Armed Conflict, Intimate Partner Violence, and Mental Distress of Women in Northeastern Uganda: A Mixed Methods Study

Abstract

As global mental health research and programming proliferate, research that prioritizes women’s voices and examines marginalized women’s mental health outcomes in relation to exposure to violence at community and relational levels of the socioecological model is needed. In a mixed methods, transnational study, we examined armed conflict exposure, intimate partner violence (IPV), and depressive symptoms among 605 women in Northeastern Uganda. We used analysis of variance to test between groups of women who had experienced no IPV or armed conflict, IPV only, armed conflict only, and both; and linear regression to predict depressive symptoms. We used rapid ethnographic methods with a subsample (n = 21) to identify problem prioritization; and, to characterize women’s mental health experiences, we conducted follow up in-depth interviews (n = 15), which we analyzed with grounded theory methods. Thirty percent of the sample met the cut-off for probable major depressive disorder; women exposed to both IPV and armed conflict had significantly higher rates of depression than all other groups. While women attributed psychological symptoms primarily to IPV exposure, both past-year IPV and exposure to armed conflict were significantly associated with depressive symptoms. Women identified socioeconomic neglect as having the most impact and described three interrelated mental health experiences that contribute to thoughts of escape, including escape through suicide. Policy efforts should be interprofessional, and specialists should collaborate to advance multi-pronged interventions and gender-informed implementation strategies for women’s wellbeing.

Depressive and anxiety disorders are leading global mental health problems, with low- and middle-income countries shouldering much of the burden (Institute for Health Metrics and Evaluation, 2018). Within low- and middle-income countries, vulnerable populations include the poor, rural populations, refugees, youth, people with low levels of education, Indigenous populations, girls, and women, all of whom present with high mental distress prevalence and suffer from lack of access to care (Saxena, Thornicroft, Knapp, & Whiteford, 2007). Women, for instance, are more likely than men to develop depression, an epidemiological trend found globally, irrespective of national development level (Nolen-Hoeksema, 1987). Risk of developing depression is multifactorial and influenced by a host of biopsychosocial vulnerabilities. However, exposure(s) to adverse environmental experiences, such as armed conflict and intimate partner violence (IPV), appreciably exacerbates potential for depression, warranting increased attention in global mental health research and programming. IPV disproportionately affects women, especially in low- and middle-income countries (World Health Organization, 2013). We examined mental health outcomes and their relation to exposure to violence, including armed conflict, among women from a low-income country, Uganda.

Armed Conflict and Mental Health

Environmental exposure to armed conflict, whether between recognized nations, national factions, or non-state involved conflicts, has wide-reaching mental health consequences for conflict-affected civilians (Human Security Research Group, 2014). Refugees or asylum seekers forcibly displaced world-wide in 2015 totaled 65.3 million people (United Nations High Commissioner for Refugees, 2016), and the vast majority of these displacements affected people in low-income nations (Internal Displacement Monitoring Center, 2016). In a meta-analysis review of 161 international studies, exposure to armed conflict was associated with mental illness across cultures (Steel et al., 2009). Conflict-affected Georgians have exhibited high rates of somatic distress, which correlated with trauma exposure, PTSD, and depression (Comellas et al., 2015). In areas subjected to armed conflict and displacement in Northern Uganda, 63% of women reported suicidal ideation (Medecins Sans Frontieres, 2005). Uganda has a suicide mortality rate of 9.9 per 100,000 people, slightly higher than the regional prevalence (World Health Organization, 2018b). In Sudan and Uganda, severity of trauma exposure was related to PTSD symptom level (Neuner et al., 2004).

The psychological impact of exposure to violence in conflict-affected areas is attributable, not only to the initial effects of traumatic events, but also to the effects violence has on loss of material and psychosocial resources over time (Hobfoll, 1998). Loss of resources following traumatic distress can exacerbate distress, which in turn predicts more resource loss in a transactional pattern, exposing survivors to prolonged distress (Heath, Hall, Russ, Canetti, & Hobfoll, 2012). For example, in Northern Ugandan displacement camps, women’s disproportionate labor demands and the necessity of “risk-taking” (e.g., for collecting firewood and finding food) in order to provide for the family’s basic needs, served as additional stressors and placed women at increased risk of sexual assault from soldiers and rebels (Okello & Hovil, 2007), a process documented in other contexts (Akhter & Kusakabe, 2014).

Intimate Partner Violence, Armed Conflict, and Mental Health

Intimate partner violence—or intimate partner behaviors that cause psychological, physical, or sexual harm (World Health Organization, 2016)—is associated with armed conflict (Stark & Ager, 2011). The relation is challenging to measure (Peterman, Palermo, & Bredenkamp, 2011) because a lack of infrastructure and challenges with the rule of law in low- and middle-income countries, can discourage reporting, particularly in conflict-affected regions (Okello & Hovil, 2007). Notwithstanding, the prevalence of gender-based violence in emergency regions has exceeded that of non-emergency contexts, and rates of IPV have surpassed sexual violence perpetrated by non-partners (Stark & Ager, 2011). For example, in the Democratic Republic of Congo, rates of intimate partner sexual violence were especially high in conflict-affected regions and 1.8 times more likely to occur than rape perpetrated by non-partners (Peterman et al., 2011). IPV, in conjunction with other psychosocial stressors, such as poverty and partner and familial alcohol abuse, has emerged as a mechanism of post-conflict hardship for women in Uganda (Annan & Brier, 2010).

While researchers have documented the association of IPV and armed conflict across contexts, fewer researchers have examined pathways between the two types of violence. In previous research (Mootz, Stabb, & Mollen, 2017), we conceptualized pathways between armed conflict and IPV according to a socioecological model (Bronfenbrenner, 1977) from a feminist perspective. Bronfenbrenner first conceptualized the socioecological model—a stratified, nested model—to account for multilevel influences on human development. Now widely used in public health research, the World Health Organization (2018a) has employed this model to organize multifactorial variables related to prevention of interpersonal violence at individual, relational, community, and societal levels. Heise first applied the socioecological model to explain the numerous variables contributing to violence against women. We have found multilevel and intersecting pathways between exposure to armed conflict and IPV include experiencing looting and loss of resources from armed conflict (e.g., community level), male alcohol use and poor mental health of men (e.g., individual level), inequitable decision-making about women’s healthcare in couples (e.g., relational level informed by societal-level gender role expectations), and women challenging husbands (e.g., relational level informed by societal-level expectations; [Mootz et al., 2018; Mootz, Stabb, & Mollen, 2017]).

