p. i

Setting the stage for equity-sensitive monitoring of the health MDGs

Meg Wirth1 Deborah Balk2

Adam Storeygard2

Emma Sacks2 Enrique Delamonica3

Alberto Minujin3

1 Consultant to Task Force 4 (Child health and maternal health), UN Millennium Project; contact mwirth@earthlink.net 2 Center for International Earth Science Information Network (CIESIN), Columbia University

3 Global Policy Section, Division of Policy and Planning, United Nations Children’s Fund (UNICEF)

Background Paper for Task Force 4, Child health and maternal health

UN Millennium Project

FINAL DRAFT

5 November 2004

p. ii

Contents

I. Rationale 1 II. Equity Indicators for Monitoring Purposes 4

III. Methods and Data 8

A. Health Indicators B. Social Stratifiers C. Surveys and Data Sources D. Methods

IV. Results 1 Ethiopia 13 Cambodia 16 Kenya 18 Ghana 20 Dominican Republic 23 Tajikistan 24

V. Discussion 25

A. Key Findings B. Limitations C. Health information asymmetry D. Setting the Stage: Adapting the MDGs for equity-sensitive monitoring

VI. Recommendations 32 References 34 Appendix 1 Definitions of Indicators Appendix 2 Definitions of Stratifiers Appendix 3 Two major population based surveys: DHS and MICS

p. iii

Acknowledgments:

We are grateful to Paula Braveman for input at all stages of the paper, Mushtaque Chowdhury for support during the early stage conceptualization of the paper and Lynn Freedman for support throughout the process. We would also like to thank James Connolly for statistical and editorial support and Davidson Gwatkin and Jeanette Vega for comments on a near final draft of the paper.

p. iv

Setting the stage for equity -sensitive monitoring of the health MDGs

Meg Wirth, Deborah Balk, Adam Storeygard, Emma Sacks, Enrique Delamonica, Alberto Minujin

Abstract The Millennium Development Goals must be buttressed by an explicit and systematic commitment to equity in order to ensure that poor, marginalized and vulnerable groups are given opportunities for health and access to health services. This analysis seeks to ‘set the stage’ for equity-sensitive monitoring of the health-related Millennium Development Goals (MDGs) using data from the broad-scale, international household-level surveys (i.e., the Demographic and Health Survey and Multiple Indicator Cluster Survey) to demonstrate that establishing an equity baseline is necessary and feasible, even in very low-income, data-poor countries. We examine six countries using 20 health indicators and 6 social strata to ground our recommendations in current data. Use of bivariate tables shows effects of single stratifiers on different population groups. We also utilize trivariate tables to expose unexpected outcomes that occur when two stratifiers are used simultaneously. We found that inequities are complex and interactive: one cannot draw inferences about the nature or extent of inequities in the health outcomes from a single indicator. On the whole, the majority of the chosen social stratifiers yield statistically significant differences within all of the child, maternal and reproductive health indicators. This finding validates the importance of assessing health equity across dimensions (including gender, ethnic group, wealth quintiles, geographic region) rather than simply one of them. Thus, from a policy perspective, health systems that are universal and inclusive will better incorporate the vulnerable and marginalized than will a series of targeted programs. These findings from specific countries are complemented by a wider call for investments in country-owned health information systems to track trends in population health. We argue that these equity considerations must be a critical element in the implementation of health policy and Poverty Reduction Strategies guided by the MDGs.

p. 1

I. Rationale

This analysis seeks to ‘set the stage’ for equity-sensitive monitoring of the health-related

Millennium Development Goals (MDGs) using data from the broad-scale, international household-level

surveys (i.e., the Demographic and Health Survey and Multiple Indicator Cluster Survey) to demonstrate

that establishing an equity baseline is necessary and feasible , even in very low-income, data-poor

countries. We examine six countries using 20 health indicators and 6 social strata to ground our

recommendations in current data.

Increasing international attention to equity has led to recognition that, while overall health

indicators are improving in many countries, many gaps are increasing or remaining stagnant. This general

statement obscures a very complex reality on the ground—inequities may be narrowing for some health

indicators and widening for others, and inequities between social groups may exist for some health

indicators but not for others. We argue that a complete and country-specific analysis of inequities in

health is an essential baseline for tracking policy changes and their effect on development indicators.

The global consensus represented by the MDGs is a new point of departure for the development

community. The UN Millennium Project, along with scores of other decision-makers, activists, bilateral

aid organizations and communities, is already deeply immersed in efforts to reach the MDGs (Sachs et al.

2004, UNDP2003.). In keeping with the arguments set forth in the Millennium Project’s Task Force 4

Interim Report and elsewhere, we argue that this effort must be guided by an explicit and systematic

commitment to equity in order to ensure that poor, marginalized and vulnerable groups are given

opportunities for health and access to health services (Freedman et al. 2004, Gwatkin 2003). And in order

to ensure that progress toward more equitable health outcomes is quantifiable, it is critical that we take a

hard look at the way the indicators and targets are framed and the data sources available for monitoring

the health MDGs over time.

This paper is an effort to move beyond the very important stages of ‘setting the values base’ for a

focus on equity and ‘sounding the alarm’ about growing gaps in health, to putting in place the

requirements for assessing progress toward achieving greater equity. We use widely available data from

recent Demographic and Health Surveys (DHS) and Multiple Indicator Cluster Surveys (MICS) to

demonstrate a practical analysis of the nature and extent of inequalities in health in particular country

contexts; this approach is generalizable to the vast majority of resource-poor countries, which have at

least one such population-based household survey containing information on health and on important

social characteristics.

