Violence against Congolese refugee women in Rwanda and mental health: a cross-sectional study using latent class analysis.

OBJECTIVE: To examine patterns of conflict-related violence and intimate partner violence (IPV) and their associations with emotional distress among Congolese refugee women living in Rwanda.
DESIGN: Cross-sectional study.
SETTING: Two Congolese refugee camps in Rwanda. PARTICIPANTS: 548 ever-married Congolese refugee women of reproductive age (15-49 years) residing in Rwanda. PRIMARY OUTCOME MEASURE: Our primary outcome was emotional distress as measured using the Self-Report Questionnaire-20 (SRQ-20). For analysis, we considered participants with scores greater than 10 to be experiencing emotional distress and participants with scores of 10 or less not to be experiencing emotional distress.
RESULTS: Almost half of women (49%) reported experiencing physical, emotional or sexual violence during the conflict, and less than 10% of women reported experiencing of any type of violence after fleeing the conflict. Lifetime IPV was reported by approximately 22% of women. Latent class analysis derived four distinct classes of violence experiences, including the Low All Violence class, the High Violence During Conflict class, the High IPV class and the High Violence During and After Conflict class. In multivariate regression models, latent class was strongly associated with emotional distress. Compared with women in the Low All Violence class, women in the High Violence During and After Conflict class and women in the High Violence During Conflict had 2.7 times (95% CI 1.11 to 6.74) and 2.3 times (95% CI 1.30 to 4.07) the odds of experiencing emotional distress in the past 4 weeks, respectively. Furthermore, women in the High IPV class had a 4.7 times (95% CI 2.53 to 8.59) greater odds of experiencing emotional distress compared with women in the Low All Violence class.
CONCLUSIONS: Experiences of IPV do not consistently correlate with experiences of conflict-related violence, and women who experience high levels of IPV may have the greatest likelihood for poor mental health in conflict-affected settings. Published by the BMJ Publishing Group Limited. For permission to use (where not already granted under a licence) please go to http://group.bmj.com/group/rights-licensing/permissions.

This study is one of the first to examine how combinations of experiences of conflict-related violence and intimate partner violence (IPV) may result in differential effects on mental health.

This study included a large, probability-based sample of traditionally underserved refugee women.

Although time periods for experiences of violence during the conflict and after fleeing the conflict were well-defined and relatively recent, experiences of IPV were those that could have occurred by one's current or former partners.

The current analysis did not include other potentially important forms of violence against women that may also be high among refugee women populations.

Women may have under-reported experiences of violence due to stigma associated with these experiences; however, all interviews were conducted in private, safe spaces and confidentiality was reinforced throughout the survey to encourage unbiased reporting.

Introduction

Violence against women in conflict-affected settings is a serious global public health issue. Conflict-related violence, such as sexual and physical violence by armed perpetrators, has disproportionately affected women in conflict settings, with experiences of violence affecting between 24% and 73% in settings such as East Timor, Cote d'Ivoire and the Democratic Republic of Congo.1–11 Women in conflict-affected settings often also experience high levels of male-perpetrated intimate partner violence (IPV),7 12 and it is increasingly being recognised that violence against women in such settings does not solely take place in public spheres, but also in more private aspects of a woman's life, such as her home and within intimate relationships.7 10 13 Women who experience conflict-related violence have been shown to be significantly more likely to experience IPV compared with women who had not experienced conflict-related violence.14 15 Approximately 31% of women in the Eastern Democratic Republic of Congo reported experiencing physical and/or sexual IPV,4 and as much as 77% of women in Jordanian refugee camps reported psychological partner violence.16