The socioecological model has also been used to organize outcomes of exposure to interpersonal violence (Laisser, Nystrom, Lugina, & Emmelin, 2011). Although women who experience IPV at the relational level are at amplified risk of common mental disorders (Ouellet-Morin et al., 2015), the often cumulative nature of exposure to traumatic stressors, such as when the enduring IPV occurs with conflict-induced stressors at the community level, takes a toll on women’s mental wellbeing (Neuner et al., 2004). As with armed conflict, poverty (also a community-level variable) has compounded the difficulties faced by survivors of IPV, heightening vulnerability to mental illness (Goodman, Smyth, Borges, & Singer, 2009). Complex and prolonged interpersonal traumas have been closely linked to individual-level, post-trauma responses of depressive symptoms, such as negative mood and cognition (Steel et al., 2009). As such, survivors of IPV in conflict-affected areas are at higher risk of mental health challenges than those not affected by IPV (Gupta et al., 2014).

The Uganda Context

Uganda is a diverse country with nine major ethnic groups and over 50 languages spoken (Central Intelligence Agency, 2018). This study took place in Northeastern Uganda where conflict and cattle raiding (violent looting of cattle and other livestock) among subtribes of the pastoral, nomadic Karamajong, spilled over into the rural sub-region of Teso. The conflict began in the 1940s but has received little international and national attention (Internal Displacement Monitoring Center & Norwegian Refugee Council, 2010). Large-scale displacement into protected camps occurred in Teso and the Lord’s Resistance Army invaded Teso in 2003—committing mass murder, rape, and looting; they displaced civilians, burned their homes, and abducted their children (Internal Displacement Monitoring Center & Norwegian Refugee Council, 2010). The Ugandan government began a disarmament program in Karamoja in 2006, rendering the study site a predominantly post-conflict region. However, several communities remain militarized (i.e., protectorate villages); government soldiers are present to protect against reoccurring cattle rustling activities.

Of the 15 sub-regions in Uganda, the population of around 1.8 million residents in Teso is the least urbanized and one of the poorest (Uganda Bureau of Statistics, 2017). About 10% of the Teso population has not attended school, and 6% of the population has completed secondary education. Women in Teso have less access to resources than men. For example, 15% of women and 33% of men own mobile phones (Uganda Bureau of Statistics, 2017). Lifetime use of the Internet also differs by gender (5% women; 20% men), and women have a lower literacy rate (64% women; 86% men; Uganda Bureau of Statistics & Inner City Fund [ICF], 2018; ICF was founded as Inner City Fund). Uganda has the seventh highest birthrate globally (Central Intelligence Agency, 2018), and Teso has one of the uppermost fertility rates nationally (on average, six children per woman) and the highest percentage of adolescent girls who have birthed a child (Uganda Bureau of Statistics & Inner City Fund [ICF], 2018). Exposure to intimate physical, sexual, or emotional violence is slightly more prevalent in Teso (61%) than nationally (56%; Uganda Bureau of Statistics & Inner City Fund [ICF], 2018).

The Present Study

As in prior studies, in the current study, we utilized the socioecological model as a theoretical base because of its ability to concurrently examine multifactorial influences, including different forms of violence, on women’s mental health. Literature that prioritizes women’s voices and examines women’s mental health outcomes in relation to exposure to both community-level (i.e., armed conflict) and relational-level (i.e., IPV) violence in low- and middle-income countries is scarce. Mixed methods approaches that examine ways in which quantitative and qualitative results converge or diverge can develop understanding of the complexities of these difficult problems (Creswell & Clark, 2018). Furthermore, despite the documentation that poverty has a transactional relationship with mental health, few researchers have examined the effect of economic exposure (i.e., loss or destruction of property, agriculture, or other resources) to armed conflict on mental health. Understanding these issues may help identify treatment targets and inform behavioral health strategies.

We aimed to address gaps in the literature, by examining associations between exposure to armed conflict, IPV, and depressive symptoms. Because of the lack of research in these areas, we maintained a two-tailed hypothesis that (1a) there would be significant differences in either direction in depressive symptoms between groups exposed to different types of violence (i.e., armed conflict and IPV). We also hypothesized that (1b) both types of violence, (1c) amount of resources, and (1d) economic exposure to armed conflict would be significantly associated with depressive symptoms. We were informed by transnational feminist literature that critically contests Western definitions (Mohanty, 1988), so we assumed that Western-developed mental health symptom diagnoses may not hold the same meaning for rural Ugandan women as they do for Western women. Thus, we sought to establish women’s prioritization of problems and to characterize IPV survivors’ conceptualization of their mental health experiences to triangulate and contextualize quantitative data.

Method

Participants

Participants were girls and women, whose ages ranged from 13 to 49 years (M = 29.88; Table 1; Mootz et al., 2018). The majority were partnered at the time of the survey (92.8%), and almost all (98.6%) had been partnered in their lifetime. The number of children women had ranged from 0 to 13 years of age, with the average number of children around four. Almost 15% (n = 89) had received no education with nearly 60% (n = 363) identifying as being unable to read or write. Women reported they were primarily Anglican (46%) and Catholic (45.4%).

Table 1.

Participant Demographics.

