Care pathways during a child’s final illness in rural South Africa: Findings from a social autopsy study

Background Half of under-5 deaths in South Africa occur at home, however the reasons remain poorly described and data on the care pathways during fatal childhood illness is limited. This study aimed to better describe care-seeking behavior in fatal childhood illness and to assess barriers to healthcare and modifiable factors that contribute to under-5 deaths in rural South Africa. Methods We conducted a social autopsy study on all under-5 deaths in two rural South African health and demographic surveillance system sites. Descriptive analyses based on the Pathways to Survival Framework were used to characterise how caregivers move through the stages of seeking and providing care for children during their final illness and to identify modifiable factors that contributed to death. Findings Of 53 deaths, 40% occurred outside health facilities. Rates of antenatal and perinatal preventative care-seeking were high: over 70% of mothers had tested for HIV, 93% received professional assistance during delivery and 79% of children were reportedly immunised appropriately for age. Of the 48 deaths tracked through the stages of the Pathways to Survival Framework, 10% died suddenly without any care, 23% received home care of whom 80% had signs of severe or possibly severe illness, and 85% sought or attempted to seek formal care outside the home. Although half of all children left the first facility alive, only 27% were referred for further care. Conclusions Modifiable factors for preventing deaths during a child’s final illness occur both inside and outside the home. The most important modifiable factors occurring inside the home relate to caregivers’ recognition of illness and appreciation of urgency in response to the severity of the child’s symptoms and signs. Outside the home, modifiable factors relate to inadequate referral and follow-up by health professionals. Further research should focus on identifying and overcoming barriers to referral. This was a population-based, cross-sectional verbal and social autopsy (VASA) study, conducted in two health and demographic surveillance system (HDSS) sites in rural South Africa, the Agincourt HDSS and the Africa Health Research Institute (AHRI). Both sites are members of the INDEPTH Network (www.indepth-network.org) and are two of the pioneer nodes of the South African Population Research Infrastructure Network (SAPRIN) (http://saprin.mrc.ac.za). The Agincourt and AHRI HDSS sites are situated in poor rural areas of South Africa, with limited infrastructure. Together these sites cover more than 280 000 people in over 40 500 households [15]. Both sites have high levels of temporary labour migration (33–36%) with household members oscillating between their place of work and their rural home [16–18]. The majority of households are dependent on income from social grants (particularly government pensions for older adults and child grants). Mozambican immigrants account for a third of the population under surveillance in Agincourt. In Agincourt, healthcare is provided by seven primary care nurse-led clinics, 10 private general practitioners (GPs) and two health centres. Only the two health centres remain open 24 hours and include emergency services. Three district hospitals are accessible 25-60km away [17]. In AHRI, healthcare access is via 10 clinics, six private GPs all located in Mtubatuba, and one health centre. Only the health centre is open 24 hours. The nearest district hospital is 45 km away, on the other side of a nature reserve where wild animals roam free. There are also approximately 300 traditional healers operating in and around each study site [19]. We included all under-5 deaths identified during the 2017 household surveys across the two HDSS sites in this study. The Agincourt HDSS has conducted annual household surveys of key demographic and health data since 1992, while AHRI’s population health surveillance started in 2000 and now involves four-monthly monitoring. The household surveys include identifying any new pregnancies, births and in-migrations, as well as all deaths and out-migrations. Data on household and maternal characteristics were taken from the household surveys. Socioeconomic status was determined based on household asset ownership: households were divided into quintiles from 1 (poorest) to 5 (least poor). In Agincourt an absolute asset index is used to rank households as described by Kabudula et al [20,21], before assigning the household to a socioeconomic quintile. In AHRI, a wealth index is derived using principal component analysis as described by Nyirenda et al [22]. Households are assigned a wealth score which follow a standard normal distribution, and subsequently divided into quintiles. The socioeconomic quintile of each household was determined relative to all other households in their HDSS in the year that the child died. All deaths of members of the HDSS sites that are identified during the household survey are followed up and investigated using verbal autopsies (VA)–a structured interview with the caregiver of the deceased to determine biological cause of death. Verbal autopsies have been validated as a means of establishing cause of death in a rural South African population [23]. Both sites use the standardized World Health Organization VA tool, which has included 10 questions on circumstances of death since 2012 [24]. We added a locally-relevant adaptation of the INDEPTH Network Social Autopsy tools for neonatal and child deaths [5] to the WHO 2016 VA tools for neonatal and child deaths. The social autopsy questions focus on the specific actions taken by caregivers during the child’s final illness and broadly follow the stages of the care pathway: 1) identifying symptoms, 2) providing care inside the home, 3) seeking healthcare outside the home 4) determining they were referred for further care and 5) whether they accepted that referral. In addition, the social autopsy attempts to identify barriers faced in accessing healthcare at each stage of the care-seeking process (see S1 Table for the adapted social autopsy tool that was integrated into the WHO 2016 VA). Interviewers were trained in the use of the verbal and social autopsy tools, which were translated into isiZulu and Shangaan, the local languages in the two HDSS sites, and back-translated into English to ensure accuracy. The verbal and social autopsy interviews were conducted between July 2017 and February 2018 for all deaths in children younger than five years identified across the HDSS sites in the 2017 household surveys. Interviews were conducted at least one month after the death to allow for the customary mourning period to pass, and up to 18 months after the death where special arrangements were required to ensure the VA interviewers were able to interview the primary caregiver of the deceased who would be the most appropriate respondent. Descriptive data analysis was conducted using the BASICS/CDC Pathways to Survival Framework [25]. Data was pooled across the two sites to give a more representative picture of rural South Africa. However, where any differences were found, we have highlighted results for each site. Neonates that were born and died in health facilities without discharge were excluded from analysis. The sociodemographic characteristics of participants and households were reported to provide relevant context for the barriers to access and modifiable factors identified during the social autopsy interviews, however given the small number of deaths in 2017, we did not perform bivariate or multivariate analyses of associations between sociodemographic characteristics and care-seeking patterns. This study was approved by the University of Oxford Medical Sciences Inter-divisional Research Ethics Committee ({“type”:”entrez-nucleotide”,”attrs”:{“text”:”R52414″,”term_id”:”814316″}}R52414/RE001), the University of Witwatersrand Human Research Ethics Committee (M1705102) and the University of KwaZulu Natal Biomedical Research Ethics Committee (BE 290/16). Written consent was obtained from the household head for participation in the household survey. The verbal autopsy interviewers took written consent separately from primary caregiver of the deceased to participate in the verbal and social autopsy interview.