Both conflict-related violence and IPV have been shown to be associated with poor mental health,4 16 but few studies to date have examined the co-occurrence of conflict-related violence and IPV and its relationship to poor mental health. Vinck and Pham17 found both conflict-related violence and experiences of IPV to be independently associated with depression and post-traumatic stress disorder (PTSD) among women in post-conflict Liberia, and Gupta et al2 found that past-year IPV but not conflict-related violence was associated with probable PTSD. Falb et al14 suggest that there is a moderation effect where the experience of conflict-related violence increases the odds of suicide ideation among women who have experienced IPV. Furthermore, even fewer studies have examined how combinations of experiences of conflict-related violence and IPV may result in differential effects on mental health. US-based research using latent class analysis (LCA), in which similar participants are classified into groups or latent classes based on observable indicators, found patterns of violence experienced by women were differentially associated with positive screens for depression.18 To date, LCA has not yet been applied to examine patterns of violence against women in conflict-affected settings. Additional research therefore is needed to examine the impact of these patterns of violence on mental health.

Accordingly, we sought to examine patterns of conflict-related violence and IPV and their associations with emotional distress, using data from a cross-sectional survey of Congolese refugee women living in Rwanda. Eastern Congo has experienced an ongoing conflict since the early 1990s, and violence against women has been staggering during this protracted conflict.19 20

Methods

Sample population

The Reproductive Health Assessment Toolkit for Conflict-Affected Women (Toolkit)21 was implemented by the American Refugee Committee (ARC) in 2008 for use with Congolese refugee women residing in two refugee camps where ARC provides health and gender-based violence prevention and response services. The camps, Nyabiheke and Gihembe, were established in 2005 and 1997, respectively. In July and August 2008, 27 088 refugees lived in these two camps. ARC programme staff used household lists, compiled through home visits and updated monthly, within the predefined catchment area for the sampling frame; male-only households or households without women of reproductive age were excluded. Excel and/or a random numbers table were used to randomly select eligible households and individuals using simple random sampling. Trained female interviewers, selected mainly from community health workers and peer educators and trained using guidelines outlined in the Toolkit,21 visited each selected household and randomly selected one woman of reproductive age (15–49 years of age) within the household. The sampling strategy to select women of reproductive age used the KISH method22 and was implemented per technical assistance from the Centers for Disease Control (CDC) and their guidelines in the Toolkit.21 Additional details about this process and associated forms can be found elsewhere.21 After randomly selecting the participant, the interview staff then verbally obtained informed consent. Interview staff were language-matched with participants and scheduled a time for the woman to participate in the survey within a private, centralised location (eg, NGO offices)21 All surveys were verbally administered in either French or Kinyarwanda (an official language in Rwanda), and responses were recorded by female interviewers. Surveys were translated and back-translated to ensure accuracy and collected data on a variety of topics, including but not limited to emotional health, gender-based violence and sociodemographic characteristics. Participants who may have needed counselling were referred to the ARC office in each of the camps, where women were linked with appropriate resources. The ARC followed the WHO's guidelines for conducting violence surveys.23 A total of 810 women participated in the survey from these refugee camps. We used all available data and restricted our analytic sample to ever-married women (N=548) since the IPV module of the RHA was only administered to these women, which is consistent with other global demographic surveys (eg, Demographic Health Survey). This secondary data analysis of unidentifiable data was deemed exempt from review by the Yale School of Public Health human subjects committee and the Internal Review Board at the University of Illinois at Chicago.

Measures

Outcome

Emotional distress was measured using the Self-Report Questionnaire-20 (SRQ-20),24 which had been developed as a mental health screening tool by the WHO. The SRQ-20 asks respondents to indicate whether or not they have experienced common problems in the past 4 weeks (eg, “Do you have headaches?” and “Do you feel unhappy?”). Items were summed to indicate the total number of problems experienced by each participant; total scores ranged from 0 to 20. Internal consistency for the SRQ-20 was excellent (Cronbach's α=0.91). For analysis, we considered participants with scores greater than 10 to be experiencing emotional distress and participants with scores of 10 or less not to be experiencing emotional distress.25 This scale and the cut-off point employed in the current analysis have been previously validated among conflict-affected populations in Rwanda.25

