Objectives To understand the recent rise in facility deliveries in Ethiopia. Design A qualitative study. Setting Four rural communities in two regions of Ethiopia. Participants 12 narrative, 12 in-depth interviews and four focus group discussions with recently delivered women; and four focus group discussions with each of grandmothers, fathers and community health workers. Results We found that several interwoven factors led to the increase in facility deliveries, and that respondents reported that the importance of these factors varied over time. The initial catalysts were a saturation of messages around facility delivery, improved accessibility of facilities, the prohibition of traditional birth attendants, and elders having less influence on deciding the place of delivery. Once women started to deliver in facilities, the drivers of the behaviour changed as women had positive experiences. As more women began delivering in facilities, families shared positive experiences of the facilities, leading to others deciding to deliver in a facility. Conclusion Our findings highlight the need to employ strategies that act at multiple levels, and that both push and pull families to health facilities.
Data were collected between March and May 2015, from two wards (kebeles), the smallest unit of local government, in the Southern Nations, Nationalities and Peoples (SNNP) region and two in Amhara region. Amhara has shown an increase in facility deliveries from 11% (2011) to 35% (2016), and SNNP region from 7% to 33%.5 Data were collected from areas where ‘The Last Ten Kilometers’ (L10K) programme was active in supporting the Health Extension Program. Kebeles were selected from a list, provided by L10K project staff, of kebeles considered to have a reasonably functioning HEW system, that is that they had HEWs in place that were considered to be active and working well. Other selection criteria were that the kebeles were seen as typical of the district (woreda) with no unusual characteristics such as having a large hospital or a large industry close by, and were less than half an hours walk from a motorable road so that the study team could feasibly access them. We have labelled these kebele ‘A-D’ to maintain anonymity. Table 1 shows the characteristics of the selected kebeles, all of which had a subsistence farming based economy. Although the study sites were all a short walk from a motorable road, access to public transport was very limited. Characteristics of study kebele SNNPR, Southern Nations, Nationalities and Peoples Region. Data were collected as part of a study to understand how HEWs influence maternal and newborn care behaviours, of which facility delivery was one. Four trained interviewers collected data in the local language using pretested semi structured guides developed by the authors. When needed translators were used. The content of the guides was informed by a theoretical framework, which identified pathways through which HEWs could influence behaviours by modifying families capabilities, opportunities and motivation.31 Data were collected from mothers, grandmothers, fathers, HEWs and HDA leaders using narrative interviews, in-depth interviews (IDIs) and focus group discussions (FGDs). All community respondents had children or grandchildren under 12 months of age, with narrative mothers having children less than 3 months of age to facilitate recall. Using a range of both methods and respondents allowed for data triangulation and ensured we captured a range of viewpoints. Narrative interviews with mothers were used to capture personal experiences, in-depth interviews to capture perceptions of what was commonly done in the community, and focus group discussions to collect data that we felt would benefit from being discussed in a group interaction. Data were collected until saturation was reached, that is, until additional interviews provided similar information to that already obtained. Saturation was determined by frequent transcript reviews. The sample size, respondent groups and the interview content related to facility delivery are shown in table 2. In the FGD, we employed several activity oriented exercises such as sorting and ranking to encourage group interaction and participation and reduce social desirability bias, which can be a particular issue in Ethiopia.32 Data collection method, sample size and content related to facility delivery HDA, Health Development Army; HEW, health extension workers; FGD, focus group discussion. Mothers, grandmothers and fathers, from different households, were identified by the HEW/HDA leaders or through snowball sampling from the community respondents – with the first method providing the majority of respondents. Eligibility criteria were that the family had received at least one visit by an HEW or HDA leader. Mothers were selected to ensure diversity in age, educational level, parity, sex of newborn and socioeconomic status. Grandmothers could be paternal or maternal— dependent on which was closest to the family. We also aimed to get diversity in place of delivery, but located few women who admitted they delivered at home. All of the HDA leaders in the study kebeles were invited for the HDA FGDs. As there were only two HEWs per kebele, HEW FGDs included HEWs from neighbouring kebeles. Interviewers approached potential respondents in their home, or at the health post. Three respondents refused, as they were too busy. Interviews lasted from 1 to 2 hours and took place in respondents’ houses, or the health post for the HEW. FGDs were conducted with 3–7 respondents in neutral locations and lasted from 1.5 to 2.5 hours. HEWs and HDA leaders were not present during any of the interviews or FGDs with community members. Interviews and FGDs were audio-recorded and fully transcribed by the data collectors in English as soon as possible. Data collectors met regularly during fieldwork to discuss emerging themes and to receive feedback from the senior researchers. On entering a householdthe interviewer introduced themselves and the project to key people, and gave the head of household a project leaflet. They explained who they wanted to interview and read aloud a study information sheet to them in a quiet place. For FGDs the information was read aloud to all FGD respondents. The interviewers checked respondents’ comprehension, rephrased if necessary and gave the respondents an opportunity to ask questions. If the respondent agreed to be interviewed the interviewer read the consent form out loud and asked the respondent to sign to show that they were willing to be interviewed, understood the study, were happy for their words to be written down and recorded, were happy for their quotes to be used and for the information collected to be transferred to London. The interviewers also signed each form. Respondents were not directly involved in the design of the study, however the interview guides were iterative and were modified as the research progressed based on reported experiences and perceptions. Some respondents were recruited through snowball sampling, that is, where respondents suggested others they knew who were eligible for interview. Analysis began during data collection through regular team meetings and reflection. A formal analysis session was held with the data collectors in the middle and at the end of data collection, this included discussion of how our characteristics could have influenced how data were collected and interpreted. Once data were collected all transcripts were read several times to ensure familiarity with the data, to begin to identify notable constructs, and to see the data as a whole. A deductive coding template was developed in Nvivo based on the theoretical framework that guided the interview content. Interviews and focus groups were then coded inductively within these broad themes. Coding was done by identifying the underlying meaning of each section of text and how it was different or similar to others section. Codes that contained similar concepts were then put into larger themes. Themes and codes were modified by looking for patterns, links and contradictions within themes. Data credibility was checked by triangulating data between respondent groups and between data collection methods. Data analysis was done by three of the senior researchers, who discussed their coding regularly to enhance conceptual thinking and to increase coding rigour. Reflective notes were kept throughout the process.