Maternal mental health priorities, help-seeking behaviors, and resources in post-conflict settings: A qualitative study in eastern Uganda

Background: Limited knowledge exists to inform the selection and introduction of locally relevant, feasible, and effective mental health interventions in diverse socio-cultural contexts and health systems. We examined stakeholders’ perspectives on mental health-related priorities, help-seeking behaviors, and existing resources to guide the development of a maternal mental health component for integration into non-specialized care in Soroti, eastern Uganda. Methods: We employed rapid ethnographic methods (free listing and ranking; semi-structured interviews; key informant interviews and pile sorting) with community health workers (n = 24), primary health workers (n = 26), perinatal women (n = 24), traditional and religious healers (n = 10), and mental health specialists (n = 9). Interviews were conducted by trained Ateso-speaking interviewers. Two independent teams conducted analyses of interview transcripts following an inductive and thematic approach. Smith’s Salience Index was used for analysis of free listing data. Results: When asked about common reasons for visiting health clinics, the most salient responses were malaria, general postnatal care, and husbands being absent. Amongst the free listed items that were identified as mental health problems, the three highest ranked concerns were adeka na aomisio (sickness of thoughts); ipum (epilepsy), and emalaria (malaria). The terms epilepsy and malaria were used in ways that reflected both biomedical and cultural concepts of distress. Sickness of thoughts appeared to overlap substantially with major depression as described in international classification, and was perceived to be caused by unsupportive husbands, intimate partner violence, chronic poverty, and physical illnesses. Reported help-seeking for sickness of thoughts included turning to family and community members for support and consultation, followed by traditional or religious healers and health centers if the problem persisted. Conclusion: Our findings add to existing literature that describes ‘thinking too much’ idioms as cultural concepts of distress with roots in social adversity. In addition to making feasible and effective treatment available, our findings indicate the importance of prevention strategies that address the social determinants of psychological distress for perinatal women in post-conflict low-resource contexts. Research activities were conducted in Soroti district, a predominantly Ateso speaking area of the Teso sub-region of Uganda (Fig. 1, developed by BK) [26]. Uganda’s health care system formally comprises seven levels, corresponding with government administrative divisions. At the village or Health Center I level (HC I), community health workers are organized into Village Health Teams. Village Health Teams are a voluntary health force that generally do not have a physical office and who mainly engage in community-based preventive, health awareness and basic curative activities. At the parish or HC II level, health centers are staffed by primary health care workers (e.g. general nurses or midwives), who deliver basic health services to catchment areas of approximately 5000 people. At the HC III (sub-county level), health centers have a clinical officer, a maternity unit, and unlike lower levels, separate wards for male and female patients (8 beds). Health centers at levels IV-VII have greater staff capacity and provide increasingly specialized care. In Soroti district, there are 14 HC IIs, 12 HC IIIs, two HC IVs, and one 195-bed regional referral hospital. The Teso sub-region study site in eastern Uganda We estimated that including four of Soroti’s ten sub-counties in our study would give us enough opportunity to identify commonalities and differences between sub-counties and reach data saturation among different participant groups of interest. As the majority of Uganda’s population lives in rural areas where the gap between health needs and available services is greatest, we excluded all sub-counties that did not have a predominantly rural population (n = 4) [24]. For feasibility reasons, we then excluded one sub-county where the majority of residents did not speak Ateso. Of the five remaining sub-counties, we randomly selected four for inclusion. Together, these sub-counties represent a predominantly rural, Ateso-speaking population living at various distances from the regional hub in Soroti municipality. Within each sub-county, we conducted purposive and convenience sampling of stakeholders, including health workers, perinatal women seeking help at health centers, and religious and traditional healers. We aimed to include health workers from different healthcare system levels, with a focus on HCI to HCIII workers as rural women most commonly access antenatal care at these levels. HC IIIs were of particular interest, as these facilities have the highest rates of antenatal care visits. Among sub-counties with multiple HC IIs, HC IIs were randomly selected for study inclusion. In each sub-county, we started recruitment by visiting HC IIs and IIIs during working hours to introduce the study to the clinical officer in-charge of the facility. These clinical officers helped recruit other health workers from their facilities to be interviewed. Community health workers at the HC I level were recruited by liaising with the Village Health Team Sub-County Coordinator. An overview of participants and data collection activities can be found in Table 1. We first conducted free listing and ranking (cf. [27, 28]) embedded in semi-structured interviews with 22 primary and 22 community health care workers from HC Is, IIs, and IIIs in each sub-county. Free listing was conducted individually with the exception of one group exercise with community health care workers (Table ​(Table1).1). Specifically, we asked participants “what are the most common problems for which pregnant women, or women who have just given birth, visit the health center?” Participants were invited to provide as many answers as possible and were probed to include all of the (physical) health, mental health, social and spiritual or supernatural problems that may lead perinatal women to visit the health center. The interviewer wrote down answers verbatim. Among the problems that were listed, interviewers asked the participants which problems they considered to be related to thoughts, feelings or behaviors (i.e. mental health problems) and to rank the three mental health problems that they felt were most important. Participants were then asked about help-seeking and current health center resources for the three most highly prioritized maternal mental health problems. Overview of participants and methods AI Additional Information, SC1 Sub-county 1, as a confidentiality safeguard, sub-counties are numbered We conducted individual semi-structured interviews with