Mental health of women and children experiencing family violence in conflict settings: a mixed methods systematic review

Background Armed conflict has significant impacts on individuals and families living in conflict-affected settings globally. Scholars working to prevent violence within families have hypothesised that experiencing armed conflict leads to an increase in family violence and mental health problems. In this review, we assessed the prevalence of family violence in conflict settings, its association with the mental health of survivors, moderating factors, and the importance of gender relations. Methods Following PRISMA guidelines, we systematically reviewed quantitative and qualitative studies that assessed the prevalence of family violence and the association between family violence and mental health problems, within conflict settings (PROSPERO reference CRD42018114443). Results We identified 2605 records, from which 174 full text articles were screened. Twenty-nine studies that reported family violence during or up to 10 years after conflict were eligible for inclusion. Twenty one studies were quantitative, measuring prevalence and association between family violence and mental health problems. The studies were generally of high quality and all reported high prevalence of violence. The prevalence of violence against women was mostly in the range of 30–40%, the highest reported prevalence of physical abuse being 78.9% in Bosnia and Herzegovina. For violence against children, over three-quarters had ever experienced violence, the highest prevalence being 95.6% in Sri Lanka. Associations were found with a number of mental health problems, particularly post-traumatic stress disorder. The risk varied in different locations. Eight qualitative studies showed how men’s experience of conflict, including financial stresses, contributes to their perpetration of family violence. Conclusions Family violence was common in conflict settings and was associated with mental health outcomes, but the studies were too heterogenous to determine whether prevalence or risk was greater than in non-conflict settings. The review highlights an urgent need for more robust data on perpetrators, forms of family violence, and mental health outcomes in conflict-affected settings in order to help understand the magnitude of the problem and identify potential solutions to address it. Supplementary Information The online version contains supplementary material available at 10.1186/s13031-021-00410-4.


Introduction
Rates of family violence and mental ill health are hypothesised to be higher in areas of armed conflict than in non-conflict-affected settings [1], but the extent of the risks and their associations have not been systematically assessed.

Open Access
*Correspondence: d.devakumar@ucl.ac.uk 1 Institute for Global Health, University College London, London WC1N 1EH, UK Full list of author information is available at the end of the article Family violence, including violence against women (VAW) from intimate partners or other household members and violence against children, is prevalent across the world and has adverse implications for physical and mental health [2]. The WHO multi-country study on women's health and domestic VAW, showed that lifetime prevalence of physical and sexual abuse by an intimate partner ranges from 15 to 71% [3]. Rates of child abuse vary by type of abuse, definition and location. 'Base-case estimates' from 96 countries suggested at least 50% prevalence of exposure to violence in the last year, equating to approximately 500 million children aged 2-17 years [4]. A systematic review of 55 studies estimated the prevalence of sexual violence to be 8-31% in girls and 3-17% in boys [5]. Sexual violence in particular is associated with a long-term increased risk of mental and physical illness [6].
Reports from conflict settings indicate increased rates of VAW and physical and sexual abuse of children by both men and women [7][8][9]. For example, in a study of child soldiers from Sierra Leone, 44% reported having been raped [10]. There are a number of possible reasons for this. The social and economic conditions that lead to conflict at a societal level may also result in increases in family violence. Catani et al. propose that the 'cycle of violence' model, in which violence is transmitted intergenerationally within a family [11], can also apply to conflict situations in which external violence leads to violence within the family [12]. The mechanisms by which this may occur are multiple. Conflict may normalise violence in a society, increase substance abuse, and affect education, income, family composition, and gender attitudes [13][14][15].
In addition to violence, mental disorders are common in conflict, due to both the conflict itself and the family violence. Women and children living in conflict-affected settings are at increased risk of developing depressive, anxiety, and psychotic disorders [16,17]. The prevalence of mental disorders in conflict-affected populations is substantially higher than in the average population: 15.4% for post-traumatic stress disorder (PTSD) and 17.3% for depression, versus 7.6% for any anxiety disorder including PTSD and 5.3% for any mood disorder including depressive disorders [18,19].
There remains a gap in knowledge on the prevalence of family violence in conflict-affected areas and its mental health consequences. Greater understanding is needed of why and how family violence may result in an increase in mental disorders in conflict-exposed populations. This review was designed to explicitly investigate links between family violence occurring at an individual level (within families), and broader forms of violent armed conflict occurring within communities. We included both quantitative and qualitative research to both enumerate the problem and explore the reasons why armed conflict is associated with mental ill-health, expanding our conceptual understanding of the underlying causes. Our review had four objectives. First, to investigate the prevalence of family violence, including violence against women and children, in conflict-affected areas. Second, to describe the risk of mental disorder among women and children who had experienced or witnessed family violence in conflict-affected areas. Third, to examine how the association of family violence with mental health in women and children is moderated by the type of conflict, the gender of the perpetrator, and the type of violence. Finally, to understand the ways in which gender relations, as a well-recognised risk factor for violence against women and children [20], influence the mental health of women and children who experience or witness family violence.

