Main findings
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.
Stillbirth rate
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.
Cause of death
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.
Hierarchical model for cause of death
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.
Time of death
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].
Care-seeking behaviour
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%).