Stillbirth surveillance and review in rural districts in Bangladesh

The four target districts, Jamalpur, Moulvibazar, Narail and Thakurgaon, comprising a total population of 6.7 million, were targeted based on their relatively poor maternal and neonatal health indicators: uptake of antenatal care (ANC) (63.5% versus 67.7% nationally) and percentage of deliveries attended by a skilled provider (19.9% versus 31.4% nationally) [9]. In addition, these districts were the target districts for the joint Government-UN Maternal Newborn Health Initiative which focused on saving maternal and neonatal lives through improved district level planning, investments in infrastructure and supplies and also strengthening of human resources.

All grassroot level health and family planning workers, Health Assistants (HA) and Family Welfare Assistants (FWA) (each responsible for a population of 5000–6000) were trained to identify all stillbirths at household level in their area using the following definition of stillbirth: “Birth of a baby after 28 weeks of gestation and who showed no evidence of life, such as beating of the heart, pulsation of the umbilical cord, or definite movement of voluntary muscles, whether or not the umbilical cord has been cut or the placenta is attached, immediately after birth” [10]. This definition also conforms with the WHO definition of stillbirth for international comparison [11].

An expert, multi-disciplinary team under the guidance of the Directorate General of Health Services and the Directorate General of Family Planning (DGFP) developed a verbal autopsy (VA) questionnaire based on recommended WHO VA tools and the existing neonatal death audit forms already available in Bangladesh [12, 13]. The questionnaire was adapted for use by district healthcare workers and family planning workers and translated into Bangla (Additional file 1). The VA questionnaire was then field tested in one district before use in all four study districts. The VA questionnaire included 29 close-ended questions with different response categories including questions on sociodemographic characteristics of the mother, complications during pregnancy and childbirth, access to antenatal care, mode of delivery, healthcare seeking behaviour at the time of antepartum or intrapartum complications, and characteristics of the stillbirth.

To assess if a stillbirth was fresh (likely intrapartum death) or macerated (likely antepartum death), questions around the appearance at birth and whether or not mothers had felt the baby move were included in the VA questionnaire.

Field level supervisors of HAs and FWAs (i.e. Health Inspectors (HI), Assistant Health Inspectors (AHI), and Family Planning Inspectors (FPI)) who oversee a population of 25,000–30,000, received training on the use of the VA questionnaire, facilitated by a team of trained social scientists and medical doctors. Each supervisor conducted at least five VA under supervision to ensure competency in using the tool.

Upon receipt of a death notification, a trained field level supervisor was assigned to conduct a VA at household level of every fifth stillbirth (identified sequentially) which occurred in his/her assigned area. Each VA interview had to include three respondents including the mother, father or relatives who were present at the time of the stillbirth (Fig. 1).

Sociodemographic characteristics of the stillbirths and mothers as well as the circumstances around the death were described using frequencies and proportions.

An expert panel was created, comprising of an obstetrician, a physician experienced in stillbirth classifications and a paediatrician, to develop a hierarchical model to assign cause of stillbirth. The model was based on information obtained from a systematic literature review to determine cause of and factors associated with stillbirth in LMIC [8]. We considered all causes with characteristics which could potentially be identified using verbal autopsy (VA). Causes with the strongest available evidence to support the diagnosis were ranked first. Eight causes of stillbirth were identified: eclampsia, hypertension, antepartum haemorrhage, diabetes, infection, intrapartum-related hypoxia, external trauma, and ruptured uterus. A category “unknown” was used for any case where there was insufficient or no information to be able to assign a cause of death.

We developed computer-based algorithms containing the specified characteristics of each condition to assign cause of stillbirth for each identified case of stillbirth for which verbal autopsy was conducted. Criteria for diagnosis were developed and agreed with input from all researchers, following a hierarchical model (Additional file 2). For each diagnosis, we considered how likely it is for the diagnosis to be accurate using data obtained via verbal autopsy only. For example, a history of hypertension with convulsion is more likely to be an accurate indication of eclampsia than a history of fever and jaundice for the diagnosis of infection. The algorithm was applied independently by two researchers to all cases of stillbirth initially, and then to fresh and macerated stillbirths separately. There were no discrepancies in results obtained between the two researchers.

Ethical clearance was obtained from the Ethical Review Committee (ERC/CIPRB/2010/01) of the Centre for Injury Prevention and Research Bangladesh. The protocol and tools were reviewed and approved for implementation by the Directorate General of Health Services of Bangladesh. Informed, written consent was requested and obtained from each respondent before each verbal autopsy interview. Anonymity and confidentiality was maintained throughout the process. Participation was voluntary.

A total of 6333 stillbirths were identified between January 2011 and December 2012 in the study area, 3025 in year 1 and 3308 in year 2. A total of 1327 VA (20.9% of all stillbirths reported) were performed.

Using the 2011 crude birth rate of Bangladesh of 22.6 livebirths per 1000 population [13], we estimated a total of 151,420 live births in the study area per year giving an estimated stillbirth rate of 19.6 and 21.4 per 1000 births in 2011 and 2012, respectively.