Information and data are essential currency in the effort to reach the MDGs and arguably,

nowhere are data as critical as in the realm of counting who lives, who dies, who is healthy and who is

sick (Stansfield and Evans 2003). As the efforts to reach the goals gain momentum, we note that health

p. 2

information systems must be upgraded to keep pace and measure progress; population-based surveys

must be examined for all they have to offer in monitoring the MDGs, and key indicators must be framed

in equity-sensitive terms.

Of particular concern, is the ability of countries to effectively gather their own health information

and monitor health status and health system outcomes within their boundaries. While international

comparisons are an important dimension of equity analysis, they are not the focus of this paper.

Monitoring equity implies an ongoing assessment of how different groups in society are faring in absolute

terms, and how they fare relative to one another (Braveman 2003). In order to measure progress in

reducing the extent of health inequalities countries must have a clear baseline for their own health MDGs,

based upon unique country circumstances and socio-cultural dynamics. Population averages are

insufficient for monitoring the health MDGs.

For many countries lacking vital registration systems, monitoring of the MDGs will rely—at least

in the short-term—upon population-based surveys such as the Demographic Health Surveys (DHS) and

Multiple Indicator Cluster Surveys (MICS).1 As the largest worldwide survey programs, MICS and DHS

are expected to play a major role in monitoring the MDGs. However, their fitness for the task of

monitoring the MDGs—and in particular, for monitoring in an equity-sensitive manner, requires further

consideration. We use recent DHS and MICS surveys from six countries to illustrate our points—but

note that other data sources including censuses, vital registration, qualitative studies, health facility

information and surveys of refugees also must be used in order to fully describe the extent and nature of

inequalities in a given country and to chart and monitor a path forward.

This paper is based around the following four arguments and assumptions: The MDG targets

should be framed in equity-sensitive terms; equity is not simply a matter of poor versus rich; the full

complexity of social disadvantage must be considered; it is feasible to measure inequalities in health

based upon current data; yet we also need to build basic health information systems as an essential part of

strengthening health systems and donors must commit to these long-term investments.

What is unique about this analysis?

While analysis of health data on resource-poor countries from an equity perspective has become

more widespread in the last five years, this paper, we believe, is unique in several ways. First, the

analysis and recommendations are specifically tied to the United Nations Millennium Project, building

upon the country case studies selected by the Project for in-depth analysis across all sectors, from

1 Even as vital registration systems become more available, surveys will still play a useful role in monitoring health equity. Surveys tend to include more indicator and strata than is possible with vital registration, even though they capture information in a cross-section.

p. 3

education to agriculture, poverty and trade to gender, environment and health. As such, the paper

specifically addresses within country inequalities in health with the intent of demonstrating how the

MDGs can and should be monitored to flag worrisome gaps in health between population groups. Thus,

the paper looks at multiple health outcomes ranging from traditional child health indicators such as under-

five mortality rate, infant mortality rate and underweight to maternal health indicators to HIV/AIDS

knowledge and contraceptive prevalence rates. In general, most equity studies have tended to take a

single health outcome—such as child mortality, skilled birth attendants or antenatal care—and examine

the distribution of this indicator across and within countries though there are notable exceptions (Stanton

and Blanc 2004, Kunst and Houweling 2001, AbouZahr and Wardlaw 2004, Victora et al. 2000, Wagstaff

and Watanabe 2000, Wagstaff et al. 2003).

Second, the paper attempts to deepen analysis of inequities in health by showing the complex

reality in different country contexts. While inequalities in health appear to be universal, in certain

settings, some health outcomes and determinants, such as immunization, age at first marriage, and even

contraceptive prevalence rate appear to be less unevenly distributed across population groups than

conditions such as severe child undernutrition and skilled attendance at birth. We bundle together a set

of 20 health indicators that relate to the MDGs, allowing the reader to look at how the nature and extent

of inequities compare across these different types of health indicator.

Third, the paper addresses multiple dimensions of inequity, moving beyond the overly simplistic

poor versus rich dichotomy. Equity in health is not simply a matter of rich versus poor. Other forms of

disadvantage, such as ethnicity, educational level of the mother, region of the country, urban versus rural

residence and gender are considered important stratifiers of health and access to health, both on their own,

and in interaction with one another. We undertake bivariate and trivariate analysis (herein called two-way

and three-way tables) to look at each of the 20 indicators across several stratifiers to capture the

complexity of health disadvantage. Note that we avoid multivariate regression analysis in order to

simplify replication in low resource settings.

Fourth, the paper assesses the statistical significance of the inequities in health status, specifying

in which cases the gaps are a result of random variation and in which cases the inequalities are

statistically valid. While such a statistical assessment should be routine, they are generally not included. 2

Finally, the paper situates the call for equity analysis within the context of a all for donor

involvement in building the health information systems of the poorest countries in the world, making

recommendations for strengthening current population based surveys such as DHS and MICS, but also

underscoring the enormous need for vital registration systems. The paper attempts to ‘set the stage’ for

2 Similar bivariate cross-tabulations appear in the country report series of the DHS and MICS reports but no tests of statistical significance are given.

p. 4

equity-sensit ive monitoring of the MDGs by demonstrating some of the key steps that are possible with

current and available data. However, this analysis represents just a small part of the process that should

be undertaken within countries, involving stakeholders in the identification of key health and social

disparities as part of the wider process of national planning and poverty reduction.

II. Equity Indicators for Monitoring purposes

Equity-sensitive monitoring serves primarily as an ‘early warning’ system, pointing to fault lines

in a country’s social fabric or economic system (Braveman 1998). Operationally, addressing inequities

in health is defined as minimizing avoidable disparities in health and its determinants—including but not

limited to health care—between groups of people who have different levels of social advantage

(Braveman 2003). From a human rights perspective, narrowing avoidable disparities in health outcomes

or health access between population groups is an imperative for the state. Often, the health indicators for

the poorest, least educated, geographically or ethnically marginalized groups are many times lower than

those for ‘better-off strata’ of the population.

