Investigating the role of health care at birth on inequalities in neonatal survival: evidence from Bangladesh
© Neal and Matthews; licensee BioMed Central Ltd. 2013
Received: 27 June 2012
Accepted: 17 February 2013
Published: 6 March 2013
In countries such as Bangladesh many women may only seek skilled care at birth when complications become evident. This often results in higher neonatal mortality for women who give birth in institutions than for those that give birth at home. However, we hypothesise that this apparent excess mortality is concentrated among less advantaged women. The aim of this paper is to examine the association between place of birth and neonatal mortality in Bangladesh, and how this varies by socio-economic status.
The study is based on pooled data from four Bangladesh Demographic and Household Surveys, and uses descriptive analysis and binomial multivariate logistic regression. It uses regression models stratified for place of delivery to examine the impact of socio-economic status and place of residence on neonatal mortality.
Poor women from rural areas and those with no education who gave birth in institutions had much worse outcomes than those who gave birth at home. There is no difference for more wealthy women. There is a much stronger socio-economic gradient in neonatal mortality for women who gave birth in institutions than those who delivered at home.
In Bangladesh babies from lower socio-economic groups and particularly those in rural areas have very poor outcomes if born in a facility. This suggests poorer, rural and less educated women are failing to obtain the timely access to quality maternal health care services needed to improve newborn outcomes.
KeywordsNeonatal Inequalities Developing countries Maternal health services Poverty Access to health care
It is well known that the newborns of women with higher socio-economic status have improved rates of neonatal survival compared with their lower status counterparts. However, as institutional births become more common it could be argued that differing patterns of health service use between rich and poor women will exacerbate and widen the mortality gap. Poorer women are less likely to go to a facility to give birth, using services only as a last resort once complications arise. These differences between rich and poor could also be mirrored by differences between educated and non-educated women, and across other socio-economic divides.
The aim of this paper is to examine how neonatal outcomes from institutional deliveries vary for women from different socio-economic groups, and ascertain whether wealth or education makes a greater contribution to mortality gaps for institutional than home births. Our approach is to use both descriptive statistics and stratified multivariate logistic regression models to examine how socio-economic status is associated with neonatal outcomes for women giving birth at home and in institutions.
Inequities in neonatal mortality and use of health services: an explanatory model
Socio-economic inequalities in neonatal mortality are seen throughout the world, and indeed even in developed countries such as the UK newborns born into the poorest families are more likely to die than those from more wealthy backgrounds. Socio-economic status is likely to influence neonatal survival through a range of proximate determinants such as maternal nutrition, birth spacing and family practices. However, there is evidence to suggest that at a national level a much greater proportion of the variance in neonatal mortality can be explained by access to health care rather than contextual factors such as income, female literacy, water and sanitation. This may suggest that for neonates the most important pathway through which socio-economic inequalities affect outcomes is differential access to, and use of health care by rich and poor, or by educated and uneducated.
Care at birth is seen as the cornerstone of efforts to reduce neonatal mortality. However, in countries such as Bangladesh, where skilled care at birth is rare, those who receive it are much more likely to have experienced complications that are strongly associated with the probability of a neonatal death. This results in higher neonatal mortality for women giving birth in institutions in a range of settings, but this apparent disadvantage may mask importance differences in groups of women that access services. Studies have consistently demonstrated that wealthy or more educated women are much more likely to deliver in an institution or with skilled care[5–7], which may suggest that a larger proportion of these urban, richer or more educated women will have planned their birth at an institution rather than waiting for a complication to arise.
Neonatal health within the context of Bangladesh
Our analysis is based on data collected from four Demographic and Household surveys (DHS) collected in 1996/7, 1999/2000, 2004 and 2007 in Bangladesh which were pooled to produce a large sample size. These are nationally representative household surveys providing a combined sample size of 26,079 children born to 19,548 mothers within the previous five years prior to the three surveys. Full birth histories were collected from women aged between 15 and 49 years in sampled households, and data are comparable over time.
