18.3	16	33.3
Some college	36	34.6	19	39.6
College degree	2	1.9	5	10.4

	n	M	SD	n	M	SD
Woman's age	104	30.41	7.02	48	29.77	6.80
Length of relationship (months)	104	88.13	61.33	48	77.27	64.00
Total days returned to abuser	51	327.24	238.24	24	187.75	217.83

Percentages not summing to 100 reflect missing data.

DA, district attorney; GED, general educational development; M, mean; SD standard deviation.

Measures

To establish temporal sequencing, the measures were asked at entry into the study, prefaced with “During the last 4 months” and repeated every 4 months for 24 months, prefaced by “Since we talked on [date of last visit].” A 4-month interval was selected for maximum stability of the child functioning instruments and to minimize recall bias. All measures were completed within 48 hours of entry into the shelter or application for a protection order.

Danger for risk of behaviors associated with lethal violence and abuse severity

The Danger Assessment Scale⁴² is a 19-item questionnaire with a yes/no response format that is designed to assist women in determining their potential risk for becoming a femicide victim. All items refer to risk factors that have been associated with murder in situations involving abuse. Examples of questions include “Has the physical violence increased in severity or frequency?” and “Has the abuser forced the woman to have sex?” Convergent construct validity of the instrument has been supported by correlations in the moderately strong range, with instruments measuring severity and/or frequency of abuse.⁴³ Initial reliability of the instrument was 0.71⁴² and ranged from 0.60 to 0.86 in five subsequent studies.⁴³ Weighted scoring results in four ranges of danger: <8, variable danger; 8–13, increased danger; 14–17, severe danger; and ≥18, extreme danger.

The Severity of Violence Against Women Scale⁴⁴ is a 47-item instrument designed to measure threats of abuse (19 items) and physical abuse (28 items). Included are nine factors or subscales that have been demonstrated valid through factor analytic techniques: symbolic violence and mild, moderate, and serious threats (threats of violence dimension) and mild, minor, moderate, serious, and sexual violence (actual violence dimension). For each item, the woman responds using a 4-point scale to indicate how often the behavior occurred: (1) never, (2) once, (3) 2–3 times, and (4) 4 or more times. The possible range of scores is 19 to 76 for the threats of abuse and 28 to 112 for physical assault. Initial internal consistency reliability estimates ranged from 0.92 to 0.96 for a sample of 707 college female students and from 0.89 to 0.96 for a scale of 208 community women.⁴⁴

Mental health functioning

The Brief Symptom Inventory-18⁴⁵ (BSI-18) is an abbreviated version of the 53-item Brief Symptom Inventory,⁴⁶ which in turn is a shortened form of the 90-item Symptom Checklist-90-Revised (SCL-90-R).⁴⁷ The BSI-18 is an 18-item self-report scale of statements that test-takers respond to based on their level of distress over the preceding 7 days (five-point Likert-like scale from 0, not at all, to 4, extremely). The total Global Severity Index (GSI) score ranges from 0 to 72 with higher scores indicating higher levels of psychological distress.⁴⁵ Reported internal consistency estimates are acceptable (0.74 for somatization, 0.79 for anxiety, 0.84 for depression, and 0.89 for the total GSI). Concurrent validity with the SCL-90-R is high, ranging from 0.91 to 0.96 on both dimensions and total scores.

The Post-Traumatic Stress Disorder Scale⁴⁸ is a seven-item symptom scale screens for PTSD and is a subset of items from the National Institute of Mental Health Diagnostic Interview Schedule for for PTSD. The seven items were empirically derived in the context of an epidemiological study of PTSD in an urban area of the United States. The seven items selected were those that most efficiently predicted PTSD diagnostic status.⁴⁸ The screen consists of five avoidance items and two hyperarousal items. When the seven-item scale was evaluated for predictive validity in a National Epidemiological Survey, a score of four or more on the seven-item scale identified cases of PTSD with sensitivity of 78%, specificity of 97%, positive predictive value of 75%, and negative predictive value of 98%. The percentage of correctly classified respondents was 96%.⁴⁹

Child behavioral functioning

The Achenbach Child Behavior Checklist (CBCL) provides a standardized parental report of child behavioral problems,^50,51 with a form for children 18 months to 5 years and a form for youth 6 to 18 years. The CBCL is orally administered to a parent who rates the presence and frequency of certain behaviors on a three-point scale (0, not true; 1, somewhat or sometimes true, and 2, very true or often true). The CBCL consist of two broadband factors of behavioral problems: internalizing and externalizing with mean scale scores for national normative samples as well as clinically referred, borderline clinical, and nonreferred samples of children. Extensive psychometric testing has yielded very favorable information regarding the tool's validity and reliability in English and Spanish.^50,51

Return to abuser

At each time point, women were also asked to indicate whether or not they were in an intimate relationship with their abuser. To establish a baseline, at entry no women reported that they were currently in a relationship with their abuser. Due to limited number of cases of women who return to their abuser across time, return to the abuse was recoded into one dichotomous variable to measure whether or not a woman returned to her abuser at any point during the 24-month period.