Variable	Katakwi District (n = 202)	Amuria District (n = 201)	Kumi District (n = 202)	All (n = 605)
Age (years)
M (SD)	30.28 (9.09)	29.33 (9.05)	30.01 (8.55)	29.88 (8.89)
Range	17–49	13–49	18–49	13–49
Partner status (%)
Currently partnered	93.9	94.1	90.6	92.8
Ever partnered	4.0	5.0	8.4	5.8
Number of children
M (SD)	4.08 (2.29)	4.49 (2.68)	3.90 (2.45)	4.15 (2.48)
Range	1–11	0–11	0–13	0–13
Education
No education (%)	18.9	13.9	11.4	14.7
Education years M (SD)	4.05 (3.09)	4.43 (2.71)	5.06 (3.49)	4.51 (3.14)
Literacy (% literate)	34.3	44.3	40.1	39.7
Religion (%)
No religion	0.5	0.0	0.0	0.2
Born again	4.4	6.0	8.4	6.3
Islam	0.0	0.5	0.5	0.3
Catholic	48.8	61.2	25.7	45.4
Anglican	43.8	30.3	63.4	46.0
Other	2.0	1.5	2.0	1.8

Note. M = mean; SD = standard deviation.

All women who participated in the qualitative follow-up identified as subsisting off of agriculture. Their ages ranged from 19 to 46 years (M = 29.24). One woman had no education, 16 women had primary level education, and 4 women had secondary level education. The number of children they had ranged from one to eight (M = 4.19). Fifteen of the women were partnered.

Procedures

We utilized quantitative data from a study examining prevalence and pathways of risk factors (Mootz et al., 2018) and collected additional, original qualitative data to understand mental health outcomes of IPV in conflict-affected communities in the Teso sub-region of Northeastern Uganda. We formed a collaborative partnership between academic institutions in the United States and Uganda, as well as with a local nonprofit organization called Transcultural Psychosocial Uganda (TPO Uganda), to facilitate the study. We obtained Institutional Review Board approval in the United States and locally. TPO Uganda identified six bilingual (Ateso/English) research assistants from Teso to conduct the surveys and qualitative interviews. The first author reviewed procedures (including the safety protocol, which was managed by TPO Uganda) and ethical best practices for researching IPV (World Health Organization, 2001) with the team. The research team piloted the survey administration (n = 21), after which they modified items for local understandability, as needed, reviewing measures item-by-item.

We used a multistage sampling strategy, as was described in Mootz et al. (2018); to ensure varied levels of exposure; our local partners purposively selected three districts in Teso: Katakwi District (1) representing high exposure, Amuria District (2) as medium, and Kumi District (3) as low, based on the anecdotal frequency and severity of exposure to the Karamajong raids. They identified one subcounty, representative of the district’s exposure level, per district. We randomly selected eight villages within each subcounty and surveyed a minimum of 25 participants in each village (N = 605). The research assistants formed three teams of two people each, started in the geographic center of each village, and spun a pen to determine random direction, sampling every third household. Eligible participants included all women (married and single) aged 18–49 and married girls under age 18 (who were considered emancipated; this was approved by our Institutional Review Board). Team members wrote down the given names of all eligible participants, placed numbers next to the names, placed the numbers in a bag, and asked someone in the household to draw a number. Following the establishment of privacy, research teams verbally consented participants due to high illiteracy rates. The team concluded interviews by providing a list of resources and a bar of soap (compensation determined by our local partners). Team members asked women who reported experiencing intimate partner physical or sexual violence to participate in rapid ethnography and qualitative follow-up interviews. Rapid ethnographic methods consist of having participants self-identify problems they are experiencing, define those problems, and determine how those problems rank in order of importance (Bolton & Tang, 2004).

Following the population-based survey, we randomly selected seven women from each of the three surveyed districts from all interested and eligible participants (n = 21). Research pairs obtained verbal consent and began audio recording the rapid ethnographic interview session, facilitated the rapid ethnography session in Ateso, and followed a semi-structured guide. Because women sparingly described mental health experiences during the first interviews, the teams followed up with purposively selected women (n = 15), based on their initial participation level. We determined the sample size based on methods of other IPV-focused qualitative studies per site (Grabe, Grose, & Dutt, 2014; Islam, Jahan, & Hossain, 2018) and, following best practices, interviewed women until data saturation was reached (Patton, 2015). Rapid ethnography sessions and in-depth interviews lasted 30–60 minutes. Each research pair translated their respective interviews into English and transcribed the English translation, discussing translation discrepancies until the pair negotiated consensus.

Instruments

Two native speakers of Ateso, with postgraduate training in English and translation, forward and back translated the survey questions. Research team members forward and back translated the interview questions. Consensus among translators ensued following any translation discrepancies.

Demographics and intimate partner violence.

We assessed demographics and intimate partner violence with a modified Survey of Women’s Health and Life Experiences in Uganda: Woman’s Survey (World Health Organization, 2005). Demographics included, among others, educational level, employment, religion, age, and resource ownership. We measured the outcome variables of intimate partner psychological (4 items: insulting, humiliating, intimidating, and threatening), physical (6 items: slapping, kicking/dragging/beating, pushing/pulling hair, hitting, choking/burning, and threatening with weapon/used weapon), and sexual violence (3 items: raping, having sexual intercourse because of intimidation, and having demeaning sex) with items assessing specific aggressive behaviors (no or yes) (Mootz et al., 2018). We selected this measure because of its widespread use in international settings (Garcia-Moreno, Jansen, Ellsberg, Heise, & Watts, 2006). We adapted this instrument by omitting items that did not measure variables related to research hypotheses. The internal consistency reliability coefficient for the intimate partner violence outcome indicators was α = .88, similar to other low-income settings (Saddki et al., 2013).

Exposure to armed conflict.

The researchers and local team used the Exposure to Political Violence Inventory (Clark et al., 2010) to measure respondent and family exposure to conflict, including verbal, physical, sexual, relocation, abduction, and loss of life (Mootz et al., 2018). Economic exposure items inquired about property theft and damage (loss of livestock, food, and house and household items) and work lost (no or yes). We excluded husband’s armed conflict exposure experiences because our hypotheses pertained to women’s experiences. To establish cultural relevance, we drew from the literature, the first author’s previous research, and the team’s familiarity with common problems faced during conflict in this region, to inform the addition of items deemed to be important experiences of armed conflict in Teso. These items assessed exposure to forms of sexual assault violence (Mootz et al., 2017) and abduction of self or children, the latter of which was a common experience during the Lord’s Resistance Army conflict with the Ugandan government. We added an economic-exposure item inquiring about having livestock stolen, since this experience was commonplace and connected to IPV (Mootz et al., 2017). Following discussion, the first author wrote these items in English, and they were translated and back translated by two advanced Ugandan graduate students studying English literature. Internal consistency of the adapted measure was α = .84.