Exposures

We assessed whether or not women experienced physical, emotional and/or sexual violence by anyone outside their family during the conflict and after the conflict. Consistent with existing studies that have implemented the RHA Toolkit to examine violence against refugee women, we considered women to have experienced physical violence if they reported having been physically hurt (such as slapped, hit, choked, beaten or kicked) and/or shot at or stabbed. We considered women to have experienced emotional violence if they reported having been threatened with a weapon of any kind and/or detained against their will. We considered women to have experienced sexual violence if they reported having been subjected to improper sexual comments, forced to remove or stripped of their clothing, subjected to unwanted kissing or touching on sexual parts of your body and/or forced or threatened with harm to make you give or receive oral sex or have vaginal or anal sex. Participants were asked to report whether or not they experienced these types of violence (1) during the conflict and (2) after the conflict. Specific dates were not provided.

Women were also asked about their experiences with physical, emotional and sexual violence from any of their partners or ex-partners (ie, lifetime IPV). We considered women to have experienced physical IPV if they reported that any of their partners had “slapped you, twisted your arm, hit you with a fist or something else, pushed you down or kicked you, or choked you” and/or “threatened to hurt you with a weapon or himself.” We considered women to have experienced emotional IPV if they reported that any of their partners had “forbid you from participating in activities in the community such as seeing friends or family, educational opportunities, women's groups, or employment opportunities;” and we considered women to have experienced sexual IPV if they reported that any of their partners had “threatened to hurt you or used force to make you have sex with him when you did not want to.” For all violence experiences, we chose to code ‘no response’ as not having experienced that type of violence.

Sociodemographic controls

Sociodemographic characteristics included as covariates in the final model included: age, current relationship status (married and living together; married and not living together; not currently married (ie, abandoned, divorced or separated); or widowed), ability to read (not at all, with difficulty, easily) and refugee camp.

Analysis

We first generated descriptive statistics to describe sample characteristics. We then conducted LCA with maximum likelihood estimation and robust SEs to classify participants with similar violence experiences into groups or latent classes.26 Our nine indicators of violence experiences were used to empirically derive class membership. We determined the most appropriate number of classes by examining six commonly used criteria, including the Akaike Information Criterion (AIC), the Bayesian Information Criterion (BIC), the sample-size adjusted BIC (SSABIC), Lo-Mendell-Rubin Likelihood Ratio Test (LMR-LRT), entropy (a measure of the appropriateness of classification), and the usefulness and interpretability of the latent classes.26 The AIC, BIC and SSABIC indicate relative goodness of fit; lower values represent better fitting models. A non-significant p value from the LMR-LRT suggests that the model with one fewer classes is acceptable, and higher entropy values suggest better classification. The best model was considered the one with the fewest number of distinct classes to offer meaningful results.26 Multivariate logistic regression models were used to determine the effect of latent class on emotional distress, adjusted for other covariates. A complete case analysis was conducted. Analyses were conducted with SAS V.9.3 and MPlus V.4.21.

Results

Participant characteristics

Participants on average were 32 years old (SD=7.64 years); 16% were 15–24 years old and 37% were 35–49 years old (table 1). A total of 57% were married and living with their partner, and almost 44% reported they could read easily. The sample was fairly evenly divided between the two refugee camps. The average emotional distress score was 6.9 (SD=5.50), and 28% of the sample had scores greater than 10.

Table 1

Participant characteristics (N=548)

	N (%)*
Age (years)†
15–24	85 (15.9)
25–34	251 (47)
35–49	198 (37.1)
Current relationship status‡
Married, living together	309 (56.9)
Married, not living together	144 (26.5)
Other	49 (9)
Widowed	41 (7.6)
Ability to read
Not at all	184 (33.6)
With difficulty	125 (22.8)
Easily	239 (43.6)
Refugee camp
Nyabiheke Camp	302 (55.1)
Gihembe Camp	246 (44.9)
Emotional distress (SRQ-20)§
≤10	374 (72.3)
>10	143 (27.7)
Type of violence
During conflict: physical¶	191 (34.9)
During conflict: emotional	169 (30.8)
During conflict: sexual¶	96 (17.5)
After conflict: physical	44 (8)
After conflict: emotional	28 (5.1)
After conflict: sexual	39 (7.1)
IPV: physical**	93 (17.3)
IPV: emotional†	45 (8.4)
IPV: sexual†	74 (13.9)

*Valid percentages.