Design
We conducted a mixed-methods systematic review. The protocol was registered prospectively with PROSPERO (CRD42018114443) and reporting follows PRISMA guidelines [21].

Search strategy
We searched the following databases: EMBASE, International Bibliography of Social Sciences, MEDLINE (PubMed), PsycINFO, Scopus, SciELO, Social Policy and Practice, Sociology database, Global Index Medicus, Online Library of Dignity, Web of Science, Cochrane Library, CINAHL Plus, and regional databases (LILACS, African Journals online, Latin America & Iberia Database (ProQuest), Middle East & Africa Database (ProQuest)). Studies were imported into Endnote and duplicates removed. The search was restricted to articles in English, Arabic, Spanish, Portuguese, and French, with no date restrictions. Database searches were complemented by reference list screening and citation tracking of included materials in Web of Science and Google Scholar.

Search terms
The search strategy (Additional file 1: Appendix) was designed with the support of a librarian and included terms related to (1) women and children, (2) family violence, (3) mental disorders, and (4) conflict-affected areas. Terms describing conflict-affected areas were compiled using a list of countries and territories in which armed conflicts were recorded before 2018, using the Uppsala Conflict Data of the International Peace Research Institute and the Heidelberg Institute for International Conflict Research. We used the Uppsala Institute interstate and intrastate definitions of conflict: "An armed conflict is a contested incompatibility that concerns government and/or territory where the use of armed force between two parties, of which at least one is the government of a state, results in at least 25 battle-related deaths in one calendar year"; and "A conflict between a government and a non-governmental party, with no interference from other countries" [22].

Inclusion criteria
Population and setting: Our review included adult women (aged 18 years or older), and male or female children (aged under 18 years). If the study sample included men aged 18 or older, reports were only included if 90% were under 18. We did not include facility-based studies in the assessment of prevalence due to the risk of selection bias, but chose to include school studies to reflect the universal right to education. We included studies up to 10 years post-conflict to capture the ongoing effects of war. Type of study: We included primary research using qualitative (e.g. individual interviews, focus group interviews) or quantitative (e.g. cohort, case-control, crosssectional studies, or baseline data from experimental or quasi-experimental studies) research designs. Where quantitative studies were reported in more than one publication, the one with the largest sample was selected for inclusion. Exposures: We included family violence against women or children. The definition of family violence, adapted from the United Nations definition of violence against women [23] was the intentional use of physical force or power by an intimate partner, ex-partner or parent that results in physical, sexual or mental harm or suffering to women and/or children. This includes threats of violence, sexual coercion, psychological abuse and arbitrary deprivation of liberty. Family violence is an overarching term that includes intimate partner violence and child abuse, perpetuated by family members. Studies of children who witness violence against women in the home in addition to experiencing direct violence themselves were included. The comparator group for quantitative studies was no exposure to family violence. Outcomes: Our primary outcome was mental disorder, defined in accordance with the International Classification of Diseases and Related Health Problems (ICD- 11) or Diagnostic and Statistical Manual of Mental Disorders, 5th Edition (DSM-V) criteria. Articles were eligible if mental disorder was assessed using either a validated diagnostic or screening instrument, diagnosed by a clinician, or self-reported by participants. The secondary outcomes included were mental and social wellbeing of women experiencing family violence and children experiencing or witnessing family violence.