In our analysis we use asset quintiles disaggregated for urban/rural residence as a proxy for wealth. DHS normally creates five asset quintiles for the whole population whether rural or urban. The disadvantage of this is that some of the index components do not effectively capture the different patterns of rural and urban asset ownership, and Houweling et al. recommends that efforts are made to more accurately quantify assets pertinent to rural communities, e.g. land ownership and livestock. In response to this issue, asset scores and quintiles were recalculated separately for urban and rural populations. The same methodology and variables were used as in the original DHS: a principal component analysis (PCA) was run on the selected variables for urban and rural dwellers separately, and an individual score estimated for each case, which became the asset score. Quintiles were then calculated separately for urban and rural dwellers, creating a total of 10 categories. However sample sizes for institutional deliveries among the poorer quintiles were very low, so quintiles 1–4 were amalgamateda to create four groupings: Rural quintile 5 (richest) and rural quintiles 1–4, urban quintile 5 (richest) and urban quintile 1–4.
Initially crosstabulations and Pearson’s Chi Squared tests were used to examine the associations between NMR and urban/rural-quintiles and place of birth. While such descriptive approaches are valuable in identifying general patterns of socio-economic inequalities, interpretation is difficult as many of the factors affecting neonatal mortality are closely correlated. Binomial multivariate logistic regression was therefore carried out to predict neonatal mortality in each of the categories net of other effects. We used a series of six stratified models to examine the impact of wealth/urban rural residence and education on neonatal outcomes at home or in institutions. The models also controlled for year of survey and a number of factors known to be associated with neonatal mortality: sex, multiple birth, birth order, parity and antenatal care. Several other factors were originally included (e.g. maternal age) but were removed as they were not significant in any of the models and did not change the coefficient of interest. The models with level of education as explanatory variables also had urban/rural as a control variable.
Our outcome is neonatal mortality, which is defined as death of a live-born infant before the 28th day of life. Originally it had been hoped to also use early neonatal mortality (as a more direct outcome of delivery care) as an outcome variable, but small sample sizes and possible data problems made this impossible. As there may be significant correlations for many unobservable characteristics between births to the same mother the data are clustered using the mother’s identity number in order to adjust the standard errors. We did not disaggregate for the different types of facility (e.g. public, private, NGO) due to small sample size, and exploratory analysis suggested it made little difference.
Coverage for home and institutional deliveries by socio-economic grouping (numbers in brackets)
Urban quintiles 1-4
Urban quintiles 5
Rural quintiles 1-4
Rural quintiles 5
Descriptive analysis of mortality by place of birth and socio-economic grouping
Neonatal mortality rate (Confidence intervals in brackets)
Wealth quintile 5 (richest)
Wealth quintile 1–4 (all but richest)
Neonatal mortality rate (Confidence intervals in brackets)
Changes in neonatal mortality rate between surveys, disaggregated by institutional and home births
Year of survey
NMR: Total sample
NMR: Institutional births only
NMR: Home deliveries only
Results of multivariate analysis
The basic model was built on neonatal mortality as a dichotomous outcome. Six stratified models were developed:
Model 1: Institutional delivery in urban areas, with urban wealth quintiles 1–4 and 5 as explanatory variables
Model 2: Institutional delivery in rural areas, with rural wealth quintiles 1–4 and 5 as explanatory variables
Model 3: Home delivery in urban areas, with urban wealth quintiles 1–4 and 5 as explanatory variables
Model 4: Home delivery in rural areas, with rural wealth quintiles 1–4 and 5 as explanatory variables
Model 5: Institutional delivery with mother’s highest level of education as explanatory variables
Model 6: Home delivery with mother’s highest level of education as explanatory variables
Regression models identifying factors that predict neonatal mortality among institutional and home births disaggregated by urban and rural samples with wealth quintiles as explanatory variable b
Odds ratios for neonatal mortality for institutional deliveries
Model 1: Rural only
Model 2: Urban only
Year of survey (reference 2007)
Wealth quintile (reference richest quintile 5)
Sample size: 1046
Sample size: 1733
Odds ratios for neonatal mortality for home deliveries
Model 3: Rural
Model 4: Urban
Year of survey (reference 2007)
Wealth quintile (reference richest quintile 5)
Sample size: 18206
Sample size: 4943
Regression models to predict neonatal mortality for institutional and home births with mother’s education as explanatory variable c
Odds ratios for neonatal mortality for home and institutional deliveries
Model 5: Institutional birth
Model 6: Home birth
Year of survey (reference 2007)
Maternal education (reference higher/further)
Sample size: 2803
Sample size: 23149
Institutional births and poor outcomes for the poor
The markedly larger socioeconomic differentials in neonatal mortality for those born in hospitals suggests strongly that provision of care at birth is particularly failing to meet the needs of disadvantaged women and their newborns. Both the descriptive and regression results clearly demonstrate that wealth has a markedly greater impact on neonatal outcomes for those giving birth in institutions compared to those giving birth at home. This suggests poor women are either arriving too late, or the care they receive is inadequate. While the descriptive analysis for education shows a very similar pattern, evidence from the multivariate analysis is less compelling: while not significant, there is a marked increase in OR for neonatal death for no education, primary and secondary education for women giving birth at home compared to those with higher education. The lack of significance may well be as a result of small sample size, and the findings may reflect the fact that the reference group of those women who have received further education is a particularly small and elite group with very low overall neonatal mortality.
