February 2005

Maternal-Neonatal Health (MNH) and Poverty: Factors Beyond Care that Affect MNH Outcomes[1]

Thomas W. Merrick

World Bank Institute

Washington, DC

Table of Contents

Introduction

The Pathways Framework

Data on MNH and Poverty

Data and Measurement Issues

Regional and Cross-National Differences

Table 1: Comparison of 1995 and 2000 Regional and Global Totals

Table 2: 2001 Global and Regional Estimates of Neonatal Mortality

Poverty Linkages

Table 3: Attendance at Delivery by a Medically Trained Person

by Wealth Quintile

Table 4: Averages of Pathways Indicators for Countries Grouped by MMR Level

Linkages with Other Factors

Table 5: Pathways Indicators for Large Countries

Table 6: Pathways Indicators for Countries with Highest MMRs

Cross-National Analyses.

Table 7: Correlations between Pathways Variables and MMRs

Table 8: Averages of Pathways Indicators for Countries Grouped by Neonatal Mortality Rate

Review of Evidence on Pathways Variables

Other Reproductive Health Risk Factors

Table 9: Total Fertility Rates by Wealth Quintile and Region

Table 10: Contraceptive Prevalence by Wealth Quintile and Region

Table 11: Adolescent Fertility Rates by Wealth Quintile and Region

Table 12: Selected Findings on Other Reproductive Health Risks

Household and Community Factors

Health System Failures

Table 13: Selected Findings on Health System Issues

Other Sectors

Table 14: Selected Fndings on Transport Interventions

Table 15: Low Body-Mass in Reproductive-Age Women by Wealth Quintile and Region

Table 16: Child Malnutrition by Wealth Quintile and Region

Public Policy and Governance:

Table 17: Governance Indicators for Countries Grouped by MMR Level

Addressing Obstacles and Information Gaps

Policy and Program Actions

Strengthening the Evidence Base

Table 19: Estimates for Burden of Disease for sub-Saharan Africa, 2002

Table 20: Longitudinal Survey Programs

Conclusion

Annex 1: Maternal Mortality in India

Annex Table 1: MMRs and Other Indicators for Indian States

Annex 2: Country-Level Data Table

References

Introduction

Every year more than half a million maternal deaths and around four million perinatal deaths occur in low and middle-income countries, mostly among the poorest groups within these countries. There is an even larger toll of morbidities (more than 8 million each year according to Koblinsky et al., 1993) resulting from non-fatal complications of delivery. Most of these deaths and disabilities are preventable, and the interventions required to prevent them are known. The sad reality is that in many instances these interventions are either not available to poor women or so poor in quality that they are ineffective.

Other reports in this series address the “what” and “how to” of health care interventions to prevent maternal and neonatal mortality and morbidity. This paper focuses on the obstacles that prevent poor women from benefiting from the knowledge and technical expertise that is available, and on the factors beyond care that shape maternal and neonatal health (MNH) outcomes. It employs an adapted version of the “Pathways” framework from the World Bank’s guidelines for Poverty Reduction Strategy Papers to link factors at various levels—from individuals, households and communities to government policies in health and other sectors—that directly or indirectly affect MNH outcomes.

The paper begins with an explanation of what the Pathways framework is and how it will be used to guide this discussion. That is followed by a brief discussion of measurement issues and by a review of cross-national evidence about linkages between factors identified in the Pathways framework and MNH outcomes. More detailed country-level evidence on factors at each of the levels (households and communities, health and other sectors, policy) will then be reviewed, leading to recommendations for actions that could be taken to address obstacles at each of these levels as well as research needed to strengthen the evidence base to assess the impact of these actions.

In adapting the Pathways framework to MNH outcomes, the paper draws on other frameworks that have focused specifically on contextual factors affecting maternal health outcomes, for example the one developed by the Prevention of Maternal Mortality collaborators at Columbia University and in West Africa (McCarthy and Maine, 1993; Thaddeus and Maine, 1994; McCarthy, 1997) as well as broader frameworks that address a range of health outcomes (for example, Hanson et al., 2003). These frameworks address the role of household and community-level variables on outcomes.[2]

The Pathways Framework

The chart below depicts a simple framework for assessing the impact of factors inside and outside the health system that influence health, including MNH, as shown in the left-hand column of the chart. These factors operate at three levels, shown in the other three columns of the chart: (1) households and communities, (2) the health system (including health care, health finance, health promotion) and sectors other than health such as education, infrastructure (water and sanitation, transport and communications) that indirectly influence health outcomes, and (3) public policies and actions that affect health systems and outcomes directly (health reforms, for example) or indirectly (macro-economic policies).