Exclusion criteria
Population and setting: We excluded soldiers and war veterans, refugees and asylum seekers who had moved away from the conflict setting. Soldiers and war veterans were considered to be too different a population who had experienced combat directly. Type of study: We excluded opinion pieces and editorials, dissertations or theses, policy papers, reviews, general reports that did not introduce new evidence from a specific study, and conference abstracts. Exposures: Other forms of violence against women and children were beyond the scope of the review; for example, peer bullying, violence by perpetrators other than intimate partners, ex-partners, or family members, female genital mutilation or cutting, or child labour or marriage.

Screening
Titles and abstracts were screened independently by two reviewers (CR, NU) from 2nd June 2019. Full texts were screened from 31st July 2019. Data extraction began on 1st September 2019 and was completed by January 2020. Disagreement was resolved by discussion and no papers were discussed with a third reviewer. Each full text was then independently reviewed by two authors (two of AP, AU-C, CR, DD, NU, JM). Disagreement was resolved by discussion. One author was contacted for missing data, but did not respond.

Quality assessment
Two reviewers assessed each article (qualitative articles by AP and JM, with disagreements decided by DD; quantitative articles by A-UC and DD) for quality using the Joanna Briggs Institute checklists [24]. Discrepancies in scoring were discussed between the reviewing authors.

Analyses
Quantitative studies: We used descriptive statistics to summarise information about the study characteristics, samples, and methods. Ninety-five percent confidence intervals were reported for all measures of prevalence and risk ratios where available. Meta-analysis was not conducted due to study heterogeneity.
Qualitative analysis: We conducted a thematic analysis of selected articles and used NVIVO 12 to organise the data. All articles were read multiple times by two reviewers (AP, JM) to identify initial codes related to the associations between family violence and mental health and the influence of gender relations (Additional file 1: Appendix), drawing on the accounts of both research participants and the article's authors. AP and JM discussed these codes to arrive at a consensus and collaboratively developed a final codebook. One reviewer (JM) then went through each article in detail, systematically extracting all relevant quotes. As a descriptive account of the current literature on this topic, further theoretical interpretation beyond a descriptive analysis was not needed. [25].

Studies included
We identified 2603 records, of which 1682 were duplicates (Fig. 1). Of the 173 full-text articles retrieved, 29 studies conducted between 1994 and 2017 in 13 countries were included (Additional file 1: Appendix). Reasons for exclusion were mostly due to the study not including family violence or not being a full research paper (Additional file 1: Appendix).
The quantitative papers described violence against women and violence against children. We included 14 quantitative studies of violence against women in conflict from 11 countries (five in Africa, five in Asia, one in Europe). All were cross-sectional, with the exception of one cohort study [26]. The sample sizes for women within these studies ranged from 80 to 2196 individuals ( Table 1). The types of violence included varied. Eleven studies focused on intimate partner violence (IPV). The other studies had broader definitions including violence from other members of the family and other forms of violence (Table 1).
We included eight studies of violence against children in conflict-affected countries. All were from Asia except one from Uganda. All were school-based samples except two that were community-based [27,28], which were the smallest (n = 149 [27]) and largest (n = 513 [28]). Six studies were cross-sectional and two were cohorts ( Table 2).
Eight qualitative studies were included [29][30][31][32][33][34][35][36], presenting research from six countries: Côte d'Ivoire (2), Timor-Leste (2), Colombia (1), Uganda (1), Democratic Republic of the Congo (1), and Sri Lanka (1).  Liberian adults IPV from partner or spouse * The study included women receiving psychiatric treatment. These women were excluded from our analysis ** Vulnerable individuals defined as 'A person who had any of the following characteristics: widowed, divorced, or separated; living in an internally displaced persons camp; women who had suffered sexual torture; single mothers; orphans; out of school youth; child/adolescent mothers; women and adolescent girls without any source of livelihood (mainly lack of access to arable land); having a mental health problem; survivor of intimate partner violence; and survivors of recent famines or floods Five reports focused exclusively on family violence perpetrated against women, two on violence by mothers against their children, and one on both forms of family violence. Only one report discussed VAW by family members other than the husband. Seven reports used interviews or focus groups and one used a questionnaire. Terms such as depression, anxiety, post-traumatic stress disorder, or suicidal ideation were rarely used. As such, mental health problems were coded as anything that discussed a psychological symptom or negative feelings, including stress, fear, worry, unhappiness, or concern. We summarise below the results according to our four objectives, including the qualitative and quantitative results together.