The difference in neonatal outcomes from institutions between the urban and rural poor is also marked. This could be either because the characteristics of the two groups are different, or because access to services is more difficult in rural areas. In many countries rural poverty may be deeper and more chronic than in urban areas, and the population may be less educated. There is some evidence that the rural population in this study are indeed different in character: we cannot directly compare wealth between urban and rural groups using the asset index, but if we examine education, over half (53%) of the rural quintiles 1–4 had no education compared to 34% of urban quintiles 1–4. However, another plausible reason could be the increased journey time when seeking care for rural women, which could detrimentally impact on outcome. The Bangladesh Maternal Health Services and Maternal Mortality Survey 2001 found that only 26% of rural women with complications reported travelling less than 30 minutes to reach a provider or facility compared with 59% of urban respondents. Around 8% of women reported travelling for more than 3 hours, which obviously is likely to have negative consequences.
While utilisation data are often disaggregated, data on institutional outcomes are not routinely analysed – indeed the idea of inequity in outcomes from services is not well developed within the literature. This work suggests that such measures could be important, and provide a much more comprehensive picture than utilisation data alone. For instance, an apparent increase in utilisation among the poorest might mask the fact that many of these women are only arriving at hospital once complications are well advanced, thus reducing their chance of positive outcomes. Ideally such analysis should include attention to fresh stillbirths, which are often used as an indicator of care at birth. However, very limited data are available: DHS infrequently collect data on stillbirths, and quality is questionable, along with a further limitation that fresh and macerated stillbirths are not differentiated. Indeed, it is worth considering that the extremely negative outcomes experienced by poor women would be even more concerning if stillbirths could be included, as poor or delayed intrapartum care may often result in the foetus dying during labour, so will not be included as a neonatal death.
The fact that due to sample size quintiles 1–4 were grouped together probably also results in an underestimation of the disadvantage faced by the poorest in terms of neonatal outcomes. Examining the difference in mortality between the top and bottom socio-economic groupings is a commonly used method of examining inequality, but has the limitation of failing to ascertain the distribution of mortality across the population. Another measure that could address this would be calculation of concentration curves or indices for the different groups as this takes into account mortality for the whole range of wealth. Further studies using this technique may be valuable.
Analysis over time of perinatal institutional outcomes can provide valuable evidence on quality of care. It would be assumed that as uptake of skilled care at birth increases, outcomes would improve as more and more women with uncomplicated pregnancies make an active choice to give birth in hospital. However, there are suggestions that in some cases increased demand actually has a negative impact on quality, as staff and resources are placed under greater pressure. Monitoring of institutional mortality rates, particularly among the poorest could give some indication of any unintended negative consequences resulting from efforts to increase uptake of services.
Why are institutional outcomes so poor for women in Bangladesh?
While this study were able to identify groups of women who experienced particularly poor outcomes, our descriptive analysis is unable to demonstrate that any group has significantly reduced mortality as a result of utilising institutional care at birth when compared with babies born at home with similar socioeconomic characteristics. In an environment such as Bangladesh many mothers in the higher socio-economic groups may use hospital care at birth only in response to complications. While their outcomes may be better because of faster recognition of, and response to, complications than their poorer counterparts, their underlying condition will still affect the neonatal outcome.
