BACKGROUND

Over the past two decades, impressive strides have been made in reducing levels of infant and childhood mortality and in increasing levels of contraceptive use in developing countries. In contrast, progress in reducing levels of maternal mortality and in making pregnancy and childbearing safer for women, despite being a central element of the Millennium Development Goals, has been much slower.1 Although exact figures remain elusive, an estimated 525,000 women, almost all from developing countries, continue to die each year from maternal causes.2

In Bangladesh, the reduction in levels of maternal mortality and improvement of maternal health have been central policy and program goals since the fourth Population and Health Programme, which began in 1992. Efforts to address these issues have recently gained considerable momentum with the formulation of the National Strategy for Maternal Health.3 This strategy emphasizes the provision of emergency obstetric care, and is predicated on the "three delays" framework of factors that affect safe motherhood service utilization and outcomes: delays in making the decision to seek care, in reaching a medical facility and in receiving adequate treatment or management at the facility.4

Evaluation of the overall progress in improving maternal health in Bangladesh (and elsewhere) has been hampered by the absence of suitable data. Facility-based assessments provide a picture of the availability, functioning and caseloads of maternal health facilities, but only for the select minority of women who utilize such services.5 In contrast, population-based surveys, such as the Demographic and Health Surveys (DHS), provide useful but generally limited data on basic aspects of maternal health behavior related to antenatal and delivery care.6,7 To address these information gaps, in 2001 the Bangladesh Maternal Health Services and Maternal Mortality Survey (BMMS) collected national data on the levels and causes of maternal and nonmaternal mortality; women's experience with and perceived severity of maternal complications; utilization of services for antenatal, delivery, postpartum and emergency obstetric care; delays in seeking emergency obstetric care; and indicators of the use of maternal health services.

As the first national population-based survey of its kind in Bangladesh, the BMMS provides a unique picture of maternal health and care-seeking behavior in a low-resource setting. A clear strength of the BMMS is its multipurpose nature, generating nationally representative data on maternal mortality levels,8,9 maternal health-seeking behavior and women's understanding of, experience with and resolution of specific maternal complications. A second unique feature of the BMMS lies in its ability to provide a population-based perspective on emergency obstetric care needs and related behavior—including the perceived occurrence and severity of obstetric complications and associated treatment-seeking behavior. Population-based studies complement approaches that use facility-based process indicators—especially in settings where most deliveries take place outside of health facilities—by including respondents who do not seek treatment at facilities.10 A final strength of the BMMS is that it identifies and quantifies the specific delays in seeking emergency obstetric care from a sufficiently large sample to permit disaggregation by types of complications.

A potential limitation of the BMMS and other retrospective studies is their reliance on women's self-reports regarding the occurrence and severity of obstetric complications; most studies have shown that self-reports may have low correspondence with medically diagnosed morbidity.11–13 Thus, while the BMMS estimates of obstetric complications and morbidity should be treated with caution, such information is critical to assessing and understanding women's treatment-seeking behavior, which depends primarily on perceptions of these complications and their associated severity.10,14 The value of future surveys similar to the BMMS would be enhanced by the concomitant collection of maternal health facility data in surrounding areas, and by more detailed data on maternal health care costs and women's perceptions of the access to and the quality of such health services.

METHODS AND DATA

The BMMS was conducted in all six divisions of Bangladesh using a two-stage sampling approach. Wards and unions served as the primary sampling units in urban and rural areas, respectively. In the first stage, 808 primary sampling units were chosen (the probability being proportional to their size), from which 1,616 secondary sampling clusters were systematically selected. All ever-married women aged 13–49 were eligible; a total of 104,323 households and 103,796 eligible women were interviewed, representing a response rate of 97%.8 This sample size was calculated to yield accurate national estimates of maternal mortality rates.

The BMMS consisted of three questionnaires. The household questionnaire collected information on selected social and demographic characteristics of all listed persons, household characteristics and deaths to household members during the three-year period prior to the survey. The verbal autopsy questionnaire was administered to the most knowledgeable family member to obtain information on all deaths to females aged 13–49 reported during the same three-year period. This interview consisted of both structured and open-ended questions on cause of death. All ever-married women aged 13–49 were administered the woman's questionnaire, which asked about, among other items, their socioeconomic and demographic characteristics, reproductive history and knowledge of specific life-threatening obstetric complications. For all live births and stillbirths that had occurred during the three years preceding the survey (N=40,657),* women were asked about antenatal, delivery and postdelivery care, as well as their experience of maternal health complications; associated treatment- seeking behavior. A team of medical and demographic experts compiled a list of Bangla terms used to describe specific complications; a summary of obstetric complications is presented in Table 1. The 7,613 women who reported that they were pregnant at the time of the BMMS were asked whether they had discussed or decided who would assist them during delivery.