Mental health distress.

We selected the Hopkins Symptom Checklist (HSCL-25; Harvard Program in Refugee Trauma, 2004) to assess depressive symptoms (15 items). The HSCL-25 has been validated in Uganda (Bolton, Wilk, & Ndogoni, 2004) and used in a number of cross-cultural contexts (Harvard Program in Refugee Trauma, 2004). We scored items as follows: 1 = not at all, 2 = a little, 3 = quite a bit, and 4 = extremely. The item “feeling blue” was translated to “feeling sad.” Internal scale reliability was α = .887 (Depressive subscale), consistent with reliability estimates (.84–.87) found in a validation study with 2,500 participants (Derogatis, Lipman, Rickels, Uhlenhuth, & Covi, 1974). To determine prevalence of probable major depressive disorder (pMDD), we used a binary outcome of 31 or above, consistent with other studies (Kinyanda et al., 2016). For other analyses (i.e., ANOVA and linear regression), we interpreted depressive symptoms as a continuous, item sum score.

Problem priorities and mental health conceptualization.

Adapting a rapid ethnographic approach to mental health assessment (Bolton & Tang, 2004) and using semi-structured interview guides, survivors of IPV described, defined, and ranked problems they had experienced and associated with exposure to armed conflict and IPV, in terms of effect on their life. To ascertain mental health conceptualization, we conducted in-depth interviews using semi-structured interview guides. Because the results of the rapid ethnographic assessment revealed that women perceived their mental health experiences were mostly related to experiences of IPV, our follow-up interviews focused on examining how women understood their mental health as it relates to IPV. The team developed these questions, which were intended to inquire broadly about women’s experiences. Interview questions loosely followed prompts asked in the Cultural Formulation Interview (American Psychiatric Association, 2013). Example questions relating to mental health conceptualization are “How does IPV affect you emotionally?” “What does [emotional effect] mean?” and “How does [emotional effect] disturb you?”

Data Analysis

Quantitative.

We used SPSS 24.0 to summarize demographics, frequencies, and symptom means. We assessed differences (ANOVAs) in probable major depressive disorder frequency between groups of women who had experienced no IPV or armed conflict, IPV only, armed conflict only, and both. We analyzed the relation between depressive symptoms and the hypothesized predictor using multiple linear regression in R (R Core Team, 2016). We screened the predictor variables for multicollinearity using the variance inflation factor (VIF). The VIF values ranged from 1.03 to 1.73, indicating that none of the predictors would seriously distort the analyses (Hair, Black, Babin, & Anderson, 2010). Although there were missing observations, nonresponse was low. The percentage of missingness on the predictor set ranged from 1% to 5%, and there were no missing data for the outcome variable of depressive symptoms. Due to the overall low instance of missing observations and no missing outcome observations, we conducted a complete case analysis (Allison, 2002).

Qualitative.

With the intention of building local research capacity, a commonly espoused value of community-partnered research (Israel et al., 2008), and triangulating our outsider perspective with local analyst input, the first author trained the local research team in grounded theory qualitative methods of data coding and analysis. First, to analyze the problems and priorities, the team of six local research assistants divided into two sub-teams of three persons each. Team 1 members created individual slips of paper for each problem generated by the participants, and then constructed categories by grouping slips of paper together related to IPV and armed conflict. Team 1 presented their categorization rationale to Team 2 members, who served as auditors, a form of analyst triangulation that enhances rigor (Patton, 2015). Team 2 members reviewed and refined conceptualization of the categories, which were determined with consensus by the entire group.

To analyze the mental health conceptualization narratives (Patton, 2015; Strauss & Corbin, 1990), local research Team 2 created an open code list by collaboratively reviewing four transcripts sentence-by-sentence. Both teams divided the remaining transcripts and coded them individually, using the code list as a guide. The teams then organized the coding into themes, relationships, and patterns. They returned to the transcripts to document the axial coding, after which they identified overall themes. A team of three women, co-authors in the United States, similarly coded the transcripts for additional analyst triangulation purposes (Patton, 2015). One team member (Gonzalez) identifies as an Anishinaabe woman, whose passion lies in using research to support and enhance efforts to revitalize tribal cultural ways and language to strengthen communities. Two other U.S. team members (Greenfield and Gill) identify as White, heterosexual, middle class psychologists with practice and research experience with rural and Indigenous populations in the United States. Their work focuses on trauma, resilience, addiction treatment recovery, health equity, and community-partnered research with American Indian communities. The first author (who also identifies as a White, heterosexual, middle-class, female psychologist with research interests in violence against women and girls in conflict settings) provided guidance on coding procedures and cultural phrasing and context, but did not otherwise participate in the U.S. coding process, to avoid influencing the interpretation of the U.S. team with findings from the Ugandan team. She has lived in Uganda and prior to that has worked as an intercultural trainer. The Ugandan and U.S. coding teams identified the same overall themes described in the Results section.

Mixed methods.

Rooted in transnational feminist theory and the politics of intentionally valuing and highlighting—without homogenizing—marginalized women’s voices from the Global South (Mohanty, 1988), our mixed methods approach was primarily qualitative (Onwuegbuzie & Combs, 2011). We viewed women’s narratives and perspectives as core findings and utilized these data to verify quantitative outcomes. To this end, we conducted a convergent analysis, the purpose of which is to compare findings from different methods to expand understanding, by examining the extent to which the quantitative and qualitative results converged or diverged (Creswell & Clark, 2018). We first analyzed quantitative and qualitative results separately and then compared results for each hypothesis. Where possible, below we present results jointly and provide discussion of key ways in which our findings converged and diverged to provide a more in-depth appreciation of women’s mental health experiences related to interpersonal violence (Creswell & Clark, 2018).