Households and Communities: Good health is dependent not only on the provision of good medical services when required but also on healthy behavior. Healthy behavior means avoiding or minimizing risks (e.g. practicing family planning and safe sex) and requires knowledge about how to prevent disease and promote health and the ability to act on this information. Many health-promoting behaviors (for example dietary habits, sanitary practices, fertility regulation, childcare, and utilization of health services) take place within families. Such behaviors are often related to ‘household assets’ such as income, education, access to health services, roads and communication, membership in formal and informal support networks, as well as general knowledge and information. Health too is itself a household asset. Economists view these household-level determinants as demand-side factors, as opposed to the supply of health care (Ensor and Cooper, 2004).

Economic research has tended to treat all members of a family, or household, as a single unit, assuming that whatever benefits one member will benefit the entire household. This is clearly not the case in reality, as it is widely acknowledged that intra-household differences in gender and age may significantly affect how decisions are made and whether a decision is beneficial for all members (Case and Deaton, 2002). Thus an understanding of household decision-making is critical to an understanding of how policy decisions affect the welfare of families as a unit or of individual members within them.

Gender disparities in access to education, credit and political influence have considerable impact on how individual family members are valued and on the degree to which women as well as men have a voice in household decision-making. Recognition of the importance of individuals and households in producing good (or poor) reproductive health outcomes should lead policy makers to focus on the constraints faced by vulnerable households and vulnerable members within those households.

Household-level behaviors and risk factors are influenced and reinforced by conditions in the community (Tinker, 2000). Community factors include both the values and norms that shape household attitudes and behaviors and the physical and environmental conditions of the community, for example terrain and weather conditions that affect households’ capacity to produce better outcomes. Community factors that typically influence health outcomes are:

  • Gender norms, which are influenced by social and cultural values that shape the roles of and relationships between men and women;
  • Existence of effective community groups, social cohesion (sometimes called social capital) that support positive behaviors individuals and families or organize actions to improve health outcomes directly (community health insurance or pooling of resources to transport emergency cases) or indirectly (micro-credit programs);
  • Community access to public services (inside and outside the health sector); and
  • Environmental conditions (safe water, location—distance from a health facility, terrain, weather conditions).

Health System and Other Sectors: The supply of health care and health information are key determinants of maternal and neonatal health outcomes for the poor. Since these interventions are covered in detail in other papers in the series, this paper will address two potential supply-side obstacles that may prevent poor women from benefiting from these services. These are: (1) financial obstacles including fees and/or coverage of critical services in insurance benefit packages; and (2) the organizational and institutional obstacles to scaling up effective interventions so that poor women can access them.

Actions in other sectors also affect health outcomes, for example:

  • Education, either formal education or training that enhance earnings capacity of household members as well as their capacity for effective health-seeking behaviors;
  • Transport and infrastructure, for example the availability of services and the quality of roads that can affect travel time when a mother requires transport for management of an obstetric emergency;
  • Energy and communications, for example coverage in a cell-phone network so that help can be sought it an emergency;
  • Water and sanitation, important for avoiding infections; and
  • Nutrition.

Government Policies and Actions can affect both the health system and related factors in other sectors. Most countries have undertaken reform as a way to profoundly change the fundamentals of the health sector including changes in organization and accountability, revenue generation allocation and purchasing, as well as regulation. Government policies and actions in other sectors also affect health because of their influence on attitudes and behavior and the supply of related services such as education, transport, water and food security. Fees for schools and health services are examples of public policies that may have an indirect effect on MNH outcomes. Taxation may also be a strong influence on health, for example “sin taxes” on tobacco or alcohol.

