Vulnerability and agency across treatment-seeking journeys for acutely ill children: How family members navigate complex healthcare before, during and after hospitalisation in a rural Kenyan setting

Background: Child mortality rates during hospitalisation for acute illness and after discharge are unacceptably high in many under-resourced settings. Childhood vulnerability to recurrent illness, and death, is linked to their families’ situations and ability to make choices and act (their agency). We examined vulnerability and agency across treatment-seeking journeys for acutely ill children and considered the implications for policy and practice. Method: A qualitative sub-study was embedded within the prospective CHAIN Network cohort study, which is investigating mechanisms of inpatient and post-hospital discharge mortality among acutely ill young children across a spectrum of nutritional status. Primary data were collected from household members of 20 purposively selected cohort children over 18 months through formal interviews (total n = 74), complemented by informal discussions and observations. Data were analysed using narrative and thematic approaches. Results: Treatment-seeking pathways were often long and complex, particularly for children diagnosed as severely malnourished. Family members’ stories reveal that children’s carers, usually mothers, navigate diverse challenges related to intersecting vulnerabilities at individual, household and facility levels. Specific challenges include the costs of treatment-seeking, confusing and conflicting messaging on appropriate care and nutrition, and poor continuity of care. Strong power inequities were observed between family members and health staff, with many mothers feeling blamed for their child’s condition. Caregivers’ agency, as demonstrated in decision-making and actions, often drew on the social support of others but was significantly constrained by their situation and broader structural drivers. Conclusion: To support children’s care and recovery, health systems must be more responsive to the needs of families facing multiple and interacting vulnerabilities. Reducing incurred treatment costs, improving interpersonal quality of care, and strengthening continuity of care across facilities is essential. Promising interventions need to be co-designed with community representatives and health providers and carefully tested for unintended negative consequences and potential for sustainable scale-up. CHAIN (www.chainnetwork.org) is a multi-disciplinary research network aimed at understanding the mechanisms contributing to high mortality in hospital and after discharge in LMICs in order to identify interventions to improve survival [5]. The Network is leading a cohort study at nine hospital sites in Africa and South Asia, recruiting more than 3000 acutely-ill children at admission to hospital and following them up post-discharge. Scheduled follow up visits are conducted at days 45, 90 and 180 after hospital. Children are enrolled and classified in three strata by anthropometric status since this is a strong marker of survival risk that encompasses both biological and social risks: severe wasting or kwashiorkor (oedematous malnutrition) (SWK), moderate wasting (MW) and no wasting (NW). Enrolled children faced varied social risks in terms of social disruption, household locations and types, and levels of maternal education [5]. Audit and training were provided to sites to ensure treatment and referral for outpatient nutritional and other care after discharge was according to current national and WHO guidelines. The primary outcome of the cohort study is mortality. Kilifi, Kenya is one of four CHAIN sites in which in-depth qualitative work was conducted to better understand family, community and stakeholders’ perspectives on children’s illness trajectories. This qualitative research was conducted in collaboration with Resilience, Empowerment and Advocacy for Women’s and Children’s Health Research (REACH), a collaborative research ethics project aimed at better understanding the practical ethical challenges of research involving populations considered to be vulnerable. Pooled data from all four sites will be reported separately. Here we consider, in-depth, the findings from Kilifi, one of Kenya’s poorest counties. The county’s health system and population face significant structural vulnerabilities, including 68% of the population living below the poverty line [24], a shortage of health workers, regular drug stock-outs and health facility access challenges [25, 26]. Most of the population depend on small scale farming, and high levels of gender inequity have also been documented [23, 27]. The latter power relations have been shown to have important implications for children’s treatment-seeking [9]. Qualitative work in Kilifi included interviews with family members of 20 purposively selected children from the site cohort of approximately 500. Children were selected to maximise the diversity of experience, based on nutritional status, household socio-economic status, geographical location, and any recent exposure to a socially disruptive event such as death of a caregiver. Members of families who consented were visited in their homes at least three times over an 18-month post-discharge period, with most visits lasting one to three hours. A total of 74 interviews were held with the children’s primary caregivers and other family members (mostly mothers, fathers, mothers-in-law, aunts, uncles and fathers in law) between April 2017 and July 2018. Most household visits and interviews were conducted by two social science team members, with formal interviews organised to fit around domestic activities to minimise disruption to family life. Interviews covered a broad range of topics informed by the theoretical and empirical literature summarised in Fig. 1, which served as a conceptual framework. This conceptual framework incorporates ideas from the treatment-seeking literature [9, 12, 20] with Rogers et al. [16] and Lange et al.’s [17] vulnerability frameworks. For vulnerabilities, we were particularly