†14 participants (2.6%) had missing responses.

‡5 participants (0.9%) had missing responses to relationship status.

§“No response” was coded as missing; 31 participants (5.7%) were missing responses to one or more items in the SRQ-20.

¶1 participant (0.2%) had a missing response.

**17 participants (3.1%) had missing responses.

IPV, intimate partner violence; SRQ-20, Self-Report Questionnaire-20.

Experiences of violence

The highest levels of violence were experienced during the conflict (table 1). Approximately 35% of women reported experiencing physical violence, 31% reported experiencing emotional violence and 18% reported experiencing sexual violence. Almost half of women (49%) reported experiencing any type of violence during the conflict. After fleeing the conflict, experiences of violence were reportedly less common, with less than 10% of women reporting experiences of physical, emotional or sexual violence. Lifetime physical IPV was reported by 17% of the women, lifetime emotional IPV was reported by 8% of the women and lifetime sexual IPV was reported by 14% of the women. Any lifetime IPV was reported by approximately 22% of women.

LCA suggests the existence of four distinct classes (figure 1). The four-class solution was supported by the AIC, SSABIC, LMR-LRT and the interpretability of the classes. Classes were named based on their most prominent characteristics and included the Low All Violence class, the High Violence During Conflict class, the High IPV class and the High Violence During and After Conflict class. The largest class (n=322; 59.0%), which we named the Low All Violence class, was characterised by infrequent reports of during-conflict violence, after-conflict violence and IPV. The second-largest class (n=113, 20.7%), which we named the High Violence During Conflict class, was characterised by a high prevalence of violence reported during the conflict but low prevalences of after-conflict violence and of IPV. The third class (n=79; 14.5%), which we named the High IPV class, was characterised by the highest frequencies of reports of IPV; this class of respondents also experienced moderate frequencies of violence during the conflict but low frequencies of violence after the conflict. The smallest class (n=32; 5.9%), which we named the High Violence During and After Conflict class, was characterised by high frequencies of violence during and after the conflict and moderate IPV victimisation.

Figure 1

Exposure to violence during and after the conflict and lifetime exposure to intimate partner violence among ever-married women. AC, after conflict; DC, during conflict; IPV, intimate partner violence (lifetime).

Violence experiences and emotional distress

Multivariable logistic regression models suggested that latent class was strongly associated with emotional distress, even after adjusting for other common risk factors (table 2). Women in the High Violence During and After Conflict class compared with Low All Violence class had almost three times the odds of experiencing emotional distress in the past 4 weeks (95% CI 1.11 to 6.74). Additionally, women who were in the High Violence During Conflict class compared with the Low All Violence class had a 2.3 times greater odds of experiencing emotional distress (95% CI 1.30 to 4.07). Women in the High IPV class compared with women in the Low All Violence class had a 4.7 times greater odds of experiencing emotional distress (95% CI 2.53 to 8.59); furthermore, the odds of experiencing emotional distress among women in the High IPV class was approximately twice the odds of experiencing emotional distress among women in the High Violence During Conflict class (95% CI 1.04 to 3.98).