A range of studies over the last five to six years has demonstrated that inequities in health

outcomes differ across and within countries and across different health indicators (Minujin and

Delamonica 2003, Segone 2001, Evans et al. 2001, AbouZahr and Wardlaw 2004, Victora et al. 2000,

Wagstaff and Watanabe 2000, Kunst and Houweling2001). In addition, studies have shown the extent to

which even public expenditure on health (and other social services) tend to disproportionately favor the

better-off groups in society (Gwatkin 2002, Castro-Leal, Dayton, . et al. 2000). On average the rich will

tend to capture a larger share of the increment to national income from growth than the poor (Ravaillon

1995). Though the fact that the poor suffer greater ill-health than the rich is perhaps conventional

wisdom, these recent analyses enabled a quantification of the differentials in access to health care and in

health outcomes. What has been striking in the analyses, is the oft-seen stepwise gradient in which each

step up the ‘wealth ladder’ confers greater advantage in health. Child malnutrition and access to

contraception and to skilled attendance are the indicators which have yielded the greatest gaps between

rich and poor and are the most unequally distributed across wealth quintiles3.

As important as 'wealth' is in determining who has access to care and who has better health

outcomes, other stratifiers may be equally (or more) significant in monitoring inequalities and in

designing appropriate policies. Gender, race/ethnicity (perhaps measured by language) and geographic

location are significant stratifiers of health opportunity that should be tracked. Yet, tracking indicators on

their own should, where possible, be supplemented by analysis that looks at interactions between

stratifiers. Examination of interaction effects allows one to specify, for example, the cumulative 3 As measured by the concentration index.

p. 5

disadvantage conferred on a child who is in a poor family with a mother with no education, or the effect

of being from both a marginalized ethnic group and having no education. In this analysis, we find that

indeed the effects of marginalization are often magnified when looking at two categories at once.

Trends over time will also be critical for health equity analysis of the MDGs. Of particular

interest are the countries that have managed to improve average health outcomes while reducing

disparities in health (Minujin and Delamonica 2003, see Box 1). In an analysis of trends in inequalities

in health, Minujin and Delamonica (2003) draw attention to countries—‘positive outliers’ –that were able

to do just that. Indonesia, Guatemala, Senegal, Niger, Egypt and others were able to improve aggregate

levels of delivery attendance (and Guatemala and Togo for U5MR, Guatemala, Egypt, Indonesia, Uganda

and others for DPT3 coverage) while reducing disparities between rich and poor. While we do not

undertake trend analysis in this exercise, we do note that the overarching goal of an equity analysis should

go beyond simple description of gaps between population groups. Going beyond simple description

means quantifying the extent of the disparity with and intent toward monitoring the progress of policies

designed to reduce disparity.

p. 6

BOX 1: A Means for Monitoring the Equity Implications of MDG Implementation National averages are not enough to assess progress because changes in the national average do not necessarily imply improvement for all members of society. For example, a national average improvement could go together with increases in the relative gap between the poorest and the wealthiest (or some other measure of inequality). In other words, it is possible that the national average improves while the situation of the disadvantaged and vulnerable groups deteriorates. The table below shows the four possible combinations of changes in disparities and country averages. Source: Minujin and Delamonica 2003

In this context it is important to highlight that all too often the poorest are the last ones to improve their health status. The available information for the 1990s, indicates that the wealthier households benefited more from improvements in access to health services than the other social groups. The evidence shows that the poor are usually “queuing” for access to basic social services; they are at the end of the line (Vandemoortele 2000).

We use the process described in Figure 1 to demonstrate the technical aspects of an approach to monitor

equity. This methodology was developed by Braveman for WHO (1998) and encompasses a series of

steps that are designed to guide monitoring health equity in ways likely to be policy-relevant. Although

the steps described here are presented as if they are purely technical and scientific, as indicated in the

original document, they take place in the context of a highly politicized process (Braveman 1998).

Indicators used for monitor ing must meet many criteria, including being simple, reliable, sustainable,

timely, policy-relevant, and affordable (Braveman 1998). The very process of determining what stratifies

health and opportunities to be healthy exposes the ‘fault lines’ in society—pointing directly to the ‘haves’

and ‘have-nots.’ Doing so is often highly charged politically, as there are, in perhaps most contexts, very

strong interests in maintaining the status quo or suppressing identification of fault lines. “Achieving

greater equity generally requires real changes in resource allocation to favor disadvantaged groups, who

Trends DISPARITIES Narrowing Widening NATIONAL

Improving

BEST OUTCOME

Improvement for better-off children, but not for the disadvantaged

AVERAGE Worsening Worsening with an element of protection of the disadvantaged

WORST OUTCOME

p. 7

by definition, are less influential” (Braveman 1998).

Figure 1. (Source: Braveman 1998)

Our ability to measure success is heavily dependent on the quality of data available and the

indicators used. And conversely, the way that the goals are framed dictates the way countries gather and

use information and data. Are the goals as they are currently specified sufficient for measuring ‘success’?

