Data were analysed using thematic analysis with both deductive and inductive coding. The transcripts were initially read for codes based on the field guide questions, and in the next step, the data were categorised into emerging حوامل hemes. The IDIs and FGD data were analysed together. The interviews were transcribed verbatim in Sindhi and the translated into English by JA). The translated interview text was then read again for any errors, and initial coding was done. The coded text identified during the first reading was coded as nodes and sub-nodes in NVIVO version 11 (QSR international). Next, interview text was read and re-read and it was labelled under appropriate nodes. Finally, the node names were edited, and their hierarchy was changed by merging, and cutting and pasting them within the most appropriate themes. A framework was developed by identifying themes as they were explained by the codes. The women’s discussion, of their practices during pregnancy and birth, and encounters with healthcare providers during pregnancy and birth, was used to understand their experiences about care, their perinatal losses, and the circumstances under which these losses occurred. Data from the interviews with the women were triangulated with those from the health officials and community members. We translated common terms used by the participants for their symptoms, complications, or a treatment into the most suitable medical terminology. For example, we interpreted “bleeding during pregnancy” as antepartum haemorrhage, “low blood” as anaemia, “blood injection” as injectable iron, “baby drowned after birth” as hypoxia or birth asphyxia. Extreme caution was observed in interpretation of the data by agreeing on final coding, evaluation of the analysis, and by providing feedback on several drafts of the paper by the two senior researchers who are experienced in qualitative research. We had author discussions to practice reflexivity as we considered our own personal experiences, knowledge, and beliefs. Thorough reflexivity was practiced in the analysis and interpretation phase of the study, and we challenged the first author (who belonged to the same province, spoke Sindhi and had knowledge of local healthcare system) to avoid subjectivity which may have influenced the analysis and interpretation of the data. The findings are reported based on the Consolidated Criteria for Reporting Qualitative Studies .
The study was conducted in Tando Muhammad Khan District in Sindh Province, Pakistan during June and August 2018. The district was selected because it is one of the typical rural districts in the province of Sindh. The region is a predominately rural with a population of ~677,228. There are 11 mother and child health centres functioning within either basic health units, rural health centres or the district hospital. These are all public health facilities and are referred to as that throughout the paper. The district has two major private medical centres, one of which is a teaching hospital and affiliated with a private medical college. The district also has approximately 20 smaller private maternity care providers. There are 83 community midwives in the rural areas who have established stations in their villages with initial support by the provincial Maternal and Child Health Program. Approximately 404 lady health workers (LHWs) are also working in the district. LHWs are publicly funded; however almost half of the district does not have any LHWs appointed. After about 18-months training, a LHW is assigned to provide health education, basic birth spacing interventions and counselling, and nutrition education to about 100 houses in her catchment area near her home. LHWs earn about USD 300 a month and report to the nearby primary healthcare facility and the provincial LHW program. The LHW program has continued to expand since 1994, and currently about 125,000 LHWs cover about 60% of the country .Participants of the study were selected from 19 villages in 5 union councils of the district. These areas represent mostly Sindhi ethnic communities and are within ~10 to 20 kilometres from the main towns. We included four diverse groups of participants: 1) women who had a perinatal death in the previous 12 months, 2), family members of these women, 3) female medical officers and district health management officials, and 4) LHWs. Women who had had a fetal death (after 28 weeks of pregnancy) or an early newborn death (in the first seven days following birth), in the year preceding the interviews were considered eligible to be included in the study. Women were identified by the local LHWs who record all perinatal deaths in their catchment areas, regardless of the place of death; homes, private maternity centres, public health facilities, and these are then reported to the district health office and the LHW program. The LHWs contacted the women on the phone and helped fix the date and time for the interview. Further recruitment of women in the study was stopped when we reached data saturation and a reasonable representation of women from most prominent areas of the district was achieved.