Background
Intimate partner violence (IPV) is a critical human rights and public health concern. IPV refers to behavior within an intimate relationship that causes, or has the potential to cause, physical, sexual, or psychological harm, including acts of physical aggression, sexual coercion, psychological abuse, and controlling behaviors [
1]. A comprehensive meta-analysis of 141 studies from 81 countries found that 30% of women and girls aged 15 and older have experienced IPV [
2]. Consequences of IPV include physical, reproductive, and mental health issues [
3‐
5] and in severe cases, the resulting injuries can be fatal [
6]. Knowledge is accruing on how to best prevent and reduce IPV [
7]. Research from low- and middle-income countries (LMIC) has more frequently focused on preventive interventions, and has shown promising benefits of group training for men and women (e.g., participatory learning activities focused on gender roles and conflict resolution skills), community mobilization interventions, and combined livelihood and training interventions for women [
8]. With regard to efforts to reduce IPV once detected, evidence (mainly from high-income countries) suggests that women-centered care, advocacy, and home-visitation programs can reduce the risk of further victimization [
8,
9]. Although treatment of mental ill health or substance abuse may strengthen efforts to prevent and reduce IPV [
10] relatively little research has focused on this topic.
There are several reasons to think that treatment of mental disorders and substance abuse problems may be an effective strategy for prevention and reduction of IPV in LMIC either through targeting perpetrators or victims. Hazardous alcohol use [
11‐
14], common mental disorders (posttraumatic stress disorder [PTSD], depression, anxiety) [
12‐
16], and anger dysregulation [
17] are known correlates of IPV perpetration. Yet, attention to perpetrators’ mental disorders has not commonly been included in batterer intervention programs such as the Duluth model. Duluth interventions tend to focus on gender reeducation aimed at addressing the patriarchal factors underlying male perpetration of IPV. Evaluations of traditional batterer intervention programs based on this model, commonly in high-income countries, have shown conflicting results [
18]. Interventions that include components to address perpetrators’ mental ill health may strengthen the effectiveness of efforts to stop or reduce IPV, given the strong correlations of mental health with IPV perpetration [
19‐
22].
Mental health interventions may also reduce further risk for victimization by treating mental health problems among IPV survivors [
23‐
25]. Longitudinal studies suggest that the relationship between IPV and mental ill health may in fact be cyclical: mental health impacts of IPV put women at increased risk for further abuse [
26‐
29]. For example, depression may be associated with self-blame for IPV victimization, reduced self-esteem, and hopelessness. Similarly, PTSD symptoms such as emotional numbing may challenge survivors’ ability to detect or respond to IPV risks [
28,
30]. Mental health interventions therefore may reduce IPV re-victimization by targeting mental ill health in survivors [
21]. Consistent with this hypothesis, a randomized controlled trial from the United States, providing cognitive behavioral therapy to interpersonal violence survivors reduced IPV re-victimization [
30]. Both for survivors and perpetrators, mental health treatments may have additional indirect benefits for IPV reduction by conferring psychological and social skills - strengthening communication, stress management, and anger management skills, and reducing social isolation – that may reduce IPV incidence [
31].
We follow a multidimensional (“both/and”) perspective, where attention to mental health occurs as part of an analysis of the various factors – both individual and structural -- that contribute to IPV [
22]. To date, less attention has been placed on some of the individual-level factors (e.g., witnessing domestic violence in childhood, experiencing child abuse, alcohol abuse) which are strongly and consistently related to IPV [
12,
13,
21,
32]. We note that interventions addressing the mental health of survivors need to be mindful of the risk of victim blaming, i.e. pointing to individual characteristics associated with higher risk for IPV without acknowledging the broader structural forces that confer risk for IPV. We also highlight the wider constellation of risk and protective factors for IPV [
33] noting both family- and community-level factors (e.g., relationship practices, household poverty) [
12,
21] and broader socio-cultural factors (e.g., gender-inequitable social norms and traditional notions of masculinity) [
12,
13,
34] which are important correlates of IPV perpetration and victimization. We examine individual-level factors, in particular mental health, without diminishing the importance of these broader social and structural influences.
Research findings from high-income countries may not generalize to LMIC because of differences in the distribution of determinants of violence; socio-cultural context; resources available to respond to IPV; notions of mental illness; and characteristics of mental health systems. Given the potential of mental health interventions to address IPV and the gap in knowledge on this topic in LMIC, in this systematic review we synthesize findings from controlled trials of mental health interventions conducted in LMIC that included IPV as a primary or secondary outcome.