Task Force 4, argues in its Interim Report, that the health MDGs are a powerful tool for motivating

change, but that as currently specified they are insufficient in that they do not include an equity

dimension. Specifically, the Report recommends that equity criteria be integrated into the indicators,

rewording the goal to the following: Reduce by three-quarters, between 1990 and 2015, the maternal

mortality ratio, ensuring the same rate of progress or faster amongst the poor and other marginalized

groups (Freedman et al. 2004).

p. 8

Vietnam is most often cited as the country that has actively adapted the MDGs to a local

context—and done so in an equity-sensitive manner. In part, this is a reflection of the country’s already

impressive progress in reducing poverty over the last decade or so. The MDGs were transformed into the

Vietnam Development Goals (VDGs) in an effort to explicitly link these goals to the national

government’s five-year plans. The government took a strong stance on equity, noting that certain ethnic

groups are particularly disadvantaged. Accordingly, Vietnam modified the maternal health target to be

‘reduce the maternal mortality rate to 80 per 100,000 live births by 2005 and 70 by 2010 with particular

attention to disadvantaged areas’ and the child mortality targets to include ‘reduce the infant mortality

rate…. at a more rapid rate in disadvantaged regions’ (Swinkels and Turk 2002). This explicit ‘equity-

sensitive’ wording of the MDGs is all too rare. It is hoped that the data presented in this exercise will

highlight the need to look at the way that the MDG indicators are ‘patterned’ across population groups

and to devise Poverty Reduction Strategy Papers (PRSPs), national development plans and other policies

that explicitly plan to reduce disparities in health and the other socioeconomic determinants which affect

health.

III. Methods and Data

A. Health Indicators

The 20 health indicators used in this analysis were selected to match the MDG indicators as

closely as possible, with a few exceptions (see Appendix 1). For example, three childhood mortality

indicators were included to provide greater detail: neonatal mortality, infant mortality and under-five

mortality. DPT (Diphtheria, Pertussis, and Tetanus) immunization rates were added because vaccination

against these illnesses has been less the focus of massive internationa l and country-specific campaigns in

the way vaccination against measles has. Thus, we hypothesized, DPT coverage reveals something about

the strength of the health system, and in particular, the trend in coverage for DPT1, DPT2 and the final

immunization, DPT3, may be indicate which groups tend to ‘fall off’ in terms of coverage over time

(Ransom 2004). On the maternal mortality side, reproductive health care indicators were expanded

beyond skilled attendance at delivery, knowledge of HIV/AIDS, and contraceptive prevalence rates, to

also include age at first intercourse and union, but other important aspects of reproductive health (e.g.,

adolescent fertility rates and unmet need for family planning) were beyond the scope of the current study.

The nature of the health indicators chosen varies. Some of the indicators chosen are health

outcomes (underweight, child mortality), some represent access to health care or preventative

interventions (visits to a health facility, skilled attendant at birth, measles and DPT vaccination,

contraceptive prevalence) and others represent health knowledge (AIDS knowledge) and fertility-related

p. 9

or women’s status indicators (ages at first intercourse and marriage). Note that underweight is a health

outcome in its own right, but also an underlying cause of some 60 percent of child mortality, and thus an

important aspect of child health. Clearly the full complement of health indicators relevant to a country’s

epidemiological burden—including malaria, TB, anemia, etc.—would need to be explored in a complete

equity analysis (see Box 2 for a more thorough listing of criteria for picking indicators).

To get some indication of access to health system resources, this study assessed visitation rates to

a clinic. Such a variable often raises concern of unobserved heterogeneity whereby less-healthy persons

disproportionately seek health clinic services. However, our concern is tempered over the time period of

one year, given the health requirements of women in their reproductive years and those of their young

children. The question is asked, however, as one of a series of questions regarding family planning (the

follow-up question is "did a health professional at the clinic speak with you about family planning?" and

the number of women times women visited health clinics but were not counseled on family planning are

considered "missed opportunities"). These questions were included in the surveys not as a measure of

sickness, but as an indication of how much of the population services could reach. These services could

include family planning, but also other health resources, and this measure indicates who are the potential

recipients of any services provided by a health center.

BOX 2: Criteria for selecting indicators to monitor equity

§ The indicators must meet standard scientific and ethical criteria (validity, reliability, ethical and cultural acceptability), with data sources of acceptable quality and reviewed by local experts who know from experience the source’s limitations.

§ Differences in the indicators between better- and worse-off groups should be relatively likely to reflect

avoidable, unfair gaps in important conditions that could be narrowed through policy changes in any sector that influences health, not just health care.

§ Appropriately disaggregated data (disaggregated according to the social groups one wants to compare) on

the indicators should be accessible for monitoring over time at the desired geographic level. § The indicators must occur frequently enough in the groups to be monitored to permit reliable estimate of

differences between the groups. § The complexity of analyzing, presenting and interpreting information on the indicators should be

considered in light of the experience of the local analysts, decision-makers, and other key participants who must perform these functions on a routine basis, generally without outside technical support.

A range of indicators is needed to reflect important aspects of health and its major determinants, including

indicators of health status, major determinants of health, and health care itself (including financing, allocation of health care resources, utilization of health services and quality of health services).

§ When a large number of indicators meet the above criteria, an additional criterion to set priorities among

the options to be presented to non-technical audiences should be the extent to which the indicators are likely to be meaningful (i.e. to reflect important concerns) to the public and policy-makers.

Source: Braveman 1998

p. 10

B. Social Stratifiers

An equity analysis requires stratifying a population into subgroups that have different levels of

underlying social advantage. The social stratifier most often used to look at inequities is income or

wealth, often using ‘wealth quintiles’ to divide the population into equal fifths according to their level of

wealth within a country-specific wealth distribution. Wealth is usually measured based upon a set of

assets the family has, rather than in terms of monetary income or expenditure. A burgeoning literature

points to the systematic gradient in health that occurs in rich and poor countries alike—with better-off

groups usually having better health status and access to health services. The existence of this gradient

across wealth quintiles has been an important development in the literature.

But in countries where the vast majority of the population is extremely poor, and only the capital

city has a population which falls in the upper quintile, the stratification by wealth quintile is not

necessarily the most appropriate way to measure inequities in health. Even in countries where wealth

does stratify health access and health outcomes, other stratifiers should be used to analyze the distribution

of health outcomes as well.