Prevalence of family violence in conflict-affected areas
The prevalence of family violence and associations with mental health are shown in Tables 3 (for women) and 4 (for children). Amongst women, the prevalence of violence varied, but exposure to domestic violence in both conflict and post-conflict settings was common. The highest prevalence was a study from Bosnia and Herzegovina, where 78.9% of women reported physical abuse and 96.1% psychological abuse [37]. Commonly, however, reports of IPV were in the region of 30-40%. Lifetime experience of IPV was reported at 37.7% [38] from Liberia, 44.6% for those who had 2-4 trauma exposures in a study from Afghanistan [13], and 49.8%in a study from rural Côte d'Ivoire [39]. High prevalence was also reported for exposure to IPV within the last 12 months (24.4% [38], 25.1% [13] and 29.7% [39]). No obvious difference was seen by country or region of the world.

Risk of mental health problems among women and children who have experienced or witnessed family violence
Nearly all the studies found associations between family violence and adverse mental health among women (Table 3). Depression and PTSD were commonly measured outcomes.
Six studies estimated the association between family violence and mental health outcomes in children. Across all studies, associations were found between family violence and mental ill-health.

Post-traumatic stress disorder
Gupta et al. [16] found a non-significant positive association between IPV and PTSD (OR 1.6 (95% CI 0.9, 2.6) in Côte d'Ivoire. Johnson et al. [44] found that in the DRC 43.9% (95% CI 34.7, 53.0) of women not affected by IPV had PTSD, compared with 77.2% (95% CI 66.8, 87.7) who were affected by IPV. A dose-response relationship between compound exposures of violence and likelihood of PTSD was reported by Rees et al. [45], whereby severe combined physical and psychological abuse was associated with the greatest risk of PTSD (OR 3.24, 95% CI 1.90, 5.50). These findings were comparable with those found in Afghanistan. Jewkes et al. [13] reported that, of women who had no exposure to trauma, 1.35% had PTSD. Of women with one trauma exposure 1.7% had PTSD, and with 2-4 trauma exposures 1.65% had PTSD (p < 0.0001). Umubyeyi et al. [46]  In children, studies from Sri Lanka [12], Afghanistan [40,42,43], Lebanon [47] and Uganda [28] all showed increases in PTSD associated with family violence. In comparable studies, Catani et al. [12] and Catani et al. [40] showed that exposure to war predicted family violence and PTSD symptoms, and a correlation between family violence and PTSD. Similarly, Fayyad et al. found higher PTSD scores (as measured by the Child Revised Impact of Events Scale (CRIES)) in children who had experienced or witnessed family violence [47].

Other mental health problems
Umubyeyi et al. [46] found associations between major depressive episode and physical (aOR 4.63; 95% CI 2.57, 8.32), sexual (aOR 5.49; 95% CI 2.94, 10.25), and psychological violence (aOR 5.59; 95% CI 3.19, 9.80). As was the case with PTSD, a dose-response relationship between compound exposures of violence and likelihood of depression was observed. In Afghanistan, Jewkes et al. [13] found that, of women who had no exposure to trauma, 12.1% reported depression; with one trauma exposure, 16.2% reported depression; and of women who had 2-4 trauma exposures; 17.1% reported depression (p < 0.0001). In Bosnia and Herzegovina, Avdibegovic et al. [37] found that 76%     In children, the association with depression was less consistent, though often not measured. Strengths and Difficulties Questionnaire (SDQ) scores, which measure a range of psychosocial outcomes, were also higher in the studies from Lebanon [47] and Afghanistan [42]. Panter-Brick found increases in SDQ score with traumatic beatings and family violence at home [43], and those who had sustained distress compared to the low distress group (measured by CRIES) had an increased risk (adjusted OR 4.84) of living in a family with ongoing stressful domestic violence [42]. Saile et al. examined the pathway between conflict, family violence and mental health in children. They found small changes in SDQ scores, depression symptoms and PTSD [28].
Women's perspectives on armed conflict or how their experiences of conflict may have affected their mental health were reported in relation to financial stress and their concerns about the care of their children. For instance, women participating in focus group discussions in Côte d'Ivoire reported that financial stress made them lose interest in having sex with their husbands, to which some husbands might respond with sexual violence: When the woman is not at ease… you know, there is no money to properly take care of the children, she is preoccupied and she does not feel like having sex so it happens that the man rapes her [36].
Another study of returned female combatants in Northern Uganda explored the stresses women felt about the care of their children when they returned from situations in which they had been abducted and forced to marry combatants:

When I had just returned from the bush with my children, I used to have lots of thoughts [worries]
on how I will look after the children in case of sickness since their father is not there [34] .
Rees et al. [29] uniquely explored the use of violence by women as a result of the armed conflict in Timor-Leste, with the concept of 'explosive anger' as an explanation for high rates of violence against children.
3. Variation of the association of family violence with mental health by type of conflict, sex of perpetrator, type of violence

Type of conflict
In the quantitative studies, we were unable to directly answer our question as to whether the association between family violence and changes in mental health differed by type of conflict. Only the three studies from Uganda were interstate [28,48,49]. No discernable difference could be seen between these and the intrastate conflicts (Democratic Republic of Congo [44], Côte d'Ivoire [16,30,39], Lebanon [47,50], Liberia [38], Palestine [26] and Sri Lanka [12,41]), or the internationalised intrastate conflicts (Afghanistan [13,27,40,43], Bosnia and Herzegovina [37] and Iraq [51]). The qualitative studies also did not engage with this concept and there were no comparative case studies in the review.

Sex of perpetrator
Limited data were available on the prevalence of femaleperpetrated family violence, and we were unable to assess whether the association between family violence and mental health problems varied according to the sex of the perpetrator. The majority of studies measuring violence against women focused on male spouses in the context of heterosexual married relationships. Two studies, Usta et al. [50] and Hossain et al. [39], defined domestic violence as including violence from family members. In rural Côte d'Ivoire, Hossain et al. [39] demonstrated the lifetime prevalence of sexual violence from female perpetrators including female family members was 0.1%. Among women, lifetime prevalence of physical violence from female family members was 8.9%; prevalence of lifetime physical violence from male family members was also 8.9%.