Results

Armed Conflict, IPV, and Mental Distress

Thirty percent of participants reported symptoms consistent with probable major depressive disorder, according to the Uganda-validated cutoff of 31 on the HSCL (Kinyanda et al., 2011). To test Hypothesis 1a, we conducted an ANOVA to examine the variation in depressive symptoms between exposure groups; the result was significant, F(3, 530) = 17.55, p < .001, η² = .09). We then tested the pair-wise group differences using a post-hoc Tukey’s HSD comparison. We found significant differences between no exposure and exposure to both IPV and armed conflict (M_Δ = 5.21, p < .001), IPV-only and exposure to both (M_Δ = 4.40, p < .001), and armed conflict-only and exposure to both (M_Δ = 2.20, p < .01). In addition, there were significant differences between no exposure and armed conflict only (M_Δ = 3.00, p < .05), but not between no exposure and IPV only (M_Δ = 0.88, p = .88). No significant differences were found between IPV only and armed conflict only (M_Δ = 2.19, p = .06). Symptom means trended upward from no exposure, IPV only, armed conflict only, and exposure to both. There were no significant associations between district membership and any of the demographic variables.

Confirming Hypothesis 1b, we found that in the presence of other variables, exposure to armed conflict and IPV were both significant predictors of depressive symptoms. Presence of any type of past-year IPV was positively associated with depressive symptoms (β = 2.08, t = 3.77, p < .001), as was exposure to armed conflict (β = 1.80, t = 3.77, p < .05). However, familial exposure to conflict was not related to depressive symptoms (β = 0.66, t = 1.10, p = .27). Overall, the model predictors explained a reasonable amount of variance in depressive symptoms (R² = .19; see Table 2).

Table 2.

Linear Regression Model Estimates for Association of Exposure to Violence and Socioeconomic Variables With Depressive Symptoms.

Predictor	β	SE	t	p Value
Intercept	27.94	1.12	—	—
Resources	−0.67	0.12	−5.78	<.001***
Literacy	−0.04	0.55	−0.10	.9
Age	0.16	0.03	4.91	<.001***
District 2	0.67	0.66	1.03	.31
District 3	−0.24	0.74	−0.32	.75
Past-year IPV	2.08	0.55	3.77	<.001***
Respondent AC	1.80	0.75	2.38	<.05*
Respondent family AC	0.66	0.60	1.10	.27
Economic AC	0.78	0.71	1.09	.27

Note. R² = .19. Age was mean-centered for the interpretability of the intercept; age of 0 is not substantively meaningful. SE = standard error; IPV = intimate partner violence; AC = armed conflict.

^*p < .05.

^***p < .001.

Through rapid ethnographies, women supported these quantitative findings by identifying several forms of shared victimization that arose from exposure to armed conflict and IPV. Table S1, an online Supplemental File available at http://journals.sagepub.com/doi/suppl/10.1177/0361684319864366, contains qualitative results showing the problems and priorities women identified, related to armed conflict and intimate partner violence exposure. Overlapping forms of violence included socioeconomic, physical, relational (i.e., through loss of loved ones), sexual, and psychological abuses.

However, women’s attributions for, and prioritization of, mental distress provided a more nuanced understanding of their perceptions of associations. Contrasting with quantitative findings, that exposure to armed conflict was associated with depressive symptoms, women attributed the psychological outcomes of depression, anger, and traumatization of children as related only to IPV (Table S1 at http://journals.sagepub.com/doi/suppl/10.1177/0361684319864366). The exception to this was feeling restless, which women also attributed to exposure to armed conflict. However, they categorized feeling restless as having less priority.

Reports of Violence, Emotional Distress, and Wellbeing

To gain a deeper understanding of women’s perceptions of IPV and mental distress, we followed up with in-depth interviews. Overall, themes reflected interactions among mental health problems, and connections between these problems and poor physical health. Suicide and escape ideation and attempts were common.

Emotional and mental health outcomes of violence.

Women across all interviews described their mental health status as a constellation of distress, anger, “heart pain,” and having “lots of (negative) thoughts” (see Individual, Figure 1). Women described this constellation of mental health experiences as sometimes synonymous with, or as worsening, negative physical outcomes. When combined with sociopolitical realities that restricted options for women, the constellation of distress may have led women to suicide ideation and to both imagined and real attempts to leave their partners.

Figure 1.

Socioecological model of mental health problems of IPV (intimate partner violence) survivors living in conflict-affected communities in Northeastern Uganda. *Rapid ethnography results show that women prioritized these problems as most problematic.

Anger, also described as bitterness and annoyance, was one core symptom that women described as a powerful feeling, which negatively affected them, and they connected it to external circumstances (e.g., physical abuse and denial of basic necessities). Participants perceived anger as linked to other negative symptoms, including heart pain and physical illness: “When you get angered, you get ulcers and find out that they even move to the brain, thus paining the heart. Fever also arises from only that problem that angered you” (Respondent #3). As is illustrated in the previous quotation, many respondents described anger with indicators of temperature (e.g., fever, hot, burning, boiling).

IPV also affected heart pain. Participants described pain in their heart, bitterness of the heart, and many thoughts running through their heart. Most respondents described heart pain as both an emotional and physical experience. One woman expressed, “Heart paining is when my heart burns, and I feel like black ants are biting it. When my heart is paining, I cannot do anything because I just have to sleep on my chest” (Respondent #12):

Because when GBV (gender-based violence) occurs, you find that your heart starts to thump or beat fast (pukpuk) like you want to die like that. And when you get angry, a thought of committing suicide comes and you start thinking that it is better to die and rest. (Respondent #3)

Thoughts, especially, went hand-in-hand with heart pain. For some, thoughts were located in the heart. For others, thoughts contributed to, or co-occurred with, heart pains:

Thoughts are what brings heart pain. It starts slightly after having a lot of thoughts and then it intensifies after you stop thinking. When [I] am having lots of thoughts. Anger, lots of thoughts with my husband. (Respondent #8)

Thoughts bring too much anger in the heart thus giving the heart to pain…. When you get angry, you feel bad so much and the heart is not settled. You feel so terrible that you want to do something bad that very time. That is what I mean by a paining heart. (Respondent #1)

Many respondents described thoughts as overcrowding the mind, and then traveling to and collecting in the heart center:

The insults he hurls at me are so painful. This makes the heart to also start paining severely, and then the too many thoughts crowd your mind. Then the bitterness (awolanet) reaches the heart and your heart starts beating/pounding fast (pukupuk) like it wants to get off (angedun). You just have to endure. (Respondent #2)

In addition to anger and heart pains, women mentioned a disturbed mind and lots of thoughts as a third integrated key component of mental health experiences: “It means that I have lots of thoughts running in my heart and can’t find answers to them so my mind becomes jumbled with a lot of thinking and thus my mind gets disturbed” (Respondent #4). Thoughts often centered on how the women could provide for their families. They felt unable to adequately provide, abandoned by their partners, and isolated. They would then ruminate about how they might provide:

My thoughts are not stable or at peace. Firstly, in terms of helping in garden work, he doesn’t bother. I had to help people plough in their fields and in return, they helped me plough my own field. That is how I managed to plough here. You look for what to eat solely. This too disturbs my mind. Everything inclusive whether soap is all on me. Am the woman as well as the man. All these things disturb my mind. When I fall sick, no one cares. You struggle alone. This is why my mind is disturbed. (Respondent #1)

For some, having lots of thoughts on provisioning of basic needs was the most bothersome. However, others felt the constellation of symptoms was most unsettling: “To me, I’m really bothered by anger, heart pains, and lots of thoughts. These three are the great ones because I can’t even have peace when one is in me” (Respondent #14).

Escape and suicide ideation.

Many women spoke of wanting to escape their lives to reduce their emotional disturbances or pain. Escape ideation was apparent in the participants’ discussions of leaving home temporarily for momentary relief (e.g., “Better to first go away to get relieved and come back”), leaving home permanently, and suicidal ideation and attempts. Women often imagined separating from abusive partners. Insurmountable barriers to permanent escape included poverty and not having rights to children. Connections to their children often kept them in the abusive relationships. One woman shared:

Sometimes when I want to leave, children come around crying, and I end up staying…. Also, sometimes when I think of children and more, so they also tell me if I leave them, they will suffer because they only have hopes on me. (Respondent #2)

The persistent, intertwined, and chronic course of IPV and poverty provoked thoughts of suicide as a viable option for escape, an experience that most of the women (n = 10) mentioned and many talked about at length. Thoughts of suicide often followed from situations that felt inescapable or unsolvable: “At times I get disturbed, that even if I left for home and maybe the man comes wanting his cows there is no way to pay back, so it’s better I just kill myself” (Respondent #6). This woman’s narrative illustrates how poverty, in this case her family being unable to pay back the husband’s bride price if she separated from him, limited options for non life-threatening escape, begetting suicidal ideation as a final choice. Thoughts of dying were often crisis driven, connected temporally to IPV, and resulting from the aforementioned constellation of distress:

The thing that tries to remove them [suicidal thoughts] is the level of violence at home. Currently, violence is reducing in my home. But when violence increases, the heart boils so much like you want to commit suicide. But now violence seems to be reducing from my home. Thus, my heart is now getting settled, hence not able to harbor such thoughts again. (Respondent #3)

For many participants, not wanting to abandon children to their fathers prevented them from suicide attempts and completion. The same woman explained:

He had once again beaten me. That is why I wanted to commit suicide …. The old woman in that home stopped me …his grandmother…. It’s that old woman who cautioned me against committing suicide and leaving my children alone. (Respondent #3)

Thoughts of children’s wellbeing surfaced continually and kept women from separating from abusive partners or attempting suicide. At the same time, some women transferred their anger to their children (e.g., husbands battered wives and some wives beat children).

Emotional distress and physical illness.

The combination of many intrusive thoughts, anger, and heart pains contributed to participants feeling physically ill. In some cases, the term “sickness” described their poor physical health. In most cases, however, as can be seen in the previous narrative excerpts, women blurred the lines between physical and mental illness: “I find that I get a disturbed mind, I feel like am falling sick, stomach becomes hot, and my heart also gets heavy” (Respondent #9). Another woman described several physical symptoms that she perceived as connected to IPV:

And also this thing of mine [touches growth on face], when the problems get many, bending becomes difficult. It’s even growing bigger. It makes my head to pain and gives me chills, even if I am stressed over a few things. You feel a lot of coldness. (Respondent #2)

A third participant was diagnosed with ulcers and described how medical personnel confirmed the connection between emotional distress and physical illness: “I experienced ill health due to what I experienced. Yes, I went for a scan and they [doctors] confirmed I had developed wounds in my chest. They told me it was bitterness or anger” (Respondent #11). In addition to stomachaches, headaches, ulcers, and changes in temperature, other symptoms included insomnia, weakness, fainting, fatigue, and low energy. Some women explained how emotional distress and physical symptoms weakened their systems, leaving them vulnerable to other common illnesses such as malaria. One woman shared, “They [disturbed thoughts] have made me lose weight; I also easily fall sick like malaria and also faint” (Respondent #8).

Having many disturbing thoughts was connected to appetite and weight loss problems. A woman expressed, “Because I think a lot all the time. I have too many thoughts that disturb me. My life is not okay. You lose weight because of thoughts” (Respondent #1). Another participant similarly shared, “I have lacked appetite due to many issues in my life. I also easily fall sick due to a lot of thoughts” (Respondent #9).

Socioeconomic Problems and Mental Distress

Sociodemographic covariates were statistically significant predictors of depressive symptoms. Confirming Hypothesis 1c, using linear regression, having more resources was associated with lower scores of depressive symptoms (β = −0.67, t = −5.78, p < .001). Conversely, older individuals tended to have higher scores of depressive symptoms (β = 0.16, t = 4.91, p < .001). However, literate women (β = −0.05, t = −0.10, p = .9) and participants from District 2 (β = 0.67, t = 1.03, p = .31) and District 3 (β = −0.24, t = −0.32, p = .75) did not significantly differ from the reference group of women unable to read and write and participants from District 1 (β = 27.94; see Table 2). Also, Hypothesis 1d was not supported: Economic exposure to armed conflict was not significantly associated with depressive symptoms, χ²(1) = 0.04, p = .85.