Table 2

Multivariate logistic regression models exploring the associations between emotional distress and classes of violence exposure among ever-married women (N=477; 92.2% of sample with valid emotional distress data)

	OR (95% CI)
Age (years)
15–24	1.00
25–34	1.48 (0.76 to 2.91)
35–49	1.66 (0.80 to 3.42)
Current relationship status
Married, living together	1.00
Married, not living together	1.56 (0.92 to 2.63)
Other	2.63 (1.21 to 5.70)*
Widowed	4.70 (2.20 to 10.04)**
Ability to read
Not at all	1.76 (1.05 to 2.93)*
With difficulty	0.71 (0.39 to 1.31)
Easily	1.00
Length of displacement	1.15 (0.90 to 1.46)
Nyabiheke Camp	1.72 (0.21 to 13.92)
Class†
Low all violence	1.00
High violence during conflict	2.30 (1.30 to 4.07)**
High IPV	4.67 (2.53 to 8.59)**
High violence during and after conflict	2.74 (1.11 to 6.74)*

*p<0.05; **p<0.01.

†High IPV class is significantly different from High violence during conflict class (High IPV versus High violence during conflict: OR=2.03; 95% CI 1.04 to 3.98; p value=0.039) in its effect on emotional distress.

IPV, intimate partner violence.

Discussion

This refugee camp-based study found high levels of emotional distress and multiple forms of violence experienced by Congolese refugee women of reproductive age living in Rwanda. One in four women were thought to be experiencing poor mental health, an estimate higher than other reported estimates of poor mental health, including past-month suicidality (7%),14 PTSD (12–14%),2 17 and depression (11%)17 among women in conflict-affected settings. This higher prevalence seems likely attributable to the more inclusive nature of our mental health measure, which assesses general emotional distress as opposed to specific diagnoses. This prevalence, however, is lower than the estimate of depression (54%) found among refugees living in Uganda, who were primarily from the Democratic Republic of Congo.27 This discrepancy could be explained by differences in the respondents; the Congolese refugees living in Rwanda had been residing in the camps longer on average than the Congolese refugees in Uganda.

Our findings also suggested that approximately half of Congolese refugee women in Rwanda have experienced some form of violence. Not surprisingly, experiences of violence among these women were concentrated during the conflict with approximately 49% experiencing some form of violence during the conflict and 15% experiencing some form of violence after fleeing the conflict. Estimates of conflict-related sexual violence among this sample were comparable to other general population samples from the DRC.4 8 28 Nearly one-quarter of women had experienced IPV at some point in their lifetime, consistent with frequencies observed in comparable settings.2 4 17 Taken together, these findings affirm the need to address the co-occurrence of both public forms of violence (ie, conflict-related violence) and more private forms of violence (ie, IPV) experiences among refugee women.

The data highlighted four distinct latent classes of women reporting similar violence experiences, based on the co-occurrence of conflict-related violence and IPV experiences. The emergence of these distinct classes suggests that experiences of IPV do not consistently correlate with experiences of conflict-related violence. For instance, in the High IPV class, women experienced relatively low levels of during and after conflict violence, but experienced high levels of IPV. These patterns suggest the strength in a latent class approach to understand nuance in violence experiences. Our study is the first to use LCA to understand patterns of violence against women in conflict-affected settings. Future research may seek to understand risk and protective factors that may be associated with the clustering of violence experiences among women affected by conflict to help inform prevention and intervention programmes.

These classes demonstrated differential associations with emotional distress and corroborate growing research in conflict-affected settings documenting that emotional health is not solely affected by violence perpetrated by armed actors or others outside of the family, but is also a result of private forms of violence, such as IPV.2 14 Our results suggest that although women who experience elevated levels of violence may suffer from worse mental health compared with women who do not experience violence or who experience low levels of violence, women who experience high levels of IPV may have the greatest likelihood for poor mental health. These results demonstrate the value of LCA for identifying distinct classes of women to understand potential mental health consequences. Furthermore, our findings have important implications for refugee programmes as much of the current humanitarian funding tends to focus on conflict-related violence and excludes IPV. Similarly, specialised violence against women prevention and response programmes within refugee camps tend to focus solely on conflict-related violence as well. Policies and programmes must consider allocating resources for addressing IPV in addition to conflict-related violence in conflict-affected settings.