Methods
Inclusion and exclusion criteria
The protocol for this systematic review was registered with PROSPERO (2017: CRD42017064660). We included prospective, controlled studies (either via comparison group or statistical design that enabled “self control” comparisons) that assessed the impact of mental health treatments on IPV among a sample primarily composed of adolescents and adults (at least 50% of the sample was age 10 or older) in LMIC. Low and middle-income countries were defined using the latest World Bank income classifications, including both lower- and upper-middle income countries. An intervention was considered a mental health treatment if it met all of the following criteria: 1) included a mental health component, i.e. an element theorized explicitly by study authors to target mental health symptoms or substance use; 2) evaluated a pharmacological or psychosocial program delivered to individuals screened in on the basis of mental ill health or substance misuse, either by using a defined disorder diagnosis or scoring above a pre-defined cutoff on a screener for symptoms of disorder or general psychological distress; 3) measured a mental or substance use disorder or symptom as an outcome; and 4) included a measure of IPV, either physical, sexual or psychological, as a study outcome (primary or secondary). We excluded studies where violence was measured as occurring between: people in the general community; paying sexual partners; or family members that were not intimate partners (i.e. father and child, mother and daughter-in-law, elder abuse within families). We did not set any restrictions by year of publication.
Trials with all types of inactive control conditions were considered for inclusion, including placebo, waitlist, no treatment, treatment as usual, or treatment without an active mental health component. We also included studies with only one treatment arm that adequately controlled for unobserved confounding in design and analysis (e.g. regression discontinuity designs, instrumental variable approaches, difference-in-difference designs, or interrupted time series). We excluded studies that compared two or more active treatments without a control condition. We excluded non-peer reviewed literature (e.g., book chapters and dissertations). We excluded studies that did not have an abstract in English. If an article had an abstract in English but was written in any other language, the article was still eligible for inclusion.
Search strategy
Our search strategy combined terms aimed at identifying studies that: (1) were conducted in LMIC; and (2) evaluated mental health treatments (i.e. had a mental health or psychosocial component and were delivered to individuals experiencing ill mental health); and (3) assessed IPV as an outcome; and (4) were controlled prospective studies. To ensure identification of studies conducted in LMIC, we applied a set of keywords developed by Johns Hopkins University librarians to include general terms used to describe LMIC (e.g. developing country, less developed nation, third world) as well as the names of all countries classified by the World Bank as low- or middle-income.
To identify studies focused on mental health treatments, search terms included names of mental disorders, categories of disorders, and commonly abused substances, as well as general terms for mental ill health (e.g. psychological stress, aggression, mental disorders). To ensure identification of studies assessing IPV outcomes, we included various terms for IPV (e.g. partner abuse, marital abuse, rape), as well as search terms describing specific forms of abuse (psychological abuse) and the broader term violence. To identify studies with intended research design, we used Cochrane recommended search terms for randomized controlled trials (see
http://work.cochrane.org/pubmed) and added terms for non-randomized controlled studies and rigorously designed prospective observational studies that adequately controlled for confounding.
A search strategy was initially developed by selecting multiple medical subject headings (MeSH) terms and subheadings relevant to mental health problems (e.g. mental disorders; stress; psychological) and interventions (e.g. psychotherapy; psychotropic drug) in PubMed. Together with a university librarian, this search was iteratively refined by examining search strategies from relevant reviews (e.g. [
35‐
37] of the impact of mental health interventions and a search of keywords of relevant and retrieved irrelevant articles. This search strategy was then adapted for use across different databases, given databases’ different restrictions on searches (e.g., limited number of terms to search) and thesauruses (e.g., the use of MeSH in PubMed). As an example, our initial PubMed/Medline search strategy is provided in Additional file
1.
Other databases searched were Web of Science (including the Social Science Citation Index); Scopus (including Medline and Embase); Ebscohost (AfricaWide, psychINFO, CINAHL); and ProQuest (PILOTS and IBSS). In addition, we hand-searched the following regional databases, trial, and funding registries: Cochrane Central Register of Controlled Trials,
ClinicalTrials.gov, EU Clinical Trials Register, ISRCTN Registry, National Institute of Health (NIH) Reporter, and WHO databases (Western Pacific Region Index Medicus, WHO Global Index Medicus, South East Asia Regional Office, Eastern Mediterranean Regional Office, African Index Medicus). Additional hand searching included the reference list of any relevant systematic reviews or published trial protocols found through this search process, as well as forward and backward citation checks on any article eligible for inclusion. We also reached out to all authors of included articles to ask if they, as experts in this area, knew of any articles that we had missed in our search process.
Search results from all databases and registries were compiled and duplicates eliminated by a single researcher using Covidence software. Two researchers then independently: (1) screened titles and abstracts, and (2) screened the full text of any article that was found to be potentially eligible. In the case of conflicting decisions on eligibility, the two reviewers (SM, LS) discussed the discrepancy and their rationale. If consensus could not be reached, a third party (WT, JB) was consulted for a final decision.