Other important stratifiers have included age, urban-rural residence and gender, and more rarely,

ethnicity, occupation, geographic region, and education level. However few analyses have included most

or all of these stratifiers in a single study. All but age (with the exception the inherent distinction in age-

specific indicators of NNMR, IMR and CMR) and occupation are studied here (see Appendix 2 for

details). The regions used as stratifiers here correspond to survey regions - the smallest geographic areas

for which estimates are produced. The number of regions per survey varies with the size of the sample,

along with other factors. Among the countries studied here, Kenya has only seven regions, while the

Dominican Republic has 32. Especially when used in combination with another stratifier, sample sizes in

individual regions can become too small to yield meaningful results.

C. Surveys/Data Sources

Data are compiled from DHS surveys of Cambodia (2000), Dominican Republic (2002), Ethiopia

(2000), Ghana (1998), and Kenya (1998), and a MICS survey of Tajikistan (2000). These countries and

surveys were chosen because they are selected Millennium Project case studies. Tajikistan data is taken

directly from aggregate tables distributed by UNICEF, whereas most measures from DHS are calculated

from individual woman- or child-level data. The remainder were taken from DHS reports or the DHS

website. Indicator definitions were harmonized across the five DHS countries when possible. Some

indicators reported here differ from those in official DHS reports (Rutstein and Rojas 2003). For

example, values of “Don’t Know” or “Missing” were excluded from our analysis, whereas in DHS

reports, these categories are sometimes explicitly reported, or, in some types of AIDS knowledge,

p. 11

considered equivalent to “No”. Similarly, DHS reports contraceptive prevalence rates for women

currently in union, whereas we report for all women. We report mean age at marriage, as opposed to

median.

D. Methods

In general, the indicators described above were calculated for all stratifiers. However, some

indicators and stratifiers studied are based on data not available for all six countries. Others were omitted

because of indicator-specific computational difficulties. Indicators were not calculated for all

combinations of two stratifiers. For example, the cross-tabulation of ethnicity and region resulted, in

most cases, in too many very small or empty categories to allow for meaningful analysis. However,

ethnicities were recoded into dominant, not dominant, and in some cases, secondary dominant categories,

based on relevant literature 2 in order to create a set of larger classes to be used as a stratifier in

combination with regions.

The "Wealth by poverty line" variable was created using the existing wealth indices (based on

DHS/Filmer and Pritchett 2001). This approach was developed as an effort to complement the

stratification by wealth quintile with a very simple, immediately policy-relevant distinction between rich

and poor. Using data from the World Development Indicators, we determined what percentage of the

population (x percent) lives below the poverty line. Then, we took the bottom percentage x of the

population based on the wealth index for that country to create the SES/Wealth by poverty line variable.

Each two-way table presents the indicators described at the top for each stratifier at the left. In

nearly all cases, the numbers are percentages of the class that fulfill the requirement at the top (e. g.

having a DPT2 vaccine, being underweight, or using a modern form of contraception). The ‘Age at first

intercourse’ and ‘Age at first marriage’ columns are exceptions, representing means. For each

stratification class (e. g. gender, region), a difference test is also reported (labeled p-value and in grey

italics in the first line of the relevant cell). These are probabilities of the null hypothesis that the values of

an indicator for each level within a stratifier (i.e. none, primary, and secondary within education) are not

significantly different from each other. Tests of significance were not performed on the mortality rate

indicators, because they are rates rather than proportions or percentages. However, Standard errors (SE)

are given in individual DHS reports for the national mortality rates. National-level SE can be used as a

general indication of likely significance between groups.

2 These include Moyo (2004) for Ethiopia, and Brockerhoff and Hewitt (2000) for Ghana and Kenya. For Cambodia, a rough proxy was created as follows: residents of the two provinces with ethnic minorities in the majority, as shown in Elledge et al. (2001), were designated “Not dominant”. Both provinces are more than two-thirds minority, while all other provinces of Cambodia are less than one-tenth ethnic minority. No ethnic group information was available for the Dominican Republic or Tajikistan.

p. 12

The three-way tables are analogous to the two-way tables, with each class on the left broken

further into subgroups based on the other stratifiers. Some pairings of stratifiers are not generated in the

three-way tables. For example, the raw ethnicity classifications are not combined with regions, because

doing so would result in the majority of classes being null for all countries studied. Other indicators such

as infant mortality were not included in the three-way tables, because the number of events (deaths)

would be too few to construct robust rates. Also, the full set of p-values for the three-way tables were

beyond the scope of this work, although selective test statistics were run and were used to screen

particular effects reported.

V. Results

This section attempts to succinctly summarize a vast amount of data generated via the

methodology described above. Our emphasis is on within country patterns that emerge, rather than

comparisons between countries. In order to organize our findings, we present data for each country in

two categories: first, observations or patterns which were ‘expected results’ on health inequalities and

second, observations or patterns which seem to be ‘unexpected results,’ surprising variations on the more

expected results or findings which seem to overturn conventional wisdom. One caveat worth mentioning

is that as analysis in these areas is quite sparse, determining what is the conventional wisdom is not

always straightforward.