Analysis process

Analyses were conducted separately for each group to examine the unique outcomes of women by point of initial contact due to existing group differences.⁵² In order to examine differences in key outcomes by whether or not a woman returned to the abuser, a series of 7 (time) by 2 (group) within subjects analysis of variance (ANOVA) and multivariate ANOVA (MANOVA) tests were conducted. Women had decreases in physical abuse and danger as well as increases in mental health functioning following entry into the study;³⁶ as such, we only examined the main effects of group and the interaction effects of group by times. The measure of effect size associated with ANOVAs and MANOVAs is partial eta squared (η²) value, which is interpreted as small (∼0.0099), medium (∼0.0588), and large (≥0.1379). Significance levels for all analyses were set as .05 and all analyses were conducted in SPSS version 19.0.

Results

Protection orders group findings

There was a significant main effect of returning to the abuser, F(1, 23)=4.51, p=0.045, η²=0.164, and a significant interaction effect of group by time, F(6, 18)=12.26, p<0.001, η²=0.803, on reported levels of danger (see Table 2). Women who returned to the abuser had significantly higher levels of danger compared to those who did not return to the abuser. Further univariate analysis revealed, that women who returned to the abuser had significantly higher levels of danger at entry (mean [M]=16.50, standard deviation [SD]=6.00) compared with those who did not return to the abuser (M=8.00, SD=3.34). At 4 months following entry to the study, women who did not return to the abuser reported higher levels of danger (M=16.71, SD 7.52) compared with those who did return (M=6.46, SD=5.53).

Table 2.

Summary of Findings for DA Group

		Returned to abuser		Did not return
	n	M	SD	M	SD
Danger
Entry	48	16.50	6.00	4.08	3.83
4 Months	48	6.46	5.53	16.71	7.52
8 Months	48	6.58	6.32	6.46	4.52
12 Months	48	6.42	5.44	5.21	3.95
16 Months	48	5.08	4.75	4.54	3.76
20 Months	48	4.29	4.36	4.67	4.39
24 Months	48	6.04	5.04	4.04	3.63
Threats
Entry	48	42.21	13.80	44.38	12.56
4 Months	48	22.50	8.58	21.46	4.30
8 Months	48	23.08	8.65	20.83	5.18
12 Months	48	24.04	9.43	19.17	.48
16 Months	48	21.21	8.16	20.50	5.04
20 Months	48	20.54	4.66	19.17	.64
24 Months	48	22.63	9.38	19.67	2.01
Physical
Entry	48	37.92	12.09	36.29	15.20
4 Months	48	22.67	5.19	21.00	.00
8 Months	48	21.83	2.65	21.00	.00
12 Months	48	23.04	7.66	21.00	.00
16 Months	48	22.08	5.31	21.00	.00
20 Months	48	21.04	.20	21.00	.00
24 Months	48	23.00	7.95	21.42	2.04
Sexual
Entry	48	8.00	3.48	8.00	3.34
4 Months	48	6.13	.61	6.00	.00
8 Months	48	6.25	.85	6.00	.00
12 Months	48	6.21	.72	6.00	.00
16 Months	48	6.00	.00	6.00	.00
20 Months	48	6.04	.20	6.00	.00
24 Months	48	6.13	.61	6.00	.00
PTSD
Entry	48	4.96	1.83	5.33	1.79
4 Months	48	3.67	1.95	4.00	2.32
8 Months	48	3.21	2.17	3.29	1.73
12 Months	48	2.83	2.12	2.88	2.09
16 Months	48	3.13	2.05	2.83	2.55
20 Months	48	2.67	2.22	2.75	1.96
24 Months	48	3.08	1.84	2.38	2.04
Anxiety
Entry	48	11.88	6.84	11.67	7.15
4 Months	48	4.75	4.57	6.33	6.66
8 Months	48	4.46	4.11	5.71	5.94
12 Months	48	4.25	4.34	4.17	6.06
16 Months	48	3.29	3.97	3.79	5.18
20 Months	48	2.92	5.33	2.88	4.08
24 Months	48	3.50	5.78	2.54	2.96
Somatization
Entry	48	7.17	6.13	5.58	5.22
4 Months	48	4.29	5.81	3.17	4.17
8 Months	48	3.54	4.51	2.21	2.70
12 Months	48	3.25	3.89	2.42	3.37
16 Months	48	2.67	2.97	2.46	4.22
20 Months	48	2.13	4.35	2.71	4.30
24 Months	48	2.83	5.49	1.46	2.04
Depression
Entry	48	10.25	6.17	8.63	6.27
4 Months	48	5.54	5.65	4.79	5.57
8 Months	48	6.08	6.77	4.50	4.87
12 Months	48	4.63	5.56	3.67	5.47
16 Months	48	4.21	4.91	2.67	4.63
20 Months	48	3.67	5.69	3.67	5.16
24 Months	48	3.79	5.89	2.83	3.52
Internalizing
Entry	48	56.46	9.91	56.42	12.23
4 Months	48	56.00	8.99	52.96	12.70
8 Months	48	55.67	11.13	52.08	10.78
12 Months	48	52.26	12.43	48.09	11.29
16 Months	48	52.38	10.06	49.63	11.12
20 Months	48	50.67	11.65	48.17	11.20
24 Months	48	52.63	12.15	51.04	12.12
Externalizing
Entry	48	55.67	10.45	55.54	15.19
4 Months	48	54.83	11.84	52.54	11.60
8 Months	48	55.21	13.72	52.46	12.27
12 Months	48	52.87	11.67	47.13	10.07
16 Months	48	51.83	12.30	47.29	11.79
20 Months	48	51.29	11.38	47.42	12.79
24 Months	48	53.58	13.45	49.38	11.29