It is also possible that if quality of care is relatively poor for all women (not just the poorest), the potential benefits from institutional birth may be reduced or in some cases even negated. There are ongoing concerns in connection with poor quality of maternal health care covering the entire period reflected in this study. An assessment of the Bangladesh Maternal Mortality programme in 1997 found patients’ wellbeing placed at risk by a lack of drugs and sterilised equipment and poor hygiene practices. Dysfunctional staff configurations, inadequately skilled staff and poor infrastructure were also identified as barriers to quality care. More recently Chowdhury et al. discovered constraints such as inadequate drug supplies, insufficient specialist staff, human resource absenteeism and long waiting times that all contributed to poor quality care. While data is again limited, there is also growing evidence that the potential for improved outcomes from an increase in facility births in developing countries may be reduced by the high incidence of hospital-acquired (nosocomial) infections in newborns.
As this study covers a considerable period of time, the question arises as to whether the situation has improved between 1997 and 2007. While Bangladesh has made some progress since the data used in this study was collected, the proportion of women receiving skilled care in Bangladesh is still amongst the lowest in the world and it could be questioned how much impact such limited progress would have on improving neonatal outcomes. Despite this, the limited data we present seems to suggest that most gains in reducing neonatal mortality have been in hospital births, and the most recent survey (2007) suggests that at the national level there are now similar levels of neonatal mortality in institutions and at home. This may imply that either quality has improved, or women with complications are recognising and seeking care more promptly. These changes can be considered within the rather surprising context of strong reductions in maternal mortality during this period. As the increase in skilled attendance has been so modest, it has been suggested that this reduction could be partly driven by improved knowledge of, and health seeking for complications, and increased access to emergency obstetric care, which would support our premise. More research is needed to understand the factors driving reductions in maternal and neonatal mortality in Bangladesh, and how their impact can be maximised.
Efforts to improve timely uptake of services for all women (not just those with complications) is extremely challenging in contexts such as Bangladesh, and will require a number of both demand- and supply-side interventions to improve access to and quality of services and also increase understanding of the need for skilled attendance at all births. In addition, services must be accessible to all sectors of the community, which will require innovative and far-reaching solutions to solve the problems of both formal and informal health care costs.
This study highlights that in Bangladesh babies from lower socio-economic groups have extremely poor outcomes if born in a facility. The mortality difference between richer and poorer, or educated versus uneducated is much more marked among facility births compared with the equivalent differential among home births, suggesting that services are particularly failing to meet the needs of poorer and less educated women. This could be because poor and uneducated women seek care later, or the care they receive is poorer quality.
Further research in different contexts (e.g. other countries with a higher proportion of facility births) may be able to identify improved outcomes for some groups of women. Greater disaggregation of institutional outcomes by socio-economic groups would provide more comprehensive evidence on whether services are reaching the poorest than utilisation data alone. As Bangladesh increases access to skilled attendance further analysis could provide an important insight into the country’s success in providing pro-poor care. Indeed, as countries move nearer to the Millennium Development Goals the focus should shift from overall mortality rates to the key differences between mortality rates for different socioeconomic groups. As institutionalisation of births increases, more focus should be placed on ensuring services promote rather than reduce equity of outcomes.
aIt was also noted that the asset scores for quintiles 1–4 in both urban and rural areas showed little real difference in assets between these poorest four quintiles.
This study was funded by the Economic and Social Research Council (ESRC). The authors would also like to acknowledge comments and advice given by Professor John Micklewright for comments made on earlier drafts.
