Background
Deliveries that occur before arrival at a health care facility (birth before arrival or BBA) – also called accidental, unplanned or out-of-hospital births or deliveries – are associated with higher rates of maternal and neonatal morbidity and mortality compared to facility deliveries or planned home births [
1‐
3]. Maternal outcomes highlighted in a recent review include increased complications and higher morbidity including more frequent tearing, increased blood loss and increased risk of longer third stage of labour [
2]. Neonatal complications associated with BBAs include low body temperature and low blood glucose [
4] as well as low birth weight [
1,
2,
5,
6]. Across several studies, neonatal mortality following BBAs has been reported as six to 11 times greater than hospital births [
2], and BBAs constitute an outsized proportion of death given the relative rarity of a BBA event [
7]. Only one study, from Australia, has qualitatively examined BBAs; it concluded that a BBA leads to heightened feelings of anxiety and fear among mothers [
8].
While research on BBAs has been conducted in several high-income countries, there are almost no studies from low-income settings where a majority of maternal and newborn deaths occur. BBAs are uncommon globally with reported rates across high-income countries ranging from 0.08 % [
5] to 1.99 % [
4]. Temporal trends reveal a rise in BBAs in some wealthy nations [
9], including a doubling of the incidence in Finland between the 1970s and 1990s [
10] and Ireland between 2005 and 2009 [
11], but staying below 0.5 % in these cases. These increases have been attributed to closures of remote maternity wards and subsequent transport barriers for rural women [
1,
10], a larger migrant population [
11] and delayed departure to facilities among labouring women [
3] including those living near a facility [
6].
In high-income settings, the most widely agreed upon predisposing factor for BBAs is poor antenatal care attendance [
3,
4,
7,
10,
12‐
16], although this trend has not been universal [
6]. Groups found to be at increased risk for BBA include young women [
14], single women [
1,
7,
12], women from disadvantaged socio-economic backgrounds [
11,
17], young primigravida women [
18], as well as older, multigravida women [
18]. Two distinct groups of women described in the literature who appear predisposed to BBA are: women who are older, multiparous and with shorter labors [
3,
6,
10,
12,
18,
19], or women, typically of a young age, who are attempting to conceal their pregnancy [
1,
7,
12]. Quantitative research on trends and patterns in BBAs has been conducted in Australia [
2], England [
1,
6,
12], Finland [
10], France [
14,
16], Hong Kong [
17], Ireland [
7,
11], Israel [
20], Italy [
5], Japan [
13], Scotland [
3] and the United States [
4,
21]. A majority of these studies reported on BBAs that occurred in the presence of a woman’s partner, without a medical professional and within a woman’s home, but also in ambulances, emergency rooms, hospital entrances, taxis, parking lots and public bathrooms.
To our knowledge, one study has addressed BBAs in Africa. In a prospective one-to-one matching study in South Africa, BBAs were associated with distance to obstetric wards, parity, complications during pregnancy, shorter duration of labor and a previous home delivery [
22]. In a study from Zambia that compared women’s perceptions of home and facility births, respondents described how concerns about giving birth en route contributed to a preference for home delivery [
23].
Limitations within the existing BBA literature confine our understanding of the issue. First, there is limited qualitative literature on the experience of BBAs [
8]. Second, there is a paucity of literature on BBAs in low-income countries. Finally, we are not aware of BBA literature that incorporates perspectives of male partners, although the role of men in deciding whether or when to seek care outside the home has emerged as crucial in the maternal health literature [
24], including literature from Tanzania [
25]. In this paper, we aim to highlight risk factors for a BBA in a low-income setting and to describe how local reproductive health norms and practices, including how couples define and negotiate each partner’s role in careseeking for birth, can shape circumstances surrounding a BBA.
Study location
In Tanzania, home and facility births occur in relatively even proportions. As of 2010, 50.1 % of births occur in a facility, which approximates the rate of facility births reported in 1991–2 [
26,
27]. Maternal and neonatal mortality are high with 454 maternal deaths and 2586 neonatal deaths per 100,000 live births; the total fertility rate is 5.3 [
26]. One in 38 women will die due to maternal causes in her lifetime [
28]. For every 1000 births, four to five women die from pregnancy-related causes and 26 neonates die [
26]. Tanzania is among ten countries that account for a majority of the world’s “first-day deaths,” or death on the first day of life [
28]. In Morogoro region, nearly one quarter of women have no education, 34.3 % have had some primary education and 32.8 % have completed primary school [
26]. While most women can read a whole sentence, 26 % of women cannot read at all [
26]. More than 60 % of men and women are engaged in agriculture as their primary occupation [
26]. In the Eastern Zone, which encompasses Morogoro region, the infant mortality rate is 70 deaths per 1000 live births [
26]. Maternal mortality estimates are not available at the regional level in the Demographic and Health Survey [
26].