The majority of the VA conducted (65%) included the mothers of the babies who had been stillborn. The mean (SD) age of the mothers was 23.7 (±5.7) years. Most of the mothers (1032, 77.7%) had at least completed primary education and 283 (21.3%) had completed secondary education or above.

Five hundred sixty-eight (42.8%) of mothers were primiparous. The majority (802, 60.4%) reported delivery of a term stillbirth, while 421 (31.7%) reported a preterm delivery and 104 (7.8%) a post-term.

Most mothers (1059, 79.8%) attended ANC at least once during the index pregnancy and 449 (33.8%) attended four or more times (Table 1). Almost 6 in 10 (57.3%) had received care from a skilled healthcare provider (a doctor, a nurse / midwife or a medical assistant).

More than half of all stillbirths (715 (53.9%)) were born at home or at the house of a traditional birth attendant. Most of the mothers had a normal vaginal delivery (1147, 86.4%). Overall, 10.7% of stillbirths were delivered by Caesarean section and 2.5% by assisted vaginal delivery (vacuum or forceps) or were a breech delivery. 614 (46.3%) of the deliveries were assisted by skilled healthcare providers.

During VA, 1041 (78.4%) of the mothers reported at least one type of complication during pregnancy (antepartum) and 865 (65.2%) reported at least one type of complication at time of birth (Table 2).

Only 11% of women had not recognised nor reported a complication during pregnancy or at the time of birth. This was similar for both antepartum and intrapartum.

Among the 1041 mothers who experienced an antepartum complication, 725 (69.6%) had sought care. Among them, 438 (60.4%) accessed care at a health facility. The healthcare facility was a secondary or tertiary level of care in 310 (42.6%) of the cases (Fig. 2).

Among the 725 women who sought care for antepartum complications, 349 (48.1%) received care from a medical doctor trained in identifying and treating most antenatal complications.

The sex of the stillborn babies was known in 928 (69.9%) of the cases. Of these, 534 (57.5%) were males. A total of 272 (20.5%) were described by mothers as “smaller than normal at birth”.

At the time of verbal autopsy, 848 (63.9%) of the stillbirths were reported to be fresh stillbirths and 464 (34.9%) macerated at time of birth. However, when considering stricter criteria (including only those who had reported fetal movement at onset of labour and were considered to look like fresh stillbirths), only 491 cases (58.3%) could be reported as intrapartum stillbirths and 351 (41.7%) as antepartum stillbirths.

Of the 1327 cases of stillbirth, 66 (5%) were due to eclampsia, 136 (10.2%) to hypertension, 182 (13.7%) to antepartum haemorrhage, 41 (3.1%) to diabetes, 115 (8.7%) to infections, 96 (7.2%) to intrapartum-related hypoxia (only those without antepartum haemorrhage, hypertension or eclampsia identified – the hierarchical model), and 3 (0.2%) due to external trauma. No cases of ruptured uterus were reported. No cause could be assigned in 688 (51.9%) of the cases (Table 3; Additional file 3). There was no significant difference between the distribution of likely cause of death for fresh compared to macerated stillbirths (p = 0.444), after accounting for intrapartum-related hypoxia, which may not be a cause of death in macerated stillbirths.

To the best of our knowledge, this is one of the largest studies to implement a stillbirth surveillance and review programme at population level in a low- or middle-income country. This study shows that district level surveillance of all stillbirths can be introduced successfully and provide a population-based, contemporaneous stillbirth rate using verbal autopsy. This can provide information on type of stillbirth (ante or intrapartum) and on likely cause of, and factors contributing to, stillbirth.

Community-based surveillance was implemented successfully in four districts in Bangladesh where 53.9% of identified stillbirths occurred at home. Two thirds of all mothers had accessed care during pregnancy for complications recognised by them (or their family) to require health care. Two thirds of these attended a healthcare facility and half of the mothers reported that they had received care from a highly trained healthcare provider.

Antepartum haemorrhage and hypertension or eclampsia were identified as the commonest causes of stillbirth accounting for almost 30% of stillbirths. Maternal infections were the third most common cause of stillbirth and identified in almost 10% of cases. However, using information obtained via verbal autopsy, did not allow the identification of a clear cause of death in half of all stillbirths. About two thirds of the cases (63.9%) were intrapartum stillbirths. There were no substantial differences in cause of stillbirth between fresh (intrapartum) and macerated stillbirths (antepartum death) except for intrapartum-related hypoxia which was present in at least 11.3% of stillbirths.

The stillbirth rate (SBR) obtained was 20.4 per 1000 births. The neonatal mortality rate (NMR) in the study area during the same period was 24.4 per 1000 live births [14]. Bangladesh does not have a system for vital registration data, but in this study, specific effort was made to identify stillbirths. As a proxy, we estimated the SBR/NMR ratio, which was 0.84. This is in line with high-resource settings with better civil registration systems and vital statistics. The median ratio of SBR to NMR in high income countries is 0.9 (IQR: 0.65–1.15) [4]. This may indicate that the majority of stillbirths in this study population were identified with the introduction of the population-based stillbirth surveillance.