Given the international literature on the subject, we would expect most indicators studied to be

stratified by wealth quintile. Certain indicators such as measles immunization, we might expect to be less

stratified by wealth because measles immunization tends to be the subject of nationwide ‘vertical’ health

campaigns that should, in theory, reach everyone more or less equitably. Other indicators such as

‘underweight’ we might expect to more stratified by wealth quintile because, as the literature indicates,

underweight is more affected by access to food, which is in turn dependent upon wealth or income

(Wagstaff 2002). In general, we expect rural health outcomes to be worse than urban and we expect a

certain degree of heterogeneity amongst the regions, with certain regions falling behind in most

indicators. With regard to ethnic groups, the literature is far less complete on the subject, but we

hypothesize that a few non-dominant ethnic groups will fall behind the national averages across all the

indicators measured. Based on a much more comprehensive literature on the role of maternal education

on family health outcomes, we expect stratification across education of the mother for all of the health

indicators.

We present results from the two-way and three-way tables in appendices and attempt to

summarize the main themes from these tables. Values based on less than 25 cases (250 live births in the

case of mortality rates) are not shown. Values based on 25 to 50 cases (250 to 5000 live births) are shown

p. 13

in parentheses. We summarize the most salient results from these tables, but further exploration of both

the two-way and three-way tables by experts with an interest in a particular country will likely yield

additional important conclusions.

ETHIOPIA

Education, Ethnicity, Residence, Region and Wealth quintile are highly significant stratifiers for

virtually all indicators; Education, ethnicity and region seem to be the most potent stratifiers of the

indicators studied for Ethiopia. For children, underweight status is high (47%) across the country. Wealth

quintile, ethnicity and gender do not seem to change this percentage to a great extent. But stratifying by

region and by level of education does reveal enormous differences, with 14% underweight in Addis and a

high of 53% underweight in SNNP district. In fact, the distribution of underweight children shows a far

greater range across ethnic groups and regions than the other indicators. Further, measles vaccination (and

DPT1-3) appear to be even more sensitive to ethnicity than underweight is, and also sensitive to

differences in region, education, urban versus rural residence.

Ethnicity significantly stratifies all the reproductive health and child health indicators (except moderately

underweight) but the relative dis/advantage of the ethnic groups is not constant across indicators (see

Table 1). The Affar ethnic group is the exception—it is disadvantaged—well below the national average

for all indicators measured, both child and reproductive health. The national average on the age at first

marriage (16) obscures a seven-year range from 13 or 14 among certain ethnic groups (who tend to

cluster in individual regions) and 20 amongst other ethnic groups. Similarly, region stratifies all the

indicators but the relative advantage of different regions is not constant.

Expected Results

% Access to Health Care Facility (Visited

in Last Year)

Age at First Marriage

% CPR - Modern Method

% Underweight% Measles

immunization

Welaita (30) Amharra (14) Somalie (0) Affar (55) Affar (2)Affar Tigray Affar Guragie Welaita*Somalie Affar Welaita Other OromoOther Sidama Other Welaita OtherAmharra Oromo Oromo Sidama GuragieOromo Welaita Sidama Tigray AmharraGuragie Other Guragie Amharra SidamaTigray Somalie Tigray Somalie SomalieSidama (49) Guragie (18) Amharra (8) Oromo (43) Tigray (67)

National Average

36 16 5 47 27

Ethnicity

Percent in each category

Table 1: Ethiopian health indicators by ethnic group (2000)

p. 14

Certain regions stand out as disadvantaged. Amhara and Somali lag farthest behind of the regions

while the ethnic groups Affar, Amharra, Somalie and Welaita (and the residual category, ‘Other’) lag

behind the national average.

Skilled birth attendance: Major differences are evident when the indicator is stratif ied by

educational level with 3 percent SBA for those with no education, 10 percent for those with primary and

45 percent for those with secondary or more. Region dramatically stratifies delivery assistance; Addis is

in a category alone with 69 percent delivery assistance; Dire Dawa, Gambela and Harari are in the 24-34

percent range and the rest difference.

Underweight: First, it is important to note the extremely high population average of 47 percent

underweight children, and 16 percent severely underweight overall in Ethiopia. The most salient

stratifiers of underweight appear to be education and region—with those with no education having 50

percent children underweight, those with primary education having 40 percent underweight and those

with secondary or more education 27 percent underweight. Addis stands out as having a rate of

underweight at 14 percent in contrast to the rest of the country which hovers around 40-50 percent.

Unexpected Results

Access to a health facility: For those with no education or a primary education, the rural/urban

split in access to a health facility holds true, but for those with a secondary or more education, rural

versus urban residence gap closes (also no gap between urban and rural for the fourth and fifth wealth

quintiles). In the Harari region, it doesn’t seem to matter how much education you have, access to a

health facility lingers around 43-48 percent. Amharra, Somalie and Welaita ethnicities have lowest access

to a facility overall (around 30%); but Amharra or Somalie women who are in the first quintile of wealth,

you have even lower access—around 20%. By contrast, Tigray women have high access regardless: the

overall access figure is 45% and even those in the poorest wealth quintile 38% have access to a health

facility.

Contraceptive prevalence: Contraceptive prevalence rates in Ethiopia are extremely low—only 6

% of the population uses any method and 5% uses a modern method. Even so, certain groups stand out

with higher rates. Over 15 % of those with at least a secondary education use a modern method, and in

regions outside of Addis, the CPR-MM difference between no education and primary education is quite

dramatic; and certain regions such as Dire Dawa, Gambela and Harari have higher CPR. Addis residents,

especially those with no education or primary education—regardless of ethnicity—have use-rates of

modern contraception much greater than the national average, suggesting widespread equitable access to

contraception in Addis. Being in the poorest four wealth quintiles does not seem to stratify access to

CPR-MM, but being in the richest group yields a clear jump up to around 10-15 %. Note, however, Addis

p. 15

region is the only one in the country that includes the highest wealth quintile. Thus, it is unclear whether

this is more an effect of wealth or Addis residence.

