Unpredictability dictates quality of maternal and newborn care provision in rural Tanzania-A qualitative study of health workers’ perspectives

Background: Health workers are the key to realising the potential of improved quality of care for mothers and newborns in the weak health systems of Sub Saharan Africa. Their perspectives are fundamental to understand the effectiveness of existing improvement programs and to identify ways to strengthen future initiatives. The objective of this study was therefore to examine health worker perspectives of the conditions for maternal and newborn care provision and their perceptions of what constitutes good quality of care in rural Tanzanian health facilities. Methods: In February 2014, we conducted 17 in-depth interviews with different cadres of health workers providing maternal and newborn care in 14 rural health facilities in Tandahimba district, south-eastern Tanzania. These facilities included one district hospital, three health centres and ten dispensaries. Interviews were conducted in Swahili, transcribed verbatim and translated into English. A grounded theory approach was used to guide the analysis, the output of which was one core category, four main categories and several sub-categories. Results: ‘It is like rain’ was identified as the core category, delineating unpredictability as the common denominator for all aspects of maternal and newborn care provision. It implies that conditions such as mothers’ access to and utilisation of health care are unreliable; that availability of resources is uncertain and that health workers have to help and try to balance the situation. Quality of care was perceived to vary as a consequence of these conditions. Health workers stressed the importance of predictability, of ‘things going as intended’, as a sign of good quality care. Conclusions: Unpredictability emerged as a fundamental condition for maternal and newborn care provision, an important determinant and characteristic of quality in this study. We believe that this finding is also relevant for other areas of care in the same setting and may be an important defining factor of a weak health system. Increasing predictability within health services, and focusing on the experience of health workers within these, should be prioritised in order to achieve better quality of care for mothers and newborns. This was a qualitative study using a grounded theory approach to guide the analysis [27], based on 17 in-depth interviews with health workers in Tandahimba district, Mtwara region, south-eastern Tanzania. It was nested within a district-wide quality improvement project, called EQUIP (Expanded Quality Management Using Information Power) [28]. Tandahimba district is located on the Makonde plateau in rural southern Tanzania, close to the border of Mozambique. The population of about 200,000 people is predominantly subsistence farmers but also engage in cashew nut farming. The road network within the district becomes muddy during the rainy season, making emergency transport difficult. The district has 32 health facilities providing maternal and child care, all government apart from one faith-based. The district hospital provides comprehensive Emergency Obstetric and Neonatal Care (EmONC) [25] as well as regular childbirth care. The three health centres are supposed to manage basic EmONC and efforts are underway to upgrade their services to include caesarean sections [29]. In reality however, the functions required to provide even basic EmONC are rarely fulfilled, especially in terms of assisted delivery and provision of Magnesium sulphate for eclampsia. The 28 dispensaries provide childbirth care according to their abilities including the administration of Oxytocin when available, both to prevent and manage postpartum haemorrhage. Capacity to provide neonatal resuscitation has recently increased in all facilities through provision of equipment and targeted training [28]. There is a severe shortage of health workers with 52% of clinical posts vacant in the district [29]. Lower cadre health workers, such as medical attendants, often have to take on the responsibilities of higher cadre health workers such as nurses and midwives; an example being to assist women during childbirth (Additional file 1) [29]. Utilisation of health care during pregnancy is high with all mothers attending Antenatal care (ANC) at least once during pregnancy, although only 47% attend at least four times [1, 2]. The proportion of mothers giving birth in health facilities has increased substantially in recent years and was estimated as high as 87% (CI 95%: 77–93) in 2014 [28]. The maternal mortality ratio however remains high at 579 deaths/100,000 live births in the Mtwara region, which is higher than the national MMR average of 432 [30]. Neonatal mortality in was estimated at around 31 per 1000 in 2013 [2]. Interviews with an open sample of 17 health workers providing maternal and newborn care were conducted in 14 health facilities between the 4th and 14th of February, 2014 (Table 1). Respondents were selected purposively, to represent different cadres of health workers and levels of health facilities, with the aim to achieve variation in the data (Table 1). Characteristics of health workers interviewed Permission to conduct the interviews was obtained from the District Medical Officer in Tandahimba district. Respondents were contacted in advance, by telephone where possible, else by going to see them in the health facility. They were asked if they would be willing to participate and when a suitable time for the interview would be. At the time of the interview, informed written consent was then sought from all respondents. An information sheet about the study was provided and read out by one of the interviewers. Respondents were informed that they could refuse to participate, withdraw or stop the interview without having to state any reason and without any consequences. All respondents chose to participate. Interviews were held in a private area of the health facility. Only the respondent and interviewers were present. Interruptions to allow respondents to attend to their patients were made and for this reason, some interviews took place over the course of a few hours. The median effective interview time was 1 h 12 min (range 1 h 3 min to max 2 h 7 min). Interviews were co-conducted in Swahili by the first (UB) and second (FH) authors. UB is a Swedish medical doctor with 2 years’ experience of working as a clinician and program manager in rural Tanzania. FH is a Tanzanian social scientist with experience from several qualitative studies in southern Tanzania. Both are fluent Swahili-speakers and minimal translation into English by FH to UB was done during the interviews. Neither UB nor FH were known to the respondents beforehand. The interview guide was semi-structured and developed in collaboration between all authors of the study. It was adapted after the first interviews as some questions did not yield sufficient response and to reflect new ideas emerging during the data collection (see Additional file 2). While the number of interviews was pre-determined, it was felt that saturation in the interview material was reached before the last few interviews were conducted as no or little new information emerged. No repeat interviews with respondents were done. Interviews were audio recorded and transcribed verbatim. Subsequent translation into English was conducted with careful attention to quality to ensure preservation of the original meaning. Extensive field-notes were taken during and at the end of each interview. Transcripts were not shared with respondents. Data analysis was conducted using Microsoft Word 2010. It was led by the first author, UB, with frequent and substantial input by the senior author, IH. A grounded theory approach was used for the analysis, the end result of which was a core category [27]. Transcripts and field-notes were initially read and re-read to capture the whole. Transcripts were initially coded inductively, using so called open coding close to the text to capture ‘what was going on’ [31]. Similar codes were categorised into higher order categories. After analysing half of the transcripts, a code list was prepared to harmonise emergent codes and categories, agreed upon through discussions between UB and IH. Subsequent transcripts were coded applying these new codes and categories where possible, using focused coding. Theoretical coding was applied early on using the conditional matrix by Corbin and Strauss [31], focusing on the conditions in which health workers provide care and the consequences these have for care provision. As the analysis progressed, theoretical coding continued using the concepts of mothers’ utilisation of care, the availability of resources in health facilities and the actions taken by health workers, i.e. clinical practice, as it became apparent that the data fitted well into these categories. Main categories contained health worker perceptions of the conditions for care provision and their perspectives of what constitutes good quality care. All main categories were linked to the core category. As the core category emerged, interviews were also reread to find more examples of the core category and to saturate the description of the links between the core category and the main categories. Results were not checked by the respondents.