When identifying problems related to armed conflict and IPV through rapid ethnographies, women emphasized socioeconomic problems far more than other experiences (Table S1; Figure 1). Exposure to armed conflict led to multiple types of property losses (homes, belongings, agricultural produce), displacement from homes and communities, and reduced access to education. Women described numerous problems associated with IPV, giving examples of husbands controlling resources and denying women and children access to land and other agricultural necessities (e.g., access to cattle for plowing), as well as access to money for food, basic necessities, medical care, and education.

Individual interviews provided further understanding of the connections among socioeconomic problems (associated with IPV) and mental distress. A woman succinctly stated, “When my husband doesn’t buy me anything, I start having lots of thoughts” (Respondent #10). A second participant expressed, “To me these problems have only brought negative effects to my family. It has created a very big economic downfall because we no longer share plans in our household as most of the time quarrels and fights” (Respondent #14). IPV affected women’s economic productivity and general family wellbeing through physical and mental illness:

These emotional problems have impacted my life negatively, though. It has caused poor feeding in my family, as you fail to farm because of stress. Your source of income also gets affected because you’ll always fail to work due to these challenges. You’ll never develop your family with all [the] nagging problems like the ones I listed. (Respondent #11)

Those emotional problems have caused delayed agriculture, as most times being wasted on overthinking. Domestic activities are affected, too, because there is no shared agenda. Because of the heart pains, it influences sickness like ulcers. (Respondent #1)

Much of women’s disturbed thought content included socioeconomic concerns such as survival needs and the wellbeing of their children. Poverty and the financial burden of family needs and children worsened mental health symptoms:

It makes me feel uneasy and unsettled and mostly it happens when he fails to provide basic necessities for my family. I start overthinking of my children as to where I would get support to raise them up. (Respondent #12)

It [lots of thoughts] disturbs me in that it pains me, also ill health of the children, school dues and feeding. It disturbs me because I now frequently fall sick. Like in a year, I have fallen sick I don’t know how many times. And when you fall sick, it’s not a minor one. It’s always major or serious. (Respondent #11)

Discussion

To our knowledge, this is the first population-based, transnational, and mixed-methods (using three methods) study to examine rural, Ugandan women’s individual-level mental health experiences in relation to community- and relational-level exposure to interpersonal violence. Thirty percent of women met the cut-off value for probable Major Depressive Disorder, which is higher than other studies in Uganda that have found prevalence rates of 17.4% (Ovuga, Boardman, & Wasserman, 2005) and 24.4% (Bolton et al., 2004). Table 3 provides a summary of key findings and shows how the quantitative, ethnographic, and interview data converge and diverge.

Table 3.

Convergence and Divergence of Key Mixed Methods Findings.

Mixed Method Analysis	Quantitative Surveys	Rapid Ethnography	In-Depth Interviews
Convergence	Exposure to IPV associated with depressive symptoms	Anger and having lots of thoughts identified as high priority problems and as associated with IPV	Anger, heart pain, and having lots of thoughts constituted the core emotional experience of IPV
	Lower resources were predictive of depressive symptoms	Socioeconomic victimization and problems most frequently discussed and designated as high priority	Having lots of thoughts/disturbed mind was a core emotional outcome and often concerned socioeconomic worries (provision of basic needs and children’s health and wellbeing)
Divergence	30% of respondents met criteria for probable major depressive disorder	Emotional outcomes infrequently identified	Women talked at length about emotional experiences when prompted
	Exposure to armed conflict associated with depressive symptoms; significant differences observed between women who had no exposure to interpersonal violence and women who had exposure to armed conflict violence only	Women associated only one symptom (feeling restless) with exposure to armed conflict and allotted the symptom as low priority	While not explicitly probed for armed conflict and mental health experiences, women did not spontaneously discuss this problem in in-depth interviews
	Economic exposure to armed conflict was not predictive of depressive symptoms	Socioeconomic problems related to armed conflict deemed most problematic	The content of having lots of thoughts/ disturbed mind did not include armed conflict experiences

Note. IPV = intimate partner violence; HSCL = Hopkins Symptom Checklist.

Both quantitative and qualitative data converged in finding that IPV was negatively related to women’s mental health. IPV was significantly associated with depressive symptoms. Women described a core emotional experience of having lots of negative thoughts, heart pain, and anger. Researchers have also reported that participants in other contexts described having lots of (negative) thoughts. For example, in Zimbabwe, kufungisisa is the experience of having a lot of disturbing thoughts, and results from, and contributes to, other mental and physical problems (Abas & Broadhead, 1997). Women’s anger, heart pain, and having lots of thoughts, and limited discussion and lower prioritization of fear, affirm conceptual variance of symptomatology associated with exposure to IPV across settings. Other studies done in sub-Saharan Africa have found that, while many depressive symptoms overlap with those described in the Diagnostic and Statistical Manual of Mental Disorders, there are variations in expression such as increased prominence of somatic symptoms and social isolation and withdrawal (i.e., over having a depressed mood; Kaaya, Lee, Mbwambo, Smith-Fawzi, & Leshabari, 2008).

The negative effects associated with socioeconomic problems on women’s mental health was another area where data converged. Women prioritized socioeconomic neglect and decline as significant problems resulting from IPV and armed conflict, both of which were intricately linked to women’s gendered position. Although economic exposure to armed conflict was not predictive of depressive symptoms, the number of material resources was negatively associated with depressive symptoms. In another study conducted in Uganda, Kinyanda and colleagues (2011) found that socioeconomic status and other sociodemographic variables were some of the strongest correlates with probable major depressive disorder, surpassing the strength of association between experiencing adverse life events and probable major depressive disorder.