Data on MNH and Poverty

Data and Measurement Issues

The paper utilizes cross-national estimates of maternal mortality from the WHO/UNICEF/UNFPA database prepared by AbouZahr and Wardlaw (2003). Estimates are provided for 172 countries and are derived from a variety of sources, including vital registration systems, direct and indirect estimating methods based on survey and census data, and statistical modeling for the 62 countries for which no national data on maternal mortality were available. This paper focuses on 142 countries in the World Bank's low and middle-income categories, and most of the 62 countries with estimated maternal mortality ratios (MMRs) are in those categories. The fact that statistical modeling was used to estimate MMRs creates some major limitations for the analysis of cross-country differences, because many of potential explanatory variables (fertility rates, GDP per capita, percentage of births assisted by a skilled attendant, and regional dummy variables) have been used to estimate the proportion of deaths that are considered “maternal” in country-level model life tables. The authors of the estimates emphasize that they should not be used for trend analysis and urge caution in cross-national comparisons for the reasons just stated.

Cross-national data on neonatal mortality rates (NMRs : deaths of liveborns during the first 28 days of life per 1000 live births) are from a compilation published in the State of the World’s Newborns 2001 (Save the Children, 2002). Data for other cross-national indicators (per-capita income, poverty, education, transport, governance, etc) are taken from the World Bank’s World Development Indicators 2003, UNDP’s 2003 Human Development Report, and other sources (see Annex 2).

In addition to cross-national comparisons and analysis, the paper will examine country-level tabulations of key indicators from the Demographic and Health Surveys that have been tabulated using a composite ‘household asset’ measure to show rich-poor differences in those indicators by wealth quintiles (Gwatkin et al, 2004). Additional evidence from country and topic-related studies will also be employed to fill out the picture of how factors beyond care may directly or indirectly impact on MNH outcomes for the poor.

Regional and Cross-National Differences

Global, regional and country-level estimates of maternal mortality (Table 1) show a clear connection between high maternal mortality ratios (MMRs) and poverty.

More than 99 percent of maternal deaths occur in developing regions, and more than 85 percent occur in the poorest countries of Sub-Saharan Africa and South Central Asia. Country-level estimates show that more than a quarter of those deaths occurred in India, and that several other poor countries in these two regions (Bangladesh, Ethiopia, the Democratic Republic of Congo, Nigeria, Pakistan, Tanzania) account for another quarter of them. The highest maternal mortality ratios are found in poor countries in Sub-Saharan Africa. With the exception of Afghanistan, all of the countries having maternal mortality ratios of 1000 or higher are found in Africa.

Table 1: Comparison of 1995 and 2000 Regional and Global Totals

Region / 2000 / 1995
Maternal Mortality Ratio / Maternal deaths
(000s) / Maternal Mortality Ratio / Maternal deaths
(000s)
WORLD TOTAL / 400 / 529,000 / 400 / 515,000
DEVELOPED REGIONS* / 20 / 2,500 / 21 / 2,800
Europe / 28 / 2.2 / 36 / 3.2
DEVELOPING REGIONS / 440 / 527,000 / 440 / 512,000
Africa / 830 / 251,000 / 1,000 / 273,000
Northern Africa / 130 / 4,600 / 200 / 7,200
Sub-Saharan Africa / 920 / 247,000 / 1,100 / 265,000
Asia / 330 / 253,000 / 280 / 217,000
Eastern Asia / 55 / 11,000 / 60 / 13,000
South-central Asia / 520 / 207,000 / 410 / 158,000
South-eastern Asia / 210 / 25,000 / 300 / 35,000
Western Asia / 190 / 9,800 / 230 / 11,000
Latin America & the Caribbean / 190 / 22,000 / 190 / 22,000
Oceania / 240 / 530 / 260 / 560

Includes Canada, United States of America, Japan, Australia and New Zealand, which are excluded from the regional averages.

Data on neonatal mortality appear to be even scarcer than those for maternal mortality. Regional patterns for neonatal mortality (Table 2) are very similar to those for maternal mortality. Africa and South Asia account for over 93 percent of global deaths.

Table 2: 2001 Global and Regional Estimates of Neonatal Mortality

Region / Number of live births (1000s) / Neonatal deaths
(1000s) / Neonatal death rate (per 1000 live births)
Africa / 28,865 / 1,205 / 42
Asia* / 76,090 / 2,561 / 34
South-Central Asia / 38,442 / 1,757 / 46
Other Asia / 37,648 / 804 / 21
Latin America and the Caribbean / 11,553 / 196 / 17
Pacific Islands* / 225 / 8 / 34
Europe / 7,374 / 44 / 6
North America / 4,098 / 18 / 4
More Developed Countries / 13,045 / 65 / 5
Less Developed Countries / 116,550 / 3,970 / 34
Global / 129,596 / 4.035 / 31

* Japan, Australia and New Zealand are included with the More Developed Countries but not in the regional sub-estimates.