Methods
This mixed methods, cross sectional study was conducted as part of a larger program evaluation of a maternal and neonatal health program designed to encourage uptake of facility-based services related to pregnancy and childbirth in Morogoro Region. Drawing on a household survey and in-depth interviews conducted in Morogoro Region, Tanzania, this study presents quantitative and qualitative data on BBAs across four rural districts (Morogoro Rural, Kilosa, Ulanga, and Mvomero District Council). Women and their partners were eligible to participate in the study if the family had experienced a birth within the preceding 14 months (a cutoff set with the intention of reducing recall bias). We define a BBA as the occurrence of a birth prior to arrival at a health facility when, according to respondents, that birth was intended to occur in a health facility. This is consistent with a BBA definition employed by Tanzanian health facilities for record-keeping, although in some facilities practitioners also require that a mother present physical evidence of a recent birth (such as a placenta or other birth remnants), which is considered a means to verify place-of-delivery intent (the authors can only speculate on how this practice serves to verify intent).
Study design
Training
Research managers trained teams of five qualitative and 16 quantitative research assistants on research methods, ethics and various facets of careseeking for maternal and neonatal health during antenatal, intrapartum and postpartum periods. Trainings were followed by pilot testing and tool revision among respondents who fit the study’s eligibility criteria (detailed below) and who lived in or near Dar es Salaam (where trainings took place). Research assistants were college-educated Tanzanians, with a relatively even breakdown of males and females trained as teachers, health practitioners and social sciences or public health graduate students.
Mixed methods design
This study employed a mixed-methods convergent parallel design, wherein qualitative and quantitative data were collected during approximately the same time period, and the aim was to compare and contrast findings across the two strands of data [
29]. Qualitative data analysis preceded and informed quantitative data analysis. Further details of analysis are presented below.
Quantitative design and sampling
The household survey (see Additional file
1) was designed to be regionally representative of rural Morogoro and self-weighting via a multistage cluster sampling survey. Data were collected from August to November in 2011. Sixty clusters were identified via probability proportional to size (PPS) sampling methods and 30–35 women were surveyed in each cluster. In each cluster, the survey team visited all households to identify eligible women. If a household had more than one eligible woman, the interviewer compiled a list of the eligible women in the household and randomly selected one from the list. Further details of our quantitative design including details of the sampling frame and clusters have been outlined in related publications [
30,
31].
Quantitative analysis
Quantitative analyses were conducted with Stata (Version 13.1, College Station, TX). Data points for 59 births before arrival and 1267 facility births were analyzed. For this analysis, a birth was categorized as a BBA if a woman responded to the question “Where did your delivery occur?” by stating that the delivery took place “On the way to a facility” or “In the home of a stranger while on the way to a facility”. A birth was categorized as a facility birth if a woman stated that a birth occurred in any type of hospital (regional, district, or local) or health center or dispensary. Due to the rarity of the outcome, few parameters could be estimated with logistic regression, and the effects of multicollinearity could have been easily amplified [
32]. To avoid estimation errors, we performed a series of analyses. In each phase of the analysis, we first estimated the effect of the most prominent predictor, then matched cases and controls on that predictor so that we could assess the effect of less prominent predictors in the next phase while controlling for previously identified predictors via matching. In this way, we matched on key variables and estimated effects on a limited number of remaining variables. To determine key variables for matching, we conducted logistic regressions to assess for risk factors associated with the outcome variable (BBA). Variables included age, parity, education and wealth, where wealth was calculated using a principal components analysis of household assets [
33]. We then used linear regression to assess the relationship of the strongest predictors of BBAs with all other predictors. Finally, we assessed candidate key variables for matching potential, that is, whether enough matches existed in the data to use a key variable as a matching variable. For example, socio-economic status was considered a key variable because it was the strongest predictor of BBA, it was highly correlated with other predictors, and it offered multiple matches among the 1267 facility births and the 59 BBAs. Our identification of key variables was also informed by existing BBA and maternal health literature, which highlights the importance of age at delivery, age at first pregnancy, parity, wealth, number of ANC visits, maternal education and partner education. To match data points, we performed the coarsened exact matching procedure using the Stata command cem, which maximizes the number of matches while retaining a meaningful level of precision in matches [
34]. We report unadjusted odds ratios for each predictor variable. At each stage of matching, unadjusted odds ratios were reported for all remaining variables, and those with
p-values <0.10 were included in a multiple logistic regression model from which adjusted odds ratios were reported.