A notable area of data divergence was the relation between exposure to armed conflict and mental health. Although our quantitative findings suggested there may be a significant association between armed conflict and depressive symptoms, women did not associate armed conflict with adverse mental health outcomes. Cross-sectional research with women in conflict-affected Cote d’Ivoire similarly found that neither personal victimization, family victimization, nor displacement due to conflict increased odds of exhibiting symptoms consistent with probable PTSD. Experiencing past-year IPV, however, increased odds of probable PTSD 3-fold (Gupta et al., 2014). IPV, as a chronic, day-to-day experience where women live with the perpetrator, is qualitatively different from exposure to violence, often perpetrated by strangers, during armed conflict (Gupta et al., 2014). Our research suggests that the socioeconomic hardship of living with male partners who perpetrate IPV may be so severe that women and their children lack basic needs, and their survival is at stake. Participants’ connections between mental distress and IPV are not surprising, given threats to their basic survival needs (Maslow, 1943). More qualitative research is needed to understand mechanisms between different forms of interpersonal violence and mental health problems among women and these areas of discrepancy.

Practice Implications

Taken together, the constellation of having lots of negative thoughts, heart pain, and anger are informative constructs that can be useful as locally contextualized screening items. In Nepal, utilizing a local concept of distress (mind-heart problems) as screening questions successfully reduced the need for further assessment efforts by half (Kohrt, Luitel, Acharya, & Jordans, 2016). Given our finding that women reported that the constellation of symptoms instigated escape ideation, clinicians who provide primary care and antenatal programs, among others, should assess suicidal risk, in addition to local conceptions of mental distress and IPV. Healthcare settings may be especially relevant entry points for women due to the many connections women made between physical and emotional outcomes.

Women’s in-depth interview narratives suggested a bidirectional relation between mental health and socioeconomic circumstances, consistent with other research (Lund et al., 2010), indicating that interventions targeting socioeconomic variables may reduce mental distress and vice-versa. In the Democratic Republic of Congo (DRC), an asset transfer program called Pigs for Peace delivered significant reductions in PTSD and anxiety symptoms. Nevertheless, while IPV decreased, differences in IPV between the intervention and delayed control groups were not significant. Also in the DRC, cognitive processing therapy with sexual assault survivors enhanced social support seeking but did not significantly improve financial networking (Hall et al., 2014). An economic empowerment intervention in the Côte d’Ivoire, combined with a gender discussion group, reduced PTSD symptoms but was less effective in reducing the PTSD symptoms of women who had experienced IPV prior to treatment (Annan, Falb, Kpebo, Hossain, & Gupta, 2017). Taken together, these innovative studies show that more research is needed to continue developing multifaceted interventions that can improve women’s socioeconomic standing and mental health and reduce IPV.

In low-income settings with challenges of poor infrastructure and low resources, the utilization of non-specialized personnel to deliver mental health and IPV interventions is essential. For instance, building a participatory peer-response network might be one approach for preventing suicide and providing support for IPV survivors with depressive symptoms. The utilization of non-specialized personnel has been successful in other low-income settings. In South Africa, community health workers (“maternal mothers”) have effectively facilitated an antenatal and postnatal intervention, which included the provision of information on maternal and child health, infectious diseases, nutrition, and alcohol use. Rotheram-Borus, Tomlinson, Roux, and Stein (2015) found surprising reductions in depressive symptoms, even without mental health being directly addressed by the intervention. In community settings in Kenya, community health workers have screened and assessed women in households for gender-based violence and mental health problems; they facilitated a brief, five-session intervention that, while ineffective in reducing exposure to violence, was successful in reducing mental distress (Bryant et al., 2017). Although studies have begun examining effectiveness of mental health treatment in other types of settings, more is needed.

Regarding implementation of mental health programming in low-income settings, our results show that survivors of interpersonal violence are embedded in complex cultural systems in which gender roles and structures are paramount. Within mental health dissemination and implementation frameworks, the organizational culture of the treatment provider is routinely accounted for (Tabak, Khoong, Chambers, & Brownson, 2012). Researchers have included little to no mention of the role of gender in implementation science (the science of translating research-to-practice) in texts, research articles, or systematic reviews (Tannenbaum, Greaves, & Graham, 2016). Enhanced methods that can account for the gendered and cyclical interaction of poverty, exposure to violence, and mental health outcomes can reveal implementation strategies that are ecologically embedded, thus improving feasibility and sustainability. For example, recent strategies have looked to the integration of systems science modeling—computer modeling showing how feedback mechanisms interact over time to regulate dynamic behaviors of a system (Hovmand, 2014)— into implementation efforts, as one such approach to address complexity (Northridge & Metcalf, 2016).

Our results invite further developments in policy. The rapid ethnographic findings are striking, in that we found nearly complete overlap between problems associated with exposure to armed conflict and IPV for women, serving as a poignant reminder for policy makers to legislate for rights of women in domestic settings, just as they would for rights of women in humanitarian crises. Thus, advocacy efforts must consider the importance of gender role socialization and cultural norms around IPV. Moreover, mental health policymaking efforts would benefit from partnerships with gender and development experts to design prevention and response strategies that are multi-pronged and systemically informed.

Limitations

Tests of group differences demonstrated a moderate effect size, signifying an added benefit of converging quantitative analyses with qualitative data to increase understanding of the intersecting roles of exposure to violence at the relational and community levels. Yet, readers should interpret the findings of this study within the limitations of its methodology. First, the quantitative data are cross-sectional, preventing direct interpretations about causality. Second, while the depressive scale of the HSCL-25 has been validated in Uganda, the validation site was urban and differed from the one under study, so perhaps there are sociolinguistic differences in the current study’s sample that affected validity in ways that are unaccounted for (Kinyanda et al., 2011). Next, although we translated, back translated, and modified items for cultural equivalence, these methods do not constitute measure equivalence. Third, inclusion criteria in the qualitative arm consisted of exposure to IPV so, while many women had dually experienced both types of violence, participant selection may have shaped the findings. Finally, stemming from women’s assessment in the rapid ethnography that the psychological outcomes from IPV had higher priority, the in-depth interview questions sacrificed breadth for depth and focused more intentionally on IPV and associated mental distress.

Conclusions

While Ugandan women who endured ongoing exposure to armed conflict and intimate partner violence were more likely to have developed depressive symptoms, they prioritized socioeconomic abuse and struggles as most problematic in affecting their mental health. Researchers, health care professionals, and policy makers can leverage local conceptions of distress for screening and adaptation of treatment. Policy efforts should incorporate gender and economic development theory, collaborate with interprofessional and local experts, and consider methods that adequately account for ecological complexity.