Means in boldface differed significantly across group, p<0.05; underscored measures indicate a significant main effect of group, p<0.05.

There was a significant effect of time by returning to the abuser on abuse severity, F (18, 6)=16.12, p<.001, η²=0.980. Women who return to the abuser had higher physical abuse at 12 months (M=23.04, SD=7.66) compared to women who did not return (M=21.00, SD=0.00). Women who did not report returning to the abuser had physical and sexual abuse scores at the bottom of the danger scale, with the exception of physical abuse scores at 24 months (M=21.42, SD=2.04), indicating that women who did not return to their abuser reported no measurable levels of physical or sexual abuse.

Regarding mental health, there was not a significant effect of returning to the abuser or an interaction of effect of returning to the abuser by time on measures of PTSD, depression, anxiety, or somatization. Similarly, there was not a significant effect of returning to the abuser or an interaction effect of returning to the abuser by time on child behavioral functioning. Children's functioning did not significantly differ as a function of whether or not their mother returned to the abuser.

Shelter group findings

In terms of danger, there was a significant main effect of returning to the abuser, F (1, 306)=4.11, p=0.048, η²=0.075, and an interaction effect of returning to the abuser by time, F (6, 306)=4.38, p<0.001, η²=0.079 (see Table 3). Women who returned to the abuser had higher reported levels of danger than those who did not return. There were no significant differences in danger scores as at baseline; however, those who returned to the abuser reported higher levels of danger at 4, 8, 20, and 24 months. There was also a significant effect of returning to the abuser on severity of abuse, F (3, 48)=4.86, p=0.005, η²=0.233, and a significant effect of returning by time on abuse, F (18, 33)=2.64, p=0.008, η²=0.590. Those who returned to the abuser reported higher levels of threats than those who did not. Those who returned to the abuser reported higher levels of physical abuse at 4, 8, 20, and 24 months. At baseline, women who did not return to the abuser reported significantly higher levels of sexual abuse compared with those who did return. Those who returned to the abuser reported higher levels of sexual abuser at 4, 8, and 20 months.

Table 3.