Qualitative design and sampling
In-depth interviews (see Additional file
2) with women and partners who experienced BBAs were selected from a larger study on careseeking during pregnancy and childbirth among women (
n = 49) and their partners (
n = 27) living near (<3 km) and far (≥3 km) from facilities. Women were neither more likely to be selected nor ineligible for the qualitative portion of the study if they were selected for the quantitative survey. Data were collected in July and August of 2011. Researchers identified women for the larger study through engagement with village health committees and canvassing villages to invite eligible mothers to participate. Women and partners for the larger study were not recruited based on their experience of a BBA. The BBA experience was identified as salient during data collection and concurrent qualitative analysis, leading to increased probing regarding BBA during later interviews. All interviews indicating a BBA experience were selected for this analysis, yielding 13 interviews (including six follow-up interviews) with four women and three partners, with one woman per district. Further details of our qualitative design and sampling have been outlined in related publications [
35].
Data collectors interviewed all respondents one-on-one, in a private location of their choosing. The interview included open-ended questions such as “Could you please walk me through your experience from the moment you sensed the baby was going to be born?” and probes such as “Please tell me more about that”. Data collectors were instructed to respect respondent autonomy - particularly in the event that sensitivities or uneasiness emerged during interviews. In the event of inconsistencies comparing women’s versus their husbands’ accounts of the birth experience, interviewers were careful not to highlight or probe on these inconsistencies with either party. A field supervisor led daily debriefing sessions with the qualitative team throughout data collection to triangulate findings, strengthen probing, build field notes, identify topics to address in future interviews and develop themes for a codebook that would later be applied to transcripts.
Qualitative analysis
Qualitative analysis drew upon 13 in-depth interviews (IDI), including six follow-up interviews, from four husband-wife pairs who experienced a BBA. In this study, follow-up interviews were conducted to clarify points that were considered unclear or insufficiently discussed in the analysis of the initial interview. Interviews with one partner had to be repeatedly cancelled due to his intoxication (from alcohol). All interviews were recorded, transcribed and quality controlled by bilingual researchers to ensure that the content of the recorded interview was reflected in the transcript. A case study approach was used to analyze and present the qualitative data [
36‐
38]. Cases were first assembled from raw case data drawn from interviews with women and their partners. In line with Patton 2002, narratives were written in a manner intended to present a fluid and coherent description of the BBA in a chronological fashion [
36]. Case study researchers emphasize that “the analyst’s first and foremost responsibility consists of doing justice to each individual case” [
36]. Due to space limitations, we could not include narratives as holistic entities within this article; however narratives can be read as supplementary data (see Additional file
3). Results of a cross-case analysis, which was conducted to inductively identify themes across narratives, are detailed in the results section.
Discussion
Drawing on quantitative and qualitative methods, this study explored the experience of BBA among women living in rural areas across Morogoro Region in Tanzania.
A factor that emerged as critically important in both surveys and interviews and which correlates with a wide body of anthropological and epidemiological literature - as well as some BBA literature - was the role of poverty and the ways in which husband-wife pairs of a low socio-economic status are at a disproportionately higher risk of experiencing a BBA [
11,
17,
22,
41‐
43]. In this study, economic constraints amplified discord on delivery location. Although Tanzania's official policy states that delivery services should be free of charge, husbands in this study described concerns related to fees and reservations regarding quality of care, factors that have been described in several studies conducted in Tanzania [
42‐
47]. Partner disagreement on delivery location preference in Tanzania has been significantly associated with reduced rates of facility births; when both partners rated the skills of government doctors and nurses as higher than that of TBAs, women were twice as likely to deliver at a health facility than in the home, even after controlling for confounders including age, wealth, and education [
39]. Women’s delayed departure illuminates how, in a context of not only unequal power relations but also severely constrained economic resources, it is largely beyond the control of laboring mothers to determine when they will depart and how they will reach a facility. It is also largely beyond the control of husbands to fulfill their socially-expected breadwinner role and provide funds for transport and birth supplies (such as razors, gloves, a plastic sheet and a new kanga or cloth) [
25].
Quantitatively, once wealth was controlled for via matching, higher parity and fewer ANC visits continued to significantly predict higher odds of BBA. The protective effect of multiple ANC visits has been consistently highlighted in both BBA literature [
3,
4,
7,
10,
12,
14,
15] and maternal health literature as a means to monitor health, promote health-seeking behavior and devise birth preparedness plans [
48]. Our qualitative interviews highlighted potential causes of this relationship: exposure to formal care through ANC visits offered educational opportunities to women regarding facility births and provided women with experiences that increased their sense of trust in and value of facility-based care. The role of parity has also been highlighted in previous BBA studies [
2,
3] that offer both biological and behavioral explanations for the relationship of parity to BBA. In the cases presented here, we see women and especially husbands making careseeking decisions based on prior experiences, with one husband describing how he delayed care to avoid the long wait times experienced during previous deliveries. Unlike earlier studies [
49], this study did not find trends by educational level.