Source: Save the Children, 2002

Death rates are highest in the South Central Asia region, at 46 per thousand live births, followed by Africa, with 42. The rate is lower for the Other Asia group because China is included there and has a substantially lower neonatal death rate (23), compared to much higher rates for the large countries in South Asia—Bangladesh (48), India (43), and Pakistan (49). While the Pacific Islands account for a small proportion of neonatal deaths, their NNM rate is comparatively high.

Poverty Linkages

Measuring maternal and neonatal mortality for sub-groups of the population within countries is even more challenging than country-level estimates. Graham and colleagues have developed a technique for estimating rich-poor differentials in maternal mortality using Demographic and Health Survey (DHS) data for 10 countries (Burkina Faso, Chad, Ethiopia, Indonesia, Kenya, Mali, Nepal, Peru, Philippines and Tanzania) with large sample sizes using wealth-quintile methodology developed at the World Bank (Graham et al, 2004; Gwatkin et al, 2004). In the country with the largest sample size and also with two surveys, Indonesia, they found that the poorest quintile accounted for one-third of all maternal deaths in both surveys, compared to fewer than 13 percent of deaths in the richest quintile. They also found a high level of association between the survival status of women and poverty status in all of the countries, and a highly significant correlation between education and survival status.

Table 3: Attendance at Delivery by a Medically Trained Person by Wealth Quintile

Region / No. of countries / Regional average / Poorest quintile / Richest quintile / Poor/rich difference
East Asia / 4 / 53.6 / 26.6 / 90.4 / 63.8
Europe/Central Asia / 6 / 94.9 / 88.4 / 99.2 / 10.8
L. America, Caribbean / 9 / 66.0 / 43.2 / 93.3 / 50.1
Middle East, N. Africa / 4 / 52.5 / 33.6 / 80.3 / 46.7
South Asia / 4 / 21.5 / 7.0 / 56.7 / 49.7
Sub-Saharan Africa / 29 / 43.5 / 24.2 / 77.1 / 53.4
All country average / 56 / 51.6 / 32.7 / 81.7 / 49.1

Source: Gwatkin et al, 2004

It is also possible to get a sense of rich-poor differences for other countries (56 in all, including the 10 for which MMRs by wealth quintiles have been calculated) by using the same DHS data for countries on deliveries attended by medically trained persons. This indicator is known to be highly correlated with both maternal and neonatal mortality (and has, in fact, been used to estimate the proportion of maternal deaths in countries lacking other maternal mortality data). Table 3 shows a 49 percentage point difference in the proportion of skilled attendance between the richest and poorest quintiles for all 56 countries for which the tabulations have been made—with the poorest quintile averaging 32.7 percent compared to 81.7 percent for the richest quintile.

Rich poor differences are greatest (64 percentage points) in East Asia, though only four countries are included in the tabulations (Cambodia, Indonesia, the Philippines, and Vietnam) and least (11 percentage points) for Europe/Central Asia (six countries: Armenia, Kazakhstan, the Kyrgyz Republic, Turkey, Turkmenistan, and Uzbekistan). That region also has the highest average level of attendance, 88 percent. South Asia (four countries: Bangladesh, India , Nepal and Pakistan) has the lowest overall level of skilled attendance, 21.5 percent and a 49.7 percentage point poor-rich differential. Latin America has the second highest overall level of skilled attendance (43 percent), but also a comparatively high (50 percentage point) differential between the rich and the poor. Sub-Saharan Africa has a higher overall average for attended deliveries (43.5 percent) than South Asia, which is puzzling given the MMR estimates for Africa. DHS data are available for 29 countries in Africa, suggesting that the issue may be poor quality of delivery care rather than under-representation of regional experience in the tabulations. Rich-poor differences are in the middle range (53 percentage points) in Africa, though attendance for the poorest quintile (25 percent) is second lowest in the tabulations after South Asia. The Middle East/North Africa (MENA) group (four countries: Egypt, Morocco, Jordan, Yemen) is in the middle of the range in terms of the regional average (52.5 percent) and rich-poor differential (46.7 percentage points).