Summary of Findings for Shelter Group

		Returned to abuser		Did not return
	n	M	SD	M	SD
Danger
Entry	104	15.69	6.61	16.85	7.76
4 Months	104	8.27	7.51	5.54	4.02
8 Months	104	7.35	5.54	4.46	4.02
12 Months	104	5.81	4.56	5.44	5.11
16 Months	104	5.17	4.91	4.42	5.06
20 Months	104	7.67	7.13	4.40	4.03
24 Months	104	7.21	6.58	4.58	5.22
Threats
Entry	104	42.17	12.96	42.88	15.28
4 Months	104	27.27	10.56	21.73	5.27
8 Months	104	24.17	8.38	19.73	2.01
12 Months	104	22.77	8.22	20.77	4.31
16 Months	104	22.69	7.07	20.58	3.82
20 Months	104	25.87	11.04	19.44	1.27
24 Months	104	23.81	8.79	20.23	4.84
Physical
Entry	104	38.00	15.43	39.33	15.02
4 Months	104	24.92	7.95	21.27	1.43
8 Months	104	23.46	6.29	21.00	.00
12 Months	104	22.52	4.92	21.38	1.65
16 Months	104	22.17	4.44	21.23	1.66
20 Months	104	23.77	7.16	21.17	1.25
24 Months	104	23.29	5.70	21.29	1.38
Sexual
Entry	104	8.29	3.77	10.21	4.75
4 Months	104	7.37	3.17	6.00	.00
8 Months	104	6.77	2.37	6.00	.00
12 Months	104	6.25	1.36	6.06	.42
16 Months	104	6.35	1.41	6.00	.00
20 Months	104	6.75	2.19	6.00	.00
24 Months	104	6.50	1.64	6.10	.69
PTSD
Entry	104	5.50	1.81	4.67	1.99
4 Months	104	3.69	2.17	3.24	2.02
8 Months	104	3.65	2.07	2.69	1.62
12 Months	104	3.42	2.19	2.85	1.90
16 Months	104	3.08	2.01	2.63	1.76
20 Months	104	3.48	2.16	2.35	1.90
24 Months	104	3.31	2.10	2.62	1.76
Anxiety
Entry	104	11.29	7.18	9.46	7.15
4 Months	104	7.02	7.30	4.63	5.72
8 Months	104	5.37	6.53	3.48	4.64
12 Months	104	5.06	6.19	3.65	5.21
16 Months	104	3.71	5.48	3.15	5.07
20 Months	104	5.02	5.86	3.79	6.26
24 Months	104	4.37	5.39	4.25	4.99
Somatization
Entry	104	7.73	6.21	7.27	6.72
4 Months	104	5.31	6.47	3.71	4.78
8 Months	104	4.31	5.47	2.79	3.85
12 Months	104	3.54	4.54	3.29	5.14
16 Months	104	3.29	5.07	3.33	5.29
20 Months	104	3.88	4.80	3.46	5.52
24 Months	104	2.92	3.92	3.54	4.47
Depression
Entry	104	12.38	6.55	9.42	6.75
4 Months	104	7.63	7.06	4.78	5.43
8 Months	104	5.98	6.44	4.04	5.14
12 Months	104	5.98	6.40	4.12	4.99
16 Months	104	4.35	5.84	3.37	4.41
20 Months	104	6.00	7.05	3.04	5.13
24 Months	104	4.54	5.88	4.00	4.62
Internalizing
Entry	104	57.67	10.21	58.83	11.22
4 Months	104	56.10	11.46	55.98	11.24
8 Months	104	54.04	10.87	54.42	11.70
12 Months	104	54.17	11.03	54.13	12.61
16 Months	104	52.76	11.67	54.78	13.04
20 Months	104	54.83	13.23	54.98	13.04
24 Months	104	52.94	12.88	55.21	11.91
Externalizing
Entry	104	57.85	12.82	57.35	11.94
4 Months	104	57.27	14.04	54.76	12.19
8 Months	104	55.46	14.23	54.94	12.40
12 Months	104	55.75	13.09	53.38	11.90
16 Months	104	53.65	12.19	54.76	12.75
20 Months	104	55.10	13.80	54.73	13.07
24 Months	104	54.50	13.73	53.67	11.16

Underscored measures indicate a significant main effect of group, p<0.05.

There was a significant effect of returning to the abuser on PTSD symptoms, F (1, 300)=6.68, p=0.013, η²=0.118. Those who returned to the abuser reported higher levels of PTSD than those who did not. There was also a significant effect of returning to the abuser on BSI measures, F (3, 48)=4.05, p=0.012, η²=0.202. Further analysis revealed a significant effect of returning to the abuser on depression scores, F (1, 50)=11.50, p<0.001, η²=0.187, with women who returned to the abuser had higher levels of depression compared with women who did not return.