Impact of gender norms on the mental health of women and children experiencing family violence
The conflict-related triggers identified in qualitative articles, such as financial stress and alcohol use, were explained by both authors and study participants as rooted in gender norms. In situations of conflict, men may not be able to fulfill their socially-expected roles such as providing financially for the household [35], and in some cases women become the main household providers [30]. This has indirect impacts on the mental health of women and children by increasing men's stress and sense of insecurity in ways that increase family violence [29,30,35].
Guruge et al. explained how the use of violence by men becomes a means of maintaining a sense of power and control in the face of the instability brought on by conflict, which threatened gender norms that men rely upon for their identity and sense of security: With death, disappearances, disability and loss of traditional sources of livelihood (such as fishing) affecting many men, women in war-affected areas of Sri Lanka have taken on being the breadwinner and head of household roles, and have become more active in their communities. When there is a disruption of established social relationships and roles in war and post-war contexts, men are known to use violence as means of re-exerting control and power to maintain roles that are consistent with social norms. [35] Related to this, economic instability often opens up new opportunities for women to earn, which men may feel threatened by, contributing to their use of violence [29,30,35]. Shuman et al., for example, discussed how change in women's roles arose in Côte d'Ivoire as an underlying cause of family violence: In post-crisis Côte d'Ivoire, economic opportunities are scarce and in some cases, women have become the financial providers for their families. While women welcomed opportunities to have more control over resources in their relationships, they also described being perceived as a threat by their partners. Among men, this perceived loss of control and traditional gender responsibilities was discussed as an underlying cause of all forms of IPV. [30] The stress reported by men was also highlighted as a reason for their increased drinking behaviour. In Côte d'Ivoire, Northern Uganda and Sri Lanka, instability and financial problems arising from the conflict were seen as triggers for problematic drinking by men, which contributed in turn to increased episodes of violence [34][35][36].
The qualitative data also support the impact of gender norms on the mental health of women and children experiencing violence directly through their experience of violence-related stigma. [29,30]. Sexual violence was frequently described as especially stigmatising for women. For instance, Annan and Brier's study of women abducted by the Lord's Resistance Army in Northern Uganda described the lower status of women who return to their communities as women who had been 'married' or sexually abused. They described how the fear of this stigma drives women to agree to be married either to violent men or as second wives, which is attached to lower status in the household, in turn increasing the risk of violence: While polygamous marriages are common, women as second wives are seen to have less power in households than first wives. One social worker observed that women's insecurities about social status combined with economic pressures to push them into relationships faster than their peers -some entering negative relationships, confirming their insecurities about having less value than other women. [29] Annan and Brier (2010) also argued that forced marriage as part of armed conflict may have similar longterm psychological impacts on women as childhood experiences of violence, with similar consequences for difficult relationships with men in adulthood.
The qualitative results predominantly related to violence against women, but one study, Kohli et al. (2015), highlighted how the stigmatisation of violence against women has affected the mental health of children by contributing to social isolation in South Kivu, DRC: Further, participants described how living in an unstable and violent household negatively affected children's interaction with community members, as the entire family could become isolated and disrespected in the community. Male and female participants in this study described households with IPV as unable to progress or have stability in their lives. [31]

Study quality
Quality assessments for both quantitative and qualitative studies are in the Additional file 1: Appendix. Overall, there was a low risk of bias in most of the quantitative studies. Bias existed in identifying and dealing with confounding, which we assessed as high or unclear in seven studies [30,37,40,45,48,50]. Sampling criteria were defined in all studies, with most adopting a version of random sampling. In some, for example O'Leary et al. [27], purposive sampling was used, leading to a likely bias.
The qualitative studies were generally of high quality, with the exception of the categories of cultural and theoretical positionality and reflexivity, for which we felt that all studies were lacking.

Discussion
Conflict has a pervasive impact on people and societies. Conflict-related violence is well documented, but less well known is how exposure to armed conflict and its social and economic consequences can influence family violence and mental health problems. Our results provided evidence that family violence is common and is associated with poor mental health outcomes, but variability in outcomes and measurement meant that we were not able to support or refute the hypothesis that the prevalence increases in conflict settings nor that it changes the association with mental disorders.
Worldwide, approximately one in three women have survived physical or sexual violence, but the prevalence varies greatly by location. By WHO region, lifetime prevalence of physical or sexual intimate partner violence is highest in the Pacific (Melanesia 51%, Micronesia 41%, Polynesia 39%), South Asia (35%) and sub-Saharan Africa (33%) [52]. According to a recent WHO report, many countries with a recent history of armed conflict, including Afghanistan and Papua New Guinea, have some of the highest prevalences of physical and sexual IPV (46% and 51%, respectively), but data from conflict settings remain poor [52]. Our review found similar proportions, but the studies varied widely, most likely representing actual differences in prevalence, but also methodological issues such as study design, data availability, and tools used.
Studies of violence against children showed the almost ubiquitous presence of violence. It is estimated that one billion children have suffered from violence in the preceding year [4]. A systematic review by Hillis et al. from 96 countries (n = 38) summarised the prevalence of violence over the past year in any country. In Africa, Asia and North America, approximately half to two-thirds of children suffered violence. Prevalence in Latin America, Oceania, and Europe was a little lower, mostly at around one-third [4]. Direct comparison is limited as the countries included in the review by Hillis et al. did not overlap with ours, with the exception of Uganda, where they reported data from a randomised controlled trial [53]. Studies in our review reported recent violence at a similar prevalence in the studies from Afghanistan. The two Sri Lankan studies appeared to show a higher prevalence, reporting 64.2% [12] and 71.6% [41] in the past month alone. Amongst Violence Against Children and Youth Surveys (VACS) studies, only one country -Ugandahas comparable data [54]. This study reported a similar prevalence of lifetime family violence (59% against girls and 68% against boys). Conflicts themselves vary and the impact on individual families is different. Other factors may also affect health. For example, in a study from Sri Lanka, the population also experienced the Indian Ocean tsunami [12].
The finding of an association between violence and mental health problems is not new [55,56], and was reported in approximately half the studies of violence against women in our review. The studies were not amenable to meta-analysis, nor could comparisons be drawn readily with non-conflict settings. The most consistent association was with PTSD. This may be due to an increased risk of PTSD, as has been commonly shown in conflict-affected populations, but may also reflect a bias in the topics that are researched. For studies that included suicide risk and substance use as additional outcome measures, the association between exposure to family violence and prevalence broadly followed the dose-response relationship observed for PTSD.
Amongst studies of violence against children, consistent associations were found between family violence and symptoms of mental health problems. Three studies constructed structural equation models to explore associations [13,26,28]. Heath et al. [26] and Jewkes et al. [13] treated violence and mental health problems (PTSD and depression) as co-outcomes and did not examine the association between the two. Jewkes et al. showed the importance of household wealth and previous childhood trauma [13]. Saile et al. [28] proposed maternal care as a mediating factor that could reduce adverse mental health symptoms.
Armed conflict can challenge dominant masculinities that cast men as providers and heads of family in ways that contribute to perpetration of family violence [31,33,36]. Gender norms undermine women's mental health during conflict by contributing to shame and stigma associated with experiencing both sexual violence and family violence more broadly, and this can also contribute to the social isolation of children [31]. However, the same destabilising effects of armed conflict can also give women new economic opportunities and contribute to their economic and social empowerment [29,30,35].