Findings should be interpreted in light of some limitations. Firstly, as cause of death was assigned hierarchically, the proportion for each cause of death could potentially change as the hierarchy changes. Secondly, verbal autopsy was used to develop a ‘clinical history’ for each case. A limitation of this is that additional clinical information that could have aided diagnosis of cause of death, including results of laboratory tests, was not available. Thus, this study allowed for the estimation of likely cause of stillbirths for those who died either at facility level or in the community in only about half of the cases (51.9%). Assigning cause of death from information obtained via verbal autopsy is known to be difficult. Studies from Bangladesh, Pakistan, Tanzania, and Ghana have reported an undetermined cause of stillbirth in 18 to 58% of cases [7, 15‐19].

Furthermore, the maternal infection category does not provide information on type of infection and whether potentially preventable (such as HIV, malaria, syphilis or tuberculosis). The questionnaire used during verbal autopsy in Bangladesh also did not include questions to ascertain whether stillborn babies were identified to have had a congenital anomaly. This is because most congenital anomalies causing death are cardiovascular and chromosomal anomalies and most often cannot be detected by parents in the community [20]. In an earlier systematic review identifying causes of and contributing factors to stillbirth in low- and middle-income countries [8], the main causes of stillbirth (ordered by frequency of reporting) were maternal factors, congenital anomalies, placental causes, asphyxia, umbilical problems, and uterine factors. However, most of the studies included in the review were hospital-based, making the comparison with the results of this study difficult.

Nevertheless, in a similar study using verbal autopsy in India, Aggarwal et al. reported causes of stillbirth to include hypertension (30%), antepartum haemorrhage (16%), underlying maternal illness (12%), congenital malformations (12%) and obstetric complications (unspecified) (10%) [21]. However, in their study, stillbirths were defined as death from 24 weeks gestation and this could have explained some of the differences in cause of death observed, particularly with regard to congenital anomalies.

Baqui et al. and Nahar et al. in Bangladesh, as well as Jehan et al. in Pakistan, similarly found that maternal haemorrhage was one of the main causes of stillbirth [7, 15, 16]. Hypertensive disorders as a major cause of stillbirth was reported by Edmond et al. in Ghana [18, 19]. In Tanzania, Hinderaker et al. found that around 42% of mothers had an infection, which is four times higher than our results [17]. However, they targeted only rural communities and had a relatively small sample of 60 stillbirths compared to the sample size in this study.

To increase the proportion of cases for which a cause of death can be determined, and with a greater level of certainty, analysis of hospital records, diagnoses and management pathways would be needed. However, this would only be feasible in cases where women have been admitted to a healthcare facility and/or for whom good patient-records are available for review.

The variability in causes of stillbirth among studies is probably due to the use of different hierarchical models in studies [7, 18]. The proportion of any cause is dependent on the proportion of other causes and if the hierarchy of causes changes, the relative importance of any cause varies in relation to the others. In earlier studies, two different hierarchical models were applied to antepartum and intrapartum stillbirths, whereas we applied a unique model to all stillbirths. In addition, different definitions of each potential cause of stillbirth were used. For example, Hinderaker et al. defined maternal infection as “all kinds of infections” [17]; whereas, the definition used in this study was “the presence of fever and jaundice or fever and premature rupture of membranes”. In addition, the prevalence of malaria, the main cause of infection in the study conducted by Hinderaker et al., was higher than in our setting.

In terms of the proportion of ante- and intrapartum stillbirths, our finding, that the proportion of intrapartum stillbirths is higher, is in line with studies conducted in Bangladesh and Pakistan [15, 16, 22, 23]. However, Baqui et al. as well as Edmond et al. found that antepartum stillbirths represent two thirds of all stillbirths [7, 18, 19]. Intrapartum-related hypoxia was estimated to be the cause of stillbirth in 23 to 25% of cases [15, 16], which is substantially higher than the 11% we found. Another study [7], which had a lower rate of intrapartum stillbirths (37.9%), assigned intrapartum-related hypoxia as the cause of stillbirth in 20.5% of stillbirth cases.

However, half of global stillbirths occur at intrapartum period [1]. The differences observed between studies may be because, in LMIC, differentiation of antepartum and intrapartum stillbirths relies mainly on the physical appearance (fresh/macerated classification) of the stillborn, which is often not a reliable way of determining time of death. Besides it has been previously noted that healthcare providers’ assessment of physical appearance at time of stillbirth may, in fact, be an unreliable method for assessing time of death [24].

Regarding care seeking for antepartum complications, we found similar results to the study conducted by Skider et al. in a rural district of Bangladesh [25]. In both studies, more than two thirds of women had accessed and received care from a trained healthcare provider. However, the proportion of women receiving care by a trained healthcare provider was lower than in our study (30% versus 48%).