When more than one complication perceived as life threatening was reported, only the most recent was considered (N=18,117). When both life-threatening and non-life-threatening complications were reported, the most recent life-threatening complication was considered as the reference complication. For women who reported no life-threatening complications, treatment-seeking behavior for the last non-life-threatening complication was examined (N=6,501). This selection process resulted in an overrepresentation of maternal complications that were perceived as life threatening in the analysis of treatment-seeking behavior, and for this reason, the data for life-threatening and non-life-threatening complications have been estimated separately.

The BMMS was implemented under the authority of the Bangladesh National Institute of Population Research and Training, and was funded by the U.S. Agency for International Development, Bangladesh. Data collection was conducted by two Bangladeshi organizations—Associates for Community and Population Research, and Mitra and Associates—both of which possessed extensive experience in large-scale demographic and health research. Each agency was responsible for data collection in three of the six divisions of Bangladesh. ORC Macro provided technical assistance and support, and the International Centre for Diarrhoeal Disease Research, Bangladesh, and Johns Hopkins University assisted with questionnaire design and data analysis. The questionnaire was pretested in November 2000, and revisions were made to both the original questionnaire and the Bangla translations. Data collection was carried out by 50 interview teams, with each team consisting of four female interviewers, a female field editor who reviewed completed questionnaires and a male supervisor. Independent quality control teams from both data-collecting organizations monitored fieldwork quality throughout the collection period, from January to June 2001.

RESULTS

Antenatal Care

Women reported having received antenatal care from a provider for 48% of all live births and stillbirths in the three years preceding the survey. In more than 85% of these cases, care was obtained from medically trained providers. Comparison with the two previous Bangladesh Demographic and Health Surveys indicates a steady increase in the proportion of women who sought antenatal care from a qualified provider—from 30% in the 1996–1997 DHS to 35% in the 1999–2000 DHS and 41% in the 2001 BMMS.6,15

The World Health Organization (WHO) recommends a minimum of four antenatal care visits, with the initial visit occurring by the fourth month of pregnancy. In our study, the number of antenatal visits to any provider was low— in only 12% of recent pregnancy outcomes had women made the recommended four antenatal visits. Moreover, the median time of first visit was at 5.4 months of pregnancy, and in only one in seven cases did women initiate ante- natal care during the first trimester. Although 84% of women reported having received advice on diet, only 54% were told where to go in the event of maternal complications and just 45% were informed about the danger signs of complications.

Delivery and Postdelivery Care

Planning for delivery is still not commonplace among pregnant Bangladeshis. Almost two-thirds of the 7,613 currently pregnant women had neither discussed nor made a decision regarding assistance at delivery; even among those in their third trimester, more than half had not reached a decision. Among those who had made a decision, 83% intended to have a medically untrained person assist in delivery.

The BMMS data show that most deliveries still occur at home and are attended by medically unskilled birth attendants. Among all live births and stillbirths in the three years preceding the survey, 91% took place outside of a health facility. Six percent of facility-based deliveries occurred at a government hospital, upazila (administrative unit) health complex, or maternal and child welfare center, and 3% at a private hospital or clinic. Only 12% of deliveries were attended by a medically trained provider. Comparison of BMMS 2001 data with 1993–1994 DHS data16reveals little change in the proportion of births assisted by a medically trained provider, in either urban or rural areas.

The BMMS found that the proportion of women reporting a cesarean delivery was 3%, part of a modest upward trend (from 2% in the 1999–2000 DHS) toward the WHO-recommended minimum of 5% of all deliveries. Yet this procedure is largely limited to women who reside in urban areas and in households of higher socioeconomic status.8 Other medical procedures at delivery are also quite infrequent: Two percent reported forceps delivery and 1% reported having received a blood transfusion at their most recent delivery. Postnatal care for mothers and infants within 42 days of delivery was uncommon, with 11% and 15% of mothers, respectively, reporting having obtained a postnatal checkup for themselves or for their baby from a medically trained provider.

Obstetric Complications

•Knowledge of complications.

Overall, women's awareness of life-threatening complications during pregnancy, delivery and the postpartum period was quite low, although it varied considerably according to type of complication (Figure 1, page 76). Fifty-six percent of women cited tetanus and 49% cited prolonged or obstructed labor as potentially life-threatening conditions, while smaller proportions mentioned retained placenta (38%) or malpresentation of the fetus (25%). Although convulsions and excessive bleeding account for more than half of all maternal deaths in Bangladesh,8 only 26% and 18% of women, respectively, cited these complications. In all, 89% of surveyed women were able to name at least one obstetric complication, and 42% could name three or more (not shown).