One author (LS) extracted information from eligible full texts into a piloted, structured Excel spreadsheet. A second author (SM) checked all the extracted information, and consulted a third reviewer (WT, JB) as necessary when her interpretation of an article varied substantially from the first reviewer or the information provided in the manuscript was unclear. The data extraction spreadsheet included entries for: sample and population characteristics (e.g. country, sample size, demographics, mental health condition targeted); study design and procedures (number and timing of assessments, statistical methodology, instruments used to measure mental ill health and IPV); intervention information (type of intervention, mode of delivery, duration and dose, if it included a gender transformative or violence specific component, control condition); study findings (intervention effects for IPV and mental ill health/ substance abuse outcomes, results of any analysis of mediators, sub-group analyses); and information related to risk of bias. Risk of bias assessments were made using the Cochrane Risk of Bias Tool by two authors independently (SM, WT). This tool included the following dimensions: selection bias (sequence generation; allocation concealment); performance bias (blinding of participants and personnel); detection bias (blinding of outcome assessment); attrition bias (incomplete outcome data); and reporting bias (selective outcome reporting).
We planned to conduct a narrative synthesis and, if a sufficient number of high-quality trials were identified with sufficient homogeneity, meta-analysis using aggregate data.
Discussion
This systematic review aimed to synthesize findings from controlled studies in LMIC of the impact of mental health treatments on the prevention and reduction of IPV. Despite our search across a wide range of databases and screening more than 1000 titles and abstracts, we found only seven studies that have evaluated the benefits of mental health treatments with regard to IPV in LMIC. These studies were conducted in five middle-income countries. Six were RCTs and one had an interrupted time series design. The research topic addressed here does seem to be an area of increasing interest: as part of our searches we identified 10 published protocols for planned or ongoing trials that likely meet inclusion criteria once completed (available upon request).
The main observation from this review is that the current literature is limited in scope, resulting in critical gaps in our knowledge. In our view, the main overall gaps concern: (1) no studies have replicated evaluations of similar treatments; (2) no studies were conducted in low-income countries; (3) there was a lack of diverse samples (e.g., no studies in humanitarian settings or with refugee populations, where rates of IPV are particularly high [
53]; no studies with older adolescents; no studies with sexual minorities); (4) limited geographical coverage (e.g., no studies from Latin America and the Caribbean, Middle East and North Africa, West Africa); and (5) a limited range of mental health concerns were targeted (e.g., no studies focused on anger dysregulation or medically unexplained complaints).
Benefits of mental health interventions for IPV outcomes
Though limited in number, studies evaluated both integrated interventions (
n = 3) and dedicated mental health treatments (
n = 4). Integrated interventions were aimed at concurrently reducing HIV/STI risk, violence against women, and alcohol or other substance misuse treatment, whereas dedicated mental health treatments targeted common mental disorders (depression, PTSD) or alcohol misuse. As the driving hypothesis for assessing the impact of these interventions on IPV is that mental health concerns are a possible pathway through which IPV perpetration or victimization can be stopped, reduced, or prevented, it should be noted that the studies included in our review produced mixed findings in relation to mental health outcomes. In evaluations of three integrated interventions, one [
40,
41] produced null findings for harmful alcohol use. In the second, improvements were found for men’s depression and suicidal thoughts, but alcohol use increased in women [
38]. In the third [
42] reductions in drug use were identified.
Somewhat more promising findings were identified for the four dedicated mental health treatments: all four showed impacts on mental health outcomes. Of the dedicated mental health treatments, the two studies that focused on depression also identified reductions in IPV: one found reductions for a combined perpetration/ victimization measure in men and women, and one found reduced IPV victimization in women [
39,
45] These benefits were not maintained at 12 months in one of the studies [
46]. The two studies that focused on alcohol misuse did not identify benefits with regard to IPV victimization or perpetration [
43,
44,
47,
48].
Recommendations
It is challenging to draw firm conclusions from this limited pool of studies. In this regard, our review echoes findings from previous broader reviews of interventions to prevent or reduce violence against women and girls, which note the lack of literature particularly from LMIC [
54,
55]. The ‘What Works to Prevent Violence’ program’s synthesis of existing knowledge [
55] specifically highlights a knowledge gap with regard to the impact of mental health on perpetration and experience of violence. Further research is urgently needed. Based on the existing studies, we believe several research strategies would improve research on this topic as it develops.