- Lawn J, Cousens S, Zupan J, Lancet Neonatal Survival Steering Team: 4 Million neonatal deaths: when? where? why?. Lancet. 2005, 365 (9462): 9-16.View ArticleGoogle Scholar
- UK Office of National Statistics: Infant and perinatal mortality in England and Wales by social and biological factors 2009. [online]. 2011, Available athttp://www.ons.gov.uk/ons/rel/child-health/infant-and-perinatal-mortality-in-england-and-wales-by-social-and-biological-factors/2009/index.html,Google Scholar
- World Health Organisation (WHO): The World Health Report 2005: Make every mother and child count. 2005, Geneva, WHOGoogle Scholar
- Govindasamy P, Stewart M, Rutstein S, Boerma T, Sommerfelt A: High-Risk Births and Maternity Care. Demographic and Health Surveys Comparative Studies No 8. 1993, Columbia, Maryland, USA: Macro International IncGoogle Scholar
- Saha S, Kabir M: Factors influencing women receiving safe child delivery care in Bangladesh. World Health and Population. 2006, 8 (4): 1-6.View ArticleGoogle Scholar
- Kabi M: Safe-delivery practices in rural Bangladesh and its associated factors: evidence from Bangladesh demographic and health survey. East Afr J of Public Health. 2007, 4 (2): 67-72.Google Scholar
- Collin S, Anwar I, Ronsmans C: A decade of inequality in maternity care antenatal care, professional attendance at delivery, and caesarean section in Bangladesh (1991–2004). Int J Equity Health. 2007, 6 (9):
- National Institute of Population Research and Training (NIPORT), ORC Macro, Johns Hopkins University and ICDDR,B: Bangladesh Maternal Health Services and Maternal Mortality Survey 2001. 2003, Dhaka, Bangladesh and Calverton, Maryland (USA): National Institute of Population Research and Training (NIPORT), ORC Macro, Johns Hopkins University and ICDDR,BGoogle Scholar
- Pitchforth E, van Teijlingen E, Graham W, Dixon-Woods M, Chowdhury M: Getting women to hospital is not enough: a qualitative study of access to emergency obstetric care in Bangladesh. Qual Saf Health Care. 2006, 15: 214-219. 10.1136/qshc.2005.017285.PubMed CentralView ArticlePubMedGoogle Scholar
- Rutstein S: Infant and Child Mortality: Levels and Trends and Demographic Differentials. World Fertility Survey Comparative Studies, no 2. 1983, Voorburg: International Statistical InstituteGoogle Scholar
- WHO: World Health Statistics. 2011, Geneva,http://www.who.int/whosis/whostat/2011/en/index.html,Google Scholar
- WHO: Success stories in developing countries. 2001, Geneva, Accessed March 2012. http://who.int/inf-new/mate3.thmGoogle Scholar
- Houweling T, Kunst A, Mackenbach J: Measuring health inequality among children in developing countries: does the choice of the indicator of economic status matter?. Int J Equity Health. 2003, 2 (8): 116-Google Scholar
- Bird K, McKay A, Shinyekwa I: Isolation and poverty: the relationship between spatially differentiated access to goods and services and poverty. 2010, CPRC Working Paper 162,http://www.odi.org.uk/publications/4534-spatial-poverty-traps-chronic-poverty-isolation-goods-services,Google Scholar
- Stanton C, Lawn JE, Rahman H, Wilczynska-Ketende K, Hill K: Stillbirth rates: delivering estimates in 190 countries. Lancet. 2006, 367 (9521): 1487-1494. 10.1016/S0140-6736(06)68586-3.View ArticlePubMedGoogle Scholar
- Iyengar SD, Iyengar K, Suhalka V, Agarwal K: Comparison of domiciliary and institutional delivery-care practices in rural Rajasthan, India. J Health Popul Nutr. 2009, 27 (2): 303-312.PubMed CentralPubMedGoogle Scholar
- Huque Z, Leppard M, Mavalandar D, Akhter H, Chowdury T: Safe Motherhood programmes in Bangladesh. Safe Motherhood Initiatives: critical issues. Edited by: Berer M, Ravindran S. 1999, London: Blackwell Science/Reproductive Health MattersGoogle Scholar
- Chowdhury S, Hossain SA, Halim A: Assessment of quality of care in maternal and newborn health services available in public health care facilities in Bangladesh. Bangladesh Med Res Counc Bull. 2009, 35 (2): 53-56.View ArticlePubMedGoogle Scholar
- Zaidi A, Huskins W, Thaver D, Bhutta Z, Abbas Z, Goldmann D: Hospital-acquired neonatal infections in developing countries. Lancet. 2005, 365 (9465): 1175-1188. 10.1016/S0140-6736(05)71881-X.View ArticlePubMedGoogle Scholar
- National Institute of Population Research and Training (NIPORT) MEASURE Evaluation, UNC-CH, USA, ICDDR,B: Bangladesh Maternal Mortality and Health Care Survey 2010. Preliminary results. 2011,http://transition.usaid.gov/bd/files/C6%20BMMS%20Survey%202010.pdf,Google Scholar
This article is published under license to BioMed Central Ltd. This is an Open Access article distributed under the terms of the Creative Commons Attribution License (http://creativecommons.org/licenses/by/2.0), which permits unrestricted use, distribution, and reproduction in any medium, provided the original work is properly cited.