interested in situational, more immediate, intrapersonal (biological or psychological), inter-personal (roles, relationships and interactions), and environmental (socio-economic and cultural, and institutional) vulnerabilities. We recognised that these vulnerabilities, and agency, across household/community and health service levels, would be shaped by broader social, economic and political forces and power-relations, or structural drivers. Quantitative data on measurable elements of vulnerability across all CHAIN sites will be presented in a future publication. Potential influences on treatment-seeking and outcomes Interview guides were used flexibly to support a rich and relaxed discussion in either Kiswahili or the local dialect, Kigiriama by native speakers conducting interviews. To support data quality, interview guides were translated, piloted and revised by the research team prior to data collection. Discussions focused on experiences and decision making at each stage of the illness, any challenges encountered, coping strategies, and how these strategies were perceived to work. The overall aim was to build the story of the illness and associated treatment-seeking from the perspective of parents and other caregivers and to understand the types of vulnerabilities to (re) hospitalisation, recurring illness or death, and agency revealed through these narratives. All formal interviews were audio-recorded, transcribed verbatim and later translated into English. All notes from the field and post-interview debrief were typed up and reviewed for issues to follow up in later interviews. We worked as a team to adopt two complementary approaches to analyse the data: a narrative approach [28, 29] and a thematic coding approach [30]. The narrative approach involved the construction of a detailed overall summary for each household, drawing on all available data, and a shorter narrative, or story. We worked with these summaries and stories to explore the overall picture of households’ pathways through care, examining household/community and health service/system influences pre- and post-admission. We investigated changes over time, and patterns of similarity and difference across households, facilitated by the construction of charts (see Table 1 for an excerpt). The thematic coding supplemented and enriched the narrative analysis. All transcriptions were coded in NVivo 10 using a coding framework based on our initial and emerging themes of interest, including treatment-seeking patterns and influences on those patterns. To support the trustworthiness of the coding process, at least two people coded each transcript, comparing results and resolving any discrepancies. Example of Household Charts comparing themes Hh3 Male, with SWK, 21 months old & unknown birth weight. About 8 months: Child occasionally gets convulsions, but the cause not clearly understood. PRE Health centre-Health centre-Private clinic-Private clinic-(both retirees living in the village) -Public dispensary-duka POST Supp Public Dispensary-Private clinic-illness continues (child still not well), duka, Public dispensary, Private clinic. Used to walk long distances to seek care, so as to reduce costs. Missed meals or reduced intake to help cover expenses for the child during treatment seeking. Siblings stopped schooling during the child’s admission. Post Could not sustain providing nourishing food as prescribed at discharge. Received support from relatives, neighbours and friends in different forms: advice, loans or foodstuff. Neighbours convinced the child’s father to accept biomedical care and send funds for the same. Took long to diagnose the problem despite several visits to local health practitioners and health facilities. Some levels of mistrust (local hws) regarding post treatment therapy. Couldn’t access care when needed during a health worker strike. Sometimes had to self-medicate due to regular drug stock-outs at local facility. Ensure child fed on nutritious food: fruits, high protein content foods-eggs, milk though couldn’t sustain. Also, asked to observe and maintain hygiene around the child- limited water sources around her area. We drew on both the narrative and coded data to identity forms of vulnerability of (re) admission, prolonged illness or death (intrapersonal, interpersonal, environmental and structural), and agency, observed at household/community levels and in health service interactions. CHAIN protocols were approved for science and ethics in all participating countries. Collaboration with REACH supported us to embed ethics research within the CHAIN cohort study. The REACH Principle Investigator (MK) is the CHAIN ethics advisor who provided pre-study and on-going ethics advice and guidance. In Kenya, the research was reviewed and approved by the Kenya Medical Research Institute (KEMRI) Scientific and Ethics Review Unit (KEMRI/SERU/3318/054). Approval was also sought and received from the Oxford Tropical Research Ethics Committee (OxTREC number 34–16). Working with families longitudinally can be ethically challenging given the forms of socio-economic and biological vulnerabilities anticipated. Still, this approach was valuable in generating the type of data needed to develop appropriate interventions. Verbal consent was sought from participants during admission to provide initial information and ask permission to visit homesteads. Written informed consent was then sought from participants on the first household visit for all in-depth interviews, observations and recordings. Consent was checked in subsequent household visits. As a form of appreciation and to compensate for time disruptions during our home visits, each household was provided with a food package after each visit. This was based on past experience [31] and was in line with the institution’s benefits guidelines. Where a child was observed to require medical attention during a research interaction, caregivers were referred to local facilities or CHAIN clinicians. In debrief meetings, ethical dilemmas encountered by frontline staff were raised and discussed, and where considered appropriate, acted upon.