Age at first intercourse & Age at first marriage are, we hypothesize, much more culturally

determined and less variable by education and wealth. The Ethiopia data reveals that age at first marriage

differs not very much or not at all between no education and primary education, but secondary education

yields a two-year increase from 16 to 18. This indicator differs not very much between urban and rural or

even by wealth. But ethnicity and residence reveal larger gaps, with ranges from 14 to 18 years.

Skilled birth attendant: Ethnicity is not a major stratifier for delivery assistance, though Amharra

and Guragie stand out as having higher percentages than the national average (8 and 13 percent

respectively). Wealth quintile does not make much difference except if one is in the highest quintile (25

percent). Note that skilled birth attendance yields quite different results from the ‘access to health care

facility in the last year’ indicator, suggesting that these questions measure two very different aspects of

the health system and are not proxies for one another.

Immunization: Immunization proves to be stratified less by wealth and more by ethnicity and

region. For DPT1 and 2, the first through fourth wealth quintiles are roughly the same level, while for

the highest quintile, the rates roughly double. For measles, there existed a three-fold difference between

richest and poorest. Yet, for both DPT3 and measles, quintiles 1-3 are about the same, quintile four is 1.5

times higher and quintile 5 is about double the lower quintiles. This suggests that an individual’s wealth

plays a larger part in the acquisition of DPT3 and measles vaccinations.

The Gambela region, which has the highest access to a health care facility (52 percent) and high

levels of delivery assistance by SBA as well as relatively high CPR using a modern method, has

surprisingly low immunization coverage levels at 13 percent for DPT3 (as compared to Addis at 82

percent). The Tigray ethnic group has dramatically higher levels of immunization coverage, almost nine

times the Affar ethnic group. Notably, measles coverage, which we expected to be relatively equitable

varied from 11 percent to 88 percent between the regions (see Figure 2).

Immunization coverage is also quite significantly different across education groups. Amongst

those with no education, DPT1 coverage is 40.2 percent, 27.7 percent and 16.4 percent respectively, while

for those with secondary or more education, coverage is 74.3, 60.0 and 55.9 percent respectively. The

same holds true for measles, with an approximately three-fold higher likelihood of measles vaccination

for those with secondary or more education as compared to those with no education. The bivariate

relationship suggests a slight favoring of male over female children for all vaccines. However, the more

in-depth view indicates an interaction with education: uneducated women significantly favor their sons

for all but DPT1, which is given at rates that are indistinguishable. Primary educated women treat their

sons and daughters more or less equally, and women with secondary or more schooling favor their

p. 16

daughters significantly for measles vaccination. Unlike the situation in many countries, it appears that

basic immunization is very inequitably distributed, suggesting that there are significant challenges in the

current implementation of even vertical programs.

Underweight: Even wealth does not appear to protect against having underweight children as in

the highest quintile underweight children are still at 37 percent. In rural areas of Affar, SNNP and Somali

regions, around one-fifth to one-quarter of children are severely underweight. And of note, even in the

highest wealth quintile, education still matters: children of mothers with no education are twice as likely

to be underweight and six times as likely to be severely underweight.

Neonatal mortality rate (NNMR), infant mortality rate (IMR) and under-five mortality rate

(U5MR): Wealth quintile is not particularly strong in terms of stratifying outcomes, nor is the urban/rural

distinction. What really stands out is level of education. There is a three fold differential in NNMR

between no education and secondary (61 and 25 per 1000 live births, respectively). IMR for children of

mothers with no education is 119 and those with secondary or more, half that. By region, Affar and

Gambela as well as Harari, Oromiya and SNNP stand out as having IMR greater than the national average

while Addis is 30 percent lower than the national average (81).

Figure 2

Children's measles coverage by region in Ethiopia (percent)

1120 20 20 25 27 28

4054 59

67

88

Affar

Oromiya

Ben-G

umz

Gambe

laSN

NP

Nationa

l avera

geAm

hara

Somali

Dire Daw

aHara

riTig

ray Addis

.

Source: Ethiopia DHS 2000

CAMBODIA

There are two salient issues that emerge from this analysis of health equity in Cambodia. The first

is that there are some outcomes in which the disparities are larger than others, in particular, delivery

p. 17

assistance and under-5 mortality. In both cases, the regional and wealth disparities are the most glaring,

implying these are the areas where future policy and monitoring efforts should be concentrated.

The second issue is that in some other outcomes, e.g. contraceptive methods and immunization,

the national averages are so low that disparities are not very large. However, as progress is made, it is

important to ensure that an equitable path is followed, lest the potential of increasing disparities (along the

“queuing model”) becomes a reality. Note that ethnic affiliation was not asked in the Cambodian DHS

and so a crude ‘recode’ is constructed based upon a several of highland regions with large percentages of

ethnic minorities. Thus, results must be interpreted with caution.

Expected results

AIDS Knowledge: Knowledge that a healthy-looking person may have AIDS and that using a

condom during sex can help prevent HIV infection is signific antly related to the level of education.

However, it is a bit surprising that AIDS knowledge is relatively high even among women with no formal

education (87 and 72 per cent respectively).

Skilled birth attendant: Not surprisingly, almost 90 percent of the births in Phnom Penh are

assisted by SBAs. This high level of coverage coexists with a national average of only a third of births

assisted by SBAs.

Access to a health facility: Similarly, the capital city is where women have the highest access to

health care facilities. This is not uncommon and the differences by region are statistically significant.

Nevertheless, the national average is very low (only 11 percent of the women visited a health care facility

in the year prior to the survey) and access in Phnom Penh is very similar to that in many other parts of the

country.