•Prevalence of complications.

The occurrence of at least one complication was reported in six of 10 live births and stillbirths. In 46% of these pregnancy outcomes, one or more complications occurred during gestation; in 35%, during delivery; and in 24%, after delivery (Table 2, page 77). Convulsions and excessive bleeding, the two most important causes of maternal death,8 were reported in 5% and 13% of cases, respectively. As would be expected, the prevalence of specific complications varied by stage of pregnancy. The most commonly reported complication during pregnancy was preeclampsia (39%); the most frequently cited complications during delivery were malpresentation or prolonged or obstructed labor (22%) and preeclampsia (11%). For the period immediately following delivery, excessive bleeding (10%) and preeclampsia (8%) were the complications cited most often.

Overall, women considered almost half of all the complications they reported to be potentially life threatening (Figure 2). The complications most likely to be perceived as life threatening were retained placenta (75%), prolonged or obstructed labor (70%), convulsions (57%) and excessive bleeding (55%).

•Treatment-seeking behavior.

In 62% of live births and stillbirths with a complication that was perceived as life threatening, women sought some form of treatment (Table 3). Treatment was most likely to have been sought by women with convulsions (77%) and those with high fever and foul-smelling vaginal discharge (75%), followed by those with excessive bleeding (65%), malpresentation or prolonged or obstructed labor (57%), and retained placenta (39%). For all types of complications, levels of treatment seeking were consistently lower when they were not perceived as life threatening.

Only one in three women who perceived their condition as life threatening, however, received care from a qualified provider—24% went outside of their home to seek care (19% from facility-based providers and 5% from providers not based at a facility) and 8% received services from providers who came to their home (Figure 3). In contrast, 29% sought care from an unqualified provider either at home or outside the home (24% vs. 5%).§ Overall, almost equivalent proportions sought treatment within or outside the home (32% vs. 30%). Most home-based treatment, however, was from unqualified providers rather than qualified providers (24% vs. 8%). As might be expected, the proportion of non-life-threatening complications for which treatment was sought from a qualified provider was even lower (22%—not shown).

Among the reasons women gave for not seeking treatment for obstetric complications perceived as life threatening, the most prominent was cost-related considerations, cited by 44%. The second most common reason was the perception that treatment was not necessary or that the condition was not serious (39%). One possible explanation for this apparent discrepancy is that different family members were involved in evaluating the seriousness of the complication; further, in-depth research on this issue is clearly warranted. Other service-related barriers to seeking care included transportation and access issues (12%), family opposition (12%) and concerns related to service quality (6%; women could give multiple reasons).

•Delays in seeking treatment.

An innovative feature of the 2001 BMMS was its attempt to assess and quantify delays in seeking treatment for obstetric complications, about which relatively little is known.4,17 Respondents reporting one or more complications perceived as life threatening were asked about four possible delays associated with the reference complication: the delay in deciding to seek care, which can be subdivided into the delay in recognizing the complication and the delay between recognition and deciding to seek treatment; travel time to the treatment facility; and waiting time at the facility. There were 3,614 reference cases of convulsions or excessive bleeding, two conditions whose effective resolution depends on timely intervention and treatment.

Sixty-one percent of women who had convulsions or excessive bleeding recognized their condition within six hours of onset, whereas 29% took six or more hours; the other 10% were unsure of the time elapsed (Figure 4). Half of the women decided to seek treatment in less than six hours; 18% took six or more hours to decide, and 32% sought no treatment. Twenty percent of women with these complications sought treatment outside the home, with three-fourths of these reporting a travel time of less than an hour. Forty-seven percent sought treatment at home. Nearly nine in 10 of those who sought facility-based treatment were seen within an hour of reaching the facility. These findings may reflect self-selection among women who sought care from providers or facilities.

•Disparities in maternal care-seeking behavior.

Socioeconomic data collected in the BMMS allow the calculation of a household wealth index and the creation of household wealth quintiles.18 There were wide disparities in maternal care-seeking behavior across the wealth quintiles, with use of health care increasing with rising socioeconomic status (Table 4, page 80). Among rural women, 22% of those in the lowest wealth quintile and 69% in the highest quintile reported having sought antenatal care from a medically trained provider; women in the highest quintile had odds more than three times as high as those in the lowest quintile of seeking such care (crude odds ratio, 3.2). Similarly, 19% of rural women in the lowest wealth quintile and 60% in the highest quintile sought treatment for a life-threatening complication; the wealthiest women had elevated odds of doing so (3.2). Even greater disparities were seen for two other care-seeking behaviors: Those in the highest wealth quintile had eight times the odds of obtaining skilled delivery care (28% vs. 3%) and nine times the odds of delivering at a health facility (21% vs. 2%) as did those in the lowest quintile.