Despite several strengths of our study, including a large, probability-based sample of refugee women, findings should be interpreted in light of some limitations. First, although the time periods for experiences of violence during the conflict and after fleeing the conflict were well-defined and relatively recent, experiences of IPV were those that could have occurred by one's current or former partners. This measurement, therefore, considered IPV occurring 20 years ago and 20 days ago as having the same effect. Additionally, the experience of IPV may be ongoing, whereas violence experienced during the conflict is relatively distant and the violence experienced after the conflict may be less distant but still in the past. These possible discrepancies in the timing of violence occurrence could help explain the impact of IPV on emotional distress within this sample. Furthermore, this analysis lacks the ability to make conclusions about the effects of chronic IPV. Additional research is needed to overcome these limitations in order to better understand the timing of such events. Second, the current analysis did not include other important forms of violence against women (eg, child marriage, transactional sex, denial of resources) that may also be high among refugee women populations. Future LCA research may seek to integrate these additional forms of gender-based violence to further understand associations with health outcomes. We also did not include socioeconomic status, as status prior to displacement was unavailable; socioeconomic status has been associated with mental health in previous research and is an important factor to consider for future work.29 Third, recall bias could have affected the reporting of both IPV as well as conflict-related violence, given that the conflict arose in the early 1990s, and could have caused under-reporting of less severe forms of violence. Women may have also misinterpreted questions; however, survey items were based on the WHO's domestic violence surveys, which have been field-tested in several vulnerable populations.23 30 Further, women may have under-reported experiences of violence due to stigma associated with these experiences. All interviews, however, were conducted in private, safe spaces and confidentiality was reinforced throughout the survey to encourage unbiased reporting. Interviewer bias is also a possibility; however, all interviewers were trained according to WHO guidelines for conducting domestic violence surveys.23 30 All interviewers were able to offer participants ethical referrals for relevant services according to each participant's wishes, and the benefits of seeking assistance from qualified service providers were continuously reinforced throughout the refugee camps. New strategies, however, are likely needed to combat low self-efficacy and help-seeking behaviours among women and girls affected by violent experiences. Last, sampling was not proportional to the size of the refugee camps and the survey's participation rate is unknown. These results may be generalisable to women residing in similar conflict-affected settings, but the extent to which is unclear.

Our study has several implications for policy, practice and future research. First, our findings suggest that humanitarian assistance programmes should emphasise the importance of employing skilled mental health workers or social workers to respond to multiple forms of violence, paying particular attention to experiences of IPV. Programmes serving refugee women and communities should also seek to respond to multiple forms of violence against women, including IPV. Currently, much of the funding and programmatic attention focuses primarily on conflict-related violence (particularly sexual violence), with less emphasis on IPV. Programmes aimed to respond to violence against women should thus be equipped to respond to the mental health needs of women with varying levels of exposure to both discrete events of violence that are more common with conflict-related violence as well as chronic, continual exposure to IPV as many women in conflict-affected settings experience ongoing IPV.31 Studies also suggest that the prevalence of IPV may be higher in conflict-affected settings and that exposure to conflict may increase a woman's vulnerability to experiencing IPV. As such, prevention and intervention programmes are needed to reduce the prevalence of IPV and mitigate its health consequences. To date, little data exist regarding promising interventions in such conflict-affected settings, but at least two randomised trials within Cote d'Ivoire demonstrate the potential to reduce IPV in conflict-affected settings.2 12 More research is needed in this area. Healthcare workers in humanitarian settings should also implement field-tested, safe and routine IPV and other violence screening for women and adolescent girls accessing healthcare because these experiences have health implications that appear to differ from the experiences of conflict-related violence. Additionally, these healthcare workers should pay particular attention to women who have experienced IPV from their current or past partner as they may have some of the poorest mental health among women in refugee camps. Acknowledging the patterns of violence women experience in conflict settings may improve mental health response services. Such response services should be combined with other prevention programmes, such as community-based programming to improve gender equitable norms and engagement of men to reduce perpetration, in order to reduce violence against women within and outside their homes.