Strengthen theoretical underpinnings
Critically, research on this topic would benefit from stronger theoretical development. This is because existing studies commonly did not detail the specific pathways through which improvements of mental health conditions were hypothesized to impact IPV perpetration or victimization. In the case of dedicated mental health interventions, IPV was included as a secondary outcome. More explicit thought regarding intervention adaptation and study design for the specific question of whether mental health interventions can address IPV is an important step to developing evidence better geared for answering this question. Several ongoing studies have begun to do this [
56,
57]. For example, one study with IPV survivors is examining whether improved social support, coping, and support seeking as a result from a group psychological intervention is associated with subsequent reductions in IPV victimization [
57]. Similarly, as noted in the introduction, evaluations of mental health interventions may assess whether changes in self-esteem, self-blame, or emotional numbing lead to reductions in IPV victimization, or whether improvements in managing strong emotions are associated with reductions in IPV perpetration.
The few benefits identified in the current literature suggest treatment strategies and/or pathways that could be targeted for more explicit study and replication. E.g., Patel, Weobong and colleagues identified reductions in physical IPV victimization amongst women after a behavioral intervention at immediate post-treatment follow-up, but not at the 12-month follow-up. Similarly, Jiang and coworkers’ small pilot trial (2014) found impacts on depression and PTSD and reduced interpersonal therapy on a combined measure of IPV perpetration and victimization in male and female earthquake survivors. They hypothesized a mental health and IPV connection based on formative qualitative research that indicated the disaster had heightened interpersonal conflicts among affected couples. This study points to how mixed-methods research may strengthen conceptual models for future testing of complex interventions [
58].
Building on improvements in conceptualizations of pathways, research on this topic would benefit from directly testing hypothesized relationships between mental health and IPV perpetration or victimization. This can, for example, be achieved through by measuring treatment effects on potential mediating variables, where reductions on mental health variables can be statistically assessed for correlation with subsequent reductions in IPV. This will be particularly important for complex interventions, where multiple components may explain prevention or reduction of IPV. In addition, theoretical models may be tested through more complex trial designs, e.g. trials with arms including combined mental health and violence protection activities, mental health and violence protection alone, against a control condition.
Improve IPV measurement
A major limitation of the current body of studies is that they may have been underpowered to detect changes in IPV. This may partly be due to the common use of dichotomous rather than continuous outcome measures for IPV. In addition, power in the existing studies was likely reduced because studies were not specifically geared to addressing changes on IPV, using samples in which IPV was present only in subsamples. IPV was commonly included as a secondary outcome measure, with studies only specifically powered to detect changes on primary outcomes. In addition, a few studies used non-standardized measures for IPV with unknown psychometric properties, including sensitivity to change.
Strengthen mental health benefits in integrated interventions
The limited mental health impacts of the integrated interventions raise important questions for future studies. The lack of clear mental health benefits from the integrated interventions (one out of three studies), may be an issue of content or dose (e.g., number of hours) of specific mental health-focused content. In integrated interventions, this dose may be smaller given that other content within the integrated intervention may target social determinants of poor mental health (i.e., IPV, social isolation, and poverty). At the same time, increasing the amount of mental health related content needs to be weighed against feasibility concerns in low-resource settings. To ensure feasibility, ongoing efforts are required to improve the scalability of existing evidence based mental health interventions in LMIC. For example, transdiagnostic interventions, by combining treatment techniques for commonly comorbid mental health concerns, reduce the number of protocols for training health workers, making it more attractive for agencies not specialized in mental health to introduce mental health interventions as part of their work [
59,
60]. Similarly, alternative intervention formats (e.g. electronic delivery or self-help formats) may increase feasibility and thereby uptake by non-specialized agencies [
61].
Consider promising strategies from HIC
Several promising strategies from HIC are also worth exploring in LMIC. For example, a more narrow systematic review of cognitive behavioral and advocacy interventions (which commonly have psychosocial support elements) with IPV survivors in high-income settings identified 12 randomized controlled trials, and found that both showed impacts on physical and psychological but not sexual and combined IPV victimization [
62]. With regard to IPV perpetrators, there is limited evidence from HIC on cognitive behavioral interventions for male perpetrators of physical IPV [
18]. In addition, an emerging literature in high-income settings has found promising results for innovative approaches, e.g. for couples in which male perpetrators receive treatment for anger dysregulation, alcohol misuse, and/ or common mental disorders [
19,
20]. IPV perpetration interventions may specifically target people at higher risk for perpetration, e.g. men with histories of childhood trauma and current anger dysregulation concerns, including harsh parenting [
63].
Limitations of systematic review
Our a priori aim was to include studies focused on reducing (symptoms of) mental disorders (i.e., studies focused on treatments evaluated with populations screened to have higher levels of symptoms or meeting criteria for disorders). Mental health is more than the absence of symptoms, and positive aspects of mental health may be targeted as protective factors against IPV. For example, positive parenting practices during childhood may be associated with reduced IPV perpetration in adulthood [
64]. Inclusion of studies focused on promoting positive aspects of mental health may have resulted in the identification of studies dedicated to the primary prevention of IPV. In addition, at the title and abstract screening phase we only included studies with an abstract in English.