Although levels of maternal health care use were somewhat higher among urban women, similar differentials across wealth quintiles were evident. Even among the wealthiest urban and rural women, many did not obtain skilled delivery care, did not deliver at a facility or did not seek skilled treatment for life-threatening complications.

DISCUSSION

Over the last 15 years, the government of Bangladesh has engaged in a major effort to improve maternal health by developing and upgrading a network of essential obstetric care facilities at the upazila and district levels. All 59 district hospitals and almost all of the 64 maternal and child welfare centers have been upgraded to provide comprehensive emergency obstetric care, but progress in upgrading upazila-level facilities has been much slower.

The BMMS and recent DHS data indicate that although the proportion of Bangladeshi women seeking antenatal care has been steadily increasing, it remains quite low. Bangladesh continues to rank near the bottom of the list of 66 countries for which DHS data on antenatal care are available. The proportion of deliveries taking place in a medical facility or assisted by a skilled medical attendant has been fairly stable over the past decade, with 90% of all deliveries occurring at home, assisted by unskilled birth attendants or family members and preceded by little planning.6,8,15,16

Women's knowledge of life-threatening obstetric complications is mixed. Among women of reproductive age interviewed in the BMMS, unprompted knowledge of specific life-threatening complications was low, although almost all knew of at least one such complication and a significant minority could name three or more. Among women who had had a recent live birth or stillbirth with a complication they perceived as life threatening, more than six in 10 had sought some form of medical treatment, suggesting a much higher level of awareness of the need for care. However, only a third sought treatment from a medically qualified provider, and the other two-thirds either did not seek care or sought care from an unqualified provider.

Of women who reported the highly time-sensitive and dangerous complications of convulsions or excessive bleeding, a significant proportion took six or more hours to decide to seek treatment, although neither travel time to reach the facility nor delays in receiving treatment once at the facility were major concerns. However, we view these findings as largely a reflection of self-selection in terms of proximity to a facility or ability to pay for treatment among the minority of women who actually made it to a facility; the major programmatic concerns with treatment seeking instead appear to be the large proportions of women who either did not seek any form of treatment for such complications or sought treatment from medically unqualified providers.

Cost concerns are an important barrier to seeking emergency obstetric care in Bangladesh. Data from the BMMS identified cost as the most commonly cited reason for not seeking treatment; cost also emerged as a central concern in a separate analysis of BMMS verbal autopsy data of factors contributing to maternal death.19 The marked socio-economic disparities in all aspects of maternity care— including emergency obstetric care—observed in this and other studies from Bangladesh20 provide further evidence of the importance of financial considerations and barriers. Although public sector maternity care services are officially free in Bangladesh, some studies have documented the myriad hidden costs to patients associated with "free" obstetric care (such as hospital fees, corruption and medical supplies), which frequently result in an untenable financial burden to families.21,22 Policy priorities for improving maternal care should include testing approaches to lessen the financial burden to poorer families, both through reforms to reduce corruption and out-of-pocket costs to poor clients and through social safety nets, such as health insurance or loans to cover maternity-related costs. Other less frequently cited programmatic concerns are issues related to service access and quality.

In contrast to this somewhat pessimistic overview of maternal health care in Bangladesh, BMMS data reveal that the estimated maternal mortality ratio was 320 per 100,000 live births for the 1998–2000 period, representing a downward trend of 22% over the preceding 12 years,9 and a level almost half that reported from local studies undertaken during the 1970s and 1980s.23–26 A partial explanation for this decline is that even though overall levels of facility-based deliveries and treatment seeking for life-threatening conditions from qualified providers remain low in Bangladesh, women who do seek effective treatment are those with the most severe life-threatening complications. Support for this explanation comes from two surveys conducted in the 1990s, which concluded that met need for emergency obstetric care among Bangladeshi women had increased significantly during this period.27,28

Another explanation for this apparent decrease in maternal mortality may lie in the sharp decline in fertility levels that has taken place within Bangladesh over the past three decades.6 This decline may have reduced childbearing at higher parities and maternal ages and lowered levels of unwanted pregnancy and abortion.29 Only 5% of maternal deaths in the BMMS data were attributable to abortion, a marked decline from the estimated 18% of maternal deaths reported in earlier studies.30,31 More research on the decline in maternal mortality levels in Bangladesh and its underlying factors is clearly warranted.