Discussion

Women who returned to the abuser reported higher danger and associated lethality risk compared to women who did not return. The demonstrated effect of returning to the abuser on women's risk of behaviors associated with lethal violence is an important addition to our understanding of abused women's danger. According to WHO, leaving an intimate partner and estrangement from an intimate partner is a risk factor for femicide.²⁵ No research was found on whether returning to the abuser and/or ending the estrangement affected the level of danger. According to this analysis, women who return to live with the abuser after accessing a shelter or protection order are in significantly higher danger of lethality compared to women who do not return.

Women from both groups also experienced similar outcomes for severity of violence. Our results confirm the intuitive idea that women who leave their abusers will in general experience lower levels of violence compared to women who return to an abusive relationship. There is little research on the experiences of violence after leaving an abusive relationship. One report notes that only 11.5% of women leaving an abusive relationship did not experience any violence or harassment after separating from the abuser for up to 3 years.⁵³ Women who had sought assistance with a protection order or shelter experienced significantly less violence if they did not return to live with the abuser.

The documented increase in risk of behaviors associated with lethal violence as well as increase in severity of violence can be applied to practice in two important ways. First, advocates can use the information to help women make decisions about relationships and safety planning. Previous research indicates that although two-thirds of women (66%) were able to accurately predict their risk of future violence, many women did not accurately predict.⁵⁴ Campbell⁵⁵ suggests reviewing levels of danger with women who believe their risk for future abuse is low. Our findings support the importance of informing women of the increased danger if they return to the abuser.

Second, professionals who screen for intimate partner violence, such as health care professionals or providers of legal and social services for abused women, should take into account whether the woman has left the relationship, received services, and returned. Based on this research, knowing the woman has returned to the abuser can potentially indicate increased danger and added likelihood for poor health and compromised functioning for the woman and her children. When Sullivan and colleagues⁵⁶ completed a needs assessment of 141 women exiting an emergency shelter for women with abusive partners, and results indicated that the women needed numerous community resources upon their shelter exit, including legal assistance, employment, and housing. Race, age, and whether a woman was returning to her assailant influenced which resources she reported needing at shelter exit. Similarly, testimonials from women in shelters worldwide⁵⁷ and in the U.S.⁵⁸ document the women want information on which to make informed decisions and referrals tailored to individual needs. To offer women exiting a shelter or receiving a protection order, information on increased likelihood of danger and poorer functioning, enables the woman to make informed decisions for herself and her children.

The mental health outcomes based on return to abuser were mixed, depending upon whether the woman sought help from a protection order or from a safe shelter. Women seeking safe shelter showed a greater incidence of mental health problems when they returned to the abuser compared to women who did not return. Researchers note personal, social, and economic resources play a role in determining mental health and functioning outcomes after leaving an abusive relationship.⁵³ At entry, women in this sample who sought help with a protection order tended to have a higher level of education, more social support, and less difficulty accessing community agencies.⁵² Greater access to personal, social, and economic resources may assist women seeking a protection order in mental health recovery, whether or not they return to the abuser.

Our methodology has limitations that may underrepresent victimization and functioning outcomes of abused women who access justice services for a protection order or seek safe shelter for the first time. Although we were present at the shelters and District Attorney's office daily, the possibility exists that women were missed. Furthermore, we only sampled from shelters designated for abused women; therefore, abused women who accessed a shelter for homeless persons would have been missed. In like fashion, abused women who seek orders of protection from a private lawyer or legal service other than the county District Attorney's office would have been missed. Another limitation of the sample could be that women who sought services either through the justice system or through safe shelters have differing characteristics, which may determine the service they use and the outcomes after receiving services. Also, we do not have information about women who do not seek services at all. Our participants were limited to English and Spanish speakers. The self-report measures used may also be a limitation. Participants may have underreported or minimized victimization or mental health symptoms. The questions may miss some episodes of victimization and incorrectly classify others. The study was limited in the types of violence against women that were measured; for example, financial and spiritual abuse was not measured. The researchers acknowledge recall bias was operant in all responses.

Conclusion

We provide evidence that women, whether they seek services for a protection order or for safe shelter, are at greater risk of behaviors associated with lethal violence and have more severe violence if they return to the abusive relationship. Women who seek services at a safe shelter have greater mental health problems if they return to the abusive relationship. Advocates and professionals who provide services to abused women should take into account increased risk during assessments and should alert women to their risk.

Author Disclosure Statement

No competing financial interests exist.