Study setting
Guinea is located in West Africa and had a population of 10·5 million in 2014. Most of Guinea’s residents are illiterate (67%), live in a rural setting (71%) and subsist below the poverty line (55%) (12,19). The country has thirty-three districts of which twenty-five were affected by the EVD outbreak. The national health system is tiered in primary, secondary, and tertiary levels (20). At the primary level, community healthcare workers (CHWs) provide healthcare and prevention services within communities (21).
The study was conducted in the district of Guéckédou, south-eastern Guinea (Figure 1). Guéckédou was the epicentre of the EVD outbreak (22) and recorded the highest EVD mortality rate countrywide, with 204 deaths out of 270 confirmed cases (76%) (2). It was also among the localities that experienced more community resistance to EVD response activities (4,8,23). Furthermore, it belongs to the most malaria-affected natural region in the country where the prevalence of malaria among children under five years of age is 61% (12). Guéckédou district consists of ten sub-districts and an urban commune.
The study sites specifically included the sub-district of Guèndembou, which had the highest EVD case burden in Guéckédou (60 reported cases), and the sub-district of Bolodou, which was less affected by the EVD outbreak (1 reported case) (2). At the time of the study (post-EVD period) Guèndembou had one health centre, one private clinic, and six health posts for target under-five population of 6,281 people. In Bolodou, there were one health centre and four health posts, for a target under-five population of 2,729 people. The number of health facilities available and functional in the post-EVD period was similar to the number in the pre-EVD context [source: Guéckédou Health District Office, 2017]
Operational definition of health-seeking behaviour and conceptual framework
We defined health-seeking behaviour as a “sequence of remedial actions that individuals undertake to rectify perceived ill-health” (24). In this study, we focused on caregivers’ decisions regarding the type of healthcare provider patients sought help from, reasons for choice of healthcare professional, and reasons for not seeking help from healthcare professionals (24).
Selection of our study variables was based on a conceptual framework produced by adapting Metta’s model and the Partners for Applied Social Sciences (PASS) model, as well as accounting for authors (BSC and AD)’s community experience of the EVD outbreak (Figure 2) (25,26). It draws on the pattern of post-EVD outbreak health-seeking behaviours as interplay among three main aspects: the perception of the post-EVD health system, post-EVD socio-economic status, and illness interpretation. The perception of post-EVD health system conditions post-EVD care-seeking behaviour is informed by i) messages and rumours at the community level about the EVD outbreak and its management by the health system; ii) personal, household, or community experience of the EVD outbreak; and iii) personal experience with health services. These factors could shape individuals’ perceived barriers or benefits (e.g., risk of EVD contamination at facility; availability of services; belief in care providers; their attitudes; quality and cost of care) and make them seek care through a given health service channel. Post-EVD socioeconomic status can be affected by the personal or household experience of the EVD outbreak (e.g., if the EVD death concerns a productive or supportive member of the household). Illness interpretation also guides individuals’ decision to resort to a given health service channel. It depends mainly on individuals’ knowledge of the illness, their perceived severity, and their perceived susceptibility to being at risk of the illness.
Study design
This was a cross-sectional survey using interview data, with closed ended questions. The survey was part of larger mixed-method study, with the qualitative component expected to be presented elsewhere.
Study participants and sampling
Caregivers of children under five years of age were interviewed. We considered as caregiver the child’s mother or the main person caring for him or her at home. Caregivers were selected through two-stage cluster sampling. In the first stage, the sub-district with the highest reported EVD case burden (Guèndembou; 60 cases) in Guéckédou and a sub-district with a low reported EVD case burden (Bolodou; one case) were selected. For the second stage, in each selected sub-district, all households with a child aged less than five years who had a fever episode (as reported by the caregiver) within the preceding 30 days were included in the study. The caregiver of one eligible child per household was selected. In households with more than one eligible child, the child who had the most recent episode was selected. All households were visited with the help of local guides, moving from the middle to the ends of each village, clockwise. The visiting process proceeded from the main village to the surrounding villages of each sub-district, clockwise, until the desired sample size was reached. In total, fourteen villages (seven in each sub-district) were visited.
The sample size was calculated using Cochran’s sample size formula, which is appropriate for cross-sectional studies (27). It based on the proportion of febrile children for whom caregivers sought health services in Guinea in 2012 (37%) (12), a confidence level of 95% and a margin error of 5%. A minimum of 358 caregivers were needed for the study. Half of the study participants were expected to come from each sub-district, to account for actual situation in each of the two sub-district.
Data collection and variables
Data were collected from 24 September to 4 October 2017 by trained interviewers using Open Data Kit (ODK) with Android mobile phones.
The study variables included: sociodemographic characteristics of caregivers and their children (caregiver’s age, education level, marital status, number of household members, main source of household income, household characteristics, age of the child, and gender of the child); EVD-related events and feelings (occurrence of EVD deaths in the household/family, whether EVD impoverished the household/family, fear of shaking hands with friends, fear of hugging friends, fear of sharing plates with friends, fear of hugging household/family members, fear of sharing plates with household/family members, fear of kissing household/family members, preference for washing hands with chlorine solution, keeping chlorine solution at home); interpretation of the child’s illness (the child could eat or breastfeed as usual [yes/no], the child could move as usual [yes/no], meaning of the illness to the caregiver, diagnosis of the illness); health-seeking options and reasons (sought care [yes/no], if yes: health-seeking place, reasons for selecting the health-seeking place); utilization of health services (if sought care, blood test performed [yes/no], medicines given [yes/no]); and perception of service quality at health facilities as compared with pre-EVD (medicine availability, antimalarial drug availability, availability of rapid test kits for malaria, waiting time, staff listening to patients, cases left without treatment, staff reliability, facility cleanliness, cost of care, quality of treatment).
Data analysis
Descriptive variables were presented as proportions or means with standard deviations (SD). Numerical values were assigned to household characteristics by adapting the method developed by the health and demographic survey (12) to measure household property scores. EVD-related feelings were also assigned numerical values to measure EVD outbreak fear among caregivers. Caregivers’ level of fear was as assessed using numerical scores, and the maximum fear score had a value of 10. Pearson’s chi-square (X2) and student t-tests were used to compare the variables between the two sub-districts.
A logistic regression using a backward stepwise model was conducted to predict caregivers’ high fear score (above the median) and care seeking behaviour. Adjusted odds ratios (AOR) were then derived with 95% confidence intervals (CIs). The level of significance was set at p < 0.05.
Data were analysed using SPSS software version 22.0 for Windows (SPSS Inc., Chicago, IL, United States).