perinatal women to gather information on help-seeking behaviors and available services for maternal mental health problems. We aimed to recruit four perinatal women from each included sub-county as they were waiting to receive antenatal care at the HC III. In these interviews, we presented participants with a fictitious case vignette about a woman in Soroti district experiencing mental health problems. The vignette was developed based on descriptions of the most highly prioritized problems identified in the free list interviews with health workers. Participants were asked to describe (1) what the woman in the case vignette could herself do about her problems; (2) their thoughts about where the woman would likely seek help (first, second, third, and so on); (3) to provide details about what services could be obtained at each place where help might be sought; and, (4) what help the health center could offer for this woman and how the health center could improve its assistance. We conducted key informant interviews with four groups of participants: traditional or religious healers, mental health specialists, community and primary health care workers, and perinatal women. Two traditional or religious healers were recruited through consultation with sub-county leaders and health center officials in each sub-county and nine mental health specialists were recruited from the regional and national referral hospitals. Specialists and healers participated in free listing and ranking exercises during their interviews that were similar in structure to the initial free listing interviews completed with primary and community healthcare workers. In this round of free listing interviews, greater focus was placed on eliciting symptoms of identified priority maternal mental health problems, what groups are particularly affected by these problems, and treatments provided for them. We then conducted a second round of key informant interviews with a subset of the community and primary health workers who participated in the free listing interviews and traditional healers who participated in the initial key informant interviews. Health workers and healers were selected if they provided rich information about the mental health needs in their sub-counties. We also conducted key informant interviews with two older (more experienced) perinatal women from each sub-county. These women were recruited by health care workers and selected based on their ability to read and their experience with the antenatal care system, as determined by their age and number of previous births. Because free listing resulted in several terms that seemed closely related, we added a pile sorting activity to these key informant interviews. Pile sorting was intended to increase our understanding of how participants would classify various symptoms. Participants were provided with notecards with names of symptoms (in both Ateso and English) derived from descriptions of the three prioritized maternal mental health problems. Participants were asked to sort the cards into piles that made sense for them, in as many or few groups as they preferred, and to name to each pile. These key informant interviews subsequently focused on different idioms of distress used to describe the three prioritized maternal mental health problems; symptoms belonging to these problems; and groups of women particularly affected by them. Finally, participants were asked about the causes of the top three problems and possible solutions for each of the causes. Perinatal women were also probed about help-seeking behaviors for the prioritized problems. Interviews were conducted using semi-structured interview guides (available as Additional files 1, 2, 3, 4, 5, 6, 7, 8) that contained interview questions, possible probes, and space for notes. They were conducted in Ateso or English, depending on the preferences of the participant. All interviews were audio recorded and transcribed verbatim. When using the guide, interviewers were instructed to still follow the narrative of the participants as much as possible. All interviews were conducted by locally recruited, Ateso-speaking interviewers who participated in a two-week training on research ethics and qualitative interviewing skills. The training provided ample opportunity for skill-based learning through feedback on role-plays, with a strong focus on asking open questions, probing, and giving non-judgmental responses. All participants provided written informed consent before participating in interviews. Ethical approval for this study was granted by the Johns Hopkins Bloomberg School of Public Health Institutional Review Board, the Mildmay Uganda Research Ethics Committee, and the Uganda National Council for Science and Technology. In addition, we obtained approval from the District Health Officer before starting recruitment. To analyze free list data and rank priority maternal mental health problems, we used Smith’s Salience Index (S), which ranks responses based on frequency of mention while also weighting items based on where it appeared on participants’ lists. Under this method, frequently mentioned items among individuals are considered to indicate common knowledge or consensus, and items listed earlier are assumed to be more important to the participant. A higher S score is indicative of greater saliency [29]. Smith Salience index is calculated for each item as the total number of items in the list minus the rank order for item A divided by the total number of items in the list. Scores range from 0 to approximately 1, with scores closer to 1 indicating greater saliency. Transcripts were analyzed using an emergent, inductive coding approach [30, 31]. This stage of the analysis was performed in parallel by two teams of analysts: one of postgraduate students based at the Johns Hopkins Bloomberg School of Public Health (JHBSPH) in the US; the other was comprised of one postgraduate student from JHBSPH and a Ugandan community psychology graduate based in Uganda. The US-based team began analysis by conducting an initial transcript review, in which they highlighted and summarized portions of text relevant to the research questions on a sub-set of interviews. Out of these summaries, they developed emergent codes and organized the codes into a framework or codebook. This codebook was utilized to code the interviews using Dedoose software [32]. The Ugandan team manually conducted line-by-line coding for the first three interviews and prepared a list of initial codes. These codes were combined and re-organized into focused codes and the revised codebook consisted of five topical and 41 response codes. New themes emerging from the data were discussed between the researchers and incorporated into the developing codebook as deemed appropriate (available as a Additional file 8). In the last step, the Uganda team coded all interviews using this final codebook, through the aid of the R package for Qualitative Data Analysis software [33].