Multiple birth narratives were elicited during in-depth interviews. Because the research design allowed for follow-up interviews, interviewers sought to probe on divergent accounts of births, particularly in instances where birth accounts appeared contradictory within or across respondents. In the interest of building trust, maintaining confidentiality, and not causing undue strain between husband-wife pairs, extensive probing in the name of triangulation was not undertaken and several inconsistencies remain unexplored. The context of this study involved social spheres and spaces wherein respondents likely placed themselves within a different “social location” [
50] than interviewers and may have therefore felt compelled to adjust accounts. This has been referred to in the literature as “a performance of identity” with an understanding that respondents may sometimes engage in presenting “what I think might make me valued by others” or by revealing a “preferred self” rather than an “essential self” [
50]. Interviewers – all urban, multi-lingual, college graduates – likely drew upon their own subjectivity during interviews, thereby influencing the production of knowledge as it related to BBA account. In analyzing and presenting these narratives, rather than seeking an objective truth, we sought to engage in a “dialogue with the transcripts, listening to them and asking questions of them” to determine a “contextual truth” [
51]. While reading for a contextual truth does not lead to a single, objective truth, we argue that it does illuminate how women and men reshape the story of their lives in an adaptive, socially desirable manner. In viewing the data this way, we recognize potential social pressures women may feel to tell researchers that they delivered in health facilities. We also appreciate the pressure men may feel to present themselves as physically present alongside women throughout labor and delivery, and as financially capable of providing for women before and during childbirth. Given this understanding, we suspect that our survey and other surveys related to careseeking for childbirth, underestimate the number of BBAs and overestimate the number of facility-based births (and potentially also the number of home births).
While it would be impossible to prevent all BBAs, strategies have been proposed to minimize incidence [
2]. Interventions could consider re-affirming the importance of birth preparedness plans (with preparedness messaging directed at women
and their husbands), expanding or improving the capacity of maternal waiting homes, and instructing families during ANC visits on a minimum amount of care required in the event of a BBA (such as keeping a baby warm, cutting a cord with a clean razor and ensuring delivery of a placenta) [
7,
12,
18]. While respondents in our study attended several ANC visits where they were instructed on the importance of bringing supplies for birth, participants did not discuss being guided on how they could gradually save funds to afford costs associated with birth, which represents an opportunity for improved birth preparedness messaging.
This study is limited in that it relies on a very small qualitative data set of four women who delivered en route and three of their partners. The study was strengthened by the use of both qualitative and quantitative methods. Qualitative methods highlighted the reality of BBAs to the research team, which informed the decision to include a measure of births en route in the survey. Quantitative measures allowed us to examine generalizable trends in the data and assess statistical trends in light of narrative themes. The uniqueness of this study stems from not only presentation of BBA data in an East African context, but also from the presentation of a male perspective on BBAs. As partners to women and fathers to children, men exert positive and negative influences over maternal health [
52]. The influence of men’s intentions and practices on childbirth has been described as “little studied” [
52] and in contexts such as Tanzania we urge that more attention be paid toward examining the role of men in careseeking for childbirth. We hope this research sparks more interest in the topic of BBAs and birth preparedness in low-income settings.
Acknowledgements
We are grateful to Neal Brandes of USAID for his guidance during the drafting of this manuscript. The authors would like to thank the data collection team including: Amrad Charles, Emmanuel Massawe, Maurus Mpunga, Rozalia Mtaturo; the Ministry of Health and Social Welfare including Neema Rusibamayila, Georgina Msemo, Helen Semu and Koheleth Winani; the MUHAS-based team consisting of Japhet Killewo (PI), Switbert Kamazima, Charles Kilewo, Rose Mpembeni, David Urassa, Aisha Omary, and Deogratias Maufi; the Jhpiego-Tanzania based team consisting of Dunstan Bishanga, Maryjane Lacoste, Chrisostom Lipingu, Marya Plotkin; the USAID team (Troy Jacobs, Raz Stevenson, Miriam Kombe); the Jhpiego-US team consisting of Eva Bazant, Elaine Charurat, Chelsea Cooper; and the JHSPH-based team consisting of Jennifer Applegate, Abdullah Baqui (PI), Carla Blauvelt, Jennifer Callaghan, Asha George, Shivam Gupta, Amnesty LeFevre and Diwakar Mohan. We thank the respondents who participated in this study.