Strengths and limitations
The main strengths of the review included its robust search strategy, its use of dual reviewers, and analysis and synthesis of both quantitative and qualitative data. The review protocol was prospectively registered and the reporting followed established standards. Limitations should, however, be noted. Studies were highly heterogeneous and meta-analysis could not be conducted and pooled estimates of the association between exposure to family violence and mental health problems could not be calculated. Violence is likely to have been underreported, and the extent of under-reporting and other biases in data collection would vary between studies and settings. Although a standardised approach for research on violence against children has been advocated (the Violence Against Children and Youth Surveys (VACS)), the studies included in the review did not follow this approach. Variation in outcomes would in part be due to the definitions used, with no set definition of family violence being accepted [57]. Our definition of family violence resulted in the inclusion of studies of violence against women and violence against children, but rarely was family violence explicitly defined within the studies.
Although we found consistent evidence of association between exposure to family violence and mental health problems for both adults and children, we can make limited inferences regarding causality due to the predominance of cross-sectional studies within the review. It is possible that both family violence and changes in mental health are outcomes of exposure to conflict, as described by Heath et al. [26] and Jewkes et al. [13], and reverse causality may occur [58].
In reviewing the qualitative studies, two forms of maltreatment of women (and children) not captured by our search terms are potentially salient for women's mental health and could be rationalised under a broader conceptualisation of family violence. First, rejection and neglect following women and children's return from conflictrelated assault or time embedded with conflict groups affects their ability to reintegrate into family life, with implications for mental health. Second, sexual violence by combatants to entrap girls and women into forced unions could be viewed as initiating acts of both intimate partnerships and IPV.

Conclusions
Our systematic review showed that family violence was common in conflict settings and was associated with mental health outcomes. However, we were unable to determine whether prevalence or risk was greater than in non-conflict settings. This gap in the review stems from a lack of comparable data and reiterates the need for standardised measurement of both violence against women and violence against children. It also draws attention to the need to break down existing silos between research on violence affecting women and that affecting children to build a more comprehensive picture of the overlapping risks and mental health outcomes of violence within families [59]. A far more robust understanding of the impacts of armed conflict on families and their experiences of violence is urgently needed to develop comprehensive support systems in conflict-affected settings.
Additional file 1: PRISMA checklist, search terms, coding framework for qualitative papers and quality assessments.