Immunization: Measles immunization improves significantly with the level of education of the

mothers. DPT coverage falls off from the first to the third dose. Of the 16 regions, measles coverage was

similar to DPT3 coverage in about half the cases. In five regions, DPT2 was similar to measles and in

another four regions it was DPT1. In relation to mortality or level of immunization there seems to be no

pattern relating them to the magnitude of the falling of DPT coverage or the similarity with measles. The

poorest wealth quintile is worst off, and the middle quintiles are invariably in between. The same patterns

were found with respect to access to clinics, delivery assistance, and mortality rates.

Underweight, both severe and moderate, is more prevalent in rural areas than in the cities. Fifty

percent of children from a family where the mother has no education were underweight and 15% percent

were severely underweight. The difference between the dominant and not dominant ethnic group was

relatively small (45 versus 54 percent). The Prey Veaeng and Mondol Kiri regions stand out as having

children underweight at rates above 50 percent. The poorest wealth quintile is worst off.

p. 18

Child mortality rates (neonatal, infant and under-five) differ by region. Phnom Penh, the capital,

consistently shows the lowest mortality. For IMR and U5MR, the next best region (a different one in each

of the three mortality indicators) is almost twice as high. The worst region for all three indicators is

Mondol Kiri, where mortality rates are about 4.5 times higher than in the capital city.

Unexpected results

Age at first marriage: Urban and more educated women marry later than the ones lacking

education or living in rural areas. However, the pattern is dominated by education as there is no more than

roughly a month delay among urban women with no education or with secondary schooling compared to

rural women with the same level education. For women who completed primary education, the difference

is larger. Nevertheless, it is only of less than 6 months.

Contraceptive prevalence rate: The level of formal education is not significant for utilization of

modern contraceptive method in Cambodia. Interestingly, CPR with a modern method is equitably

distributed across the education stratifier—there is no significant difference in access to modern methods

whether the mother has no education or secondary or more. However, access to modern contraception

does differ by region, with Banteay Mean Chey having 22 percent access and other regions having a third

of that level.

Immunization: The data shows no real difference between rural and urban children for DPT and

measles immunization, surprising especially given low national average (49 percent DPT3 and 56 percent

measles). Not completely unexpectedly, there is a lack of gender bias in DPT3, measles and underweight.

However, in the three-way analysis, males are significantly favored for measles among women with

secondary education and not those with less education. This is surprising.

Underweight: The urban bias in underweight seems to be concentrated among the households

where the mother has completed primary and secondary schooling. Among those with no formal

education, there is no difference between rural and urban underweight.

KENYA

Tables 1 and 2 reveal some consistent patterns: differences in ethnicity tend to be greater than by

any other characteristic on all indicators except underweight status. Overall, gender and education

differences (and even differences by wealth quintiles) seem more significant for urban-dwellers than

people in rural areas. (Many regions and ethnicities, however, are dropped from the three-way analysis

because of small sample size once divided into so many groups.) Wealth differences are significant on

child health, even more so than apparent differences in ethnicity.

p. 19

Expected Results

AIDS knowledge: Notably, AIDS knowledge is significantly inequitably distributed across all

analyzed population groupings in Kenya. The distribution is far greater for the knowledge about ‘condom

use’ as compared to the ‘health person can have AIDS’.

CPR modern method: Note that CPR (modern method) is stratified significantly by education,

ethnic group, region, residence and wealth. Of these, ethnicity has the largest range, but it also has the

largest number of classes.

Access to health care facilities: 70 percent of Kamba women sought health services, whereas

only about 40% of Luhya and Luo women sought health care. In contrast, the differences between the

poorest women (1st wealth quintile) and the least poor women (5th wealth quintile) ranged from 45% to

52%, not a particularly wide range, with the main difference being between the 1st and 2nd quintiles.

Skilled birth attendant: Education, Ethnicity, region, residence and wealth quintile all

significantly stratify skilled birth attendant usage. Twenty-seven percent of women with no education

used a skilled birth attendant, as compared to 36 percent of women with primary education and 72 percent

of women with secondary or more education. And region showed a slightly tighter range from 33 percent

in Western to 78 percent in Nairobi. As for ethnicity, the Mijikenda/Swahili were at a low of 27 percent

and the Kikuyu at a high of 71 percent. Though not necessarily surprising, the clear stratification across

all of these social determinants is quite striking for SBA.

Underweight: Children’s underweight status is significantly stratified by education, wealth and

ethnicity. Children who’s mothers have secondary or more schooling, or were in the 5th wealth quintile,

or were Kikuyu were least likely to be underweight (roughly 10 percent of these children were), as

compared to children of mothers with no education, in the poorest wealth quintile, or were

Mijikenda/Swahili, who respectively were 34 percent, 33 percent, and 32 percent underweight. Upon

further inspection in the three way tables, wealth differentials are significant for each educational group.

In each education level (none, primary, or secondary and more), the difference in proportions

underweight are 2-4 times as great for the children in the poorest households as they are in wealthiest

households.

The percentages of underweight children do not show significant gender differences for any

education level (if anything, there is a slight daughter-favoring in all cases, except secondary education or

more for severely underweight). As expected, rural children are more likely to be underweight, especially

in families where the mother has no or only primary education. Wealth, by poverty lines or by quintiles,

follow predictable patterns in that the non-poor fare better than the poor and the higher quintiles generally

have less underweight children, except in a few cases (2nd and 3rd quintile for severely underweight for

p. 20

primary education and 2nd quintile fared better than the 3rd quintile for all underweight indicators for

secondary education).

Unexpected Results

Age at first marriage: This indicator is significantly stratified by education, region, residence and

wealth quintile but not by ethnic group. In fact the difference in age at first marriage between the lowest

quintile (age 17) and the highest wealth quintile (age 20) is three years, quite significant for this indicator.

And the difference between the two educational-level extremes is 17 and 21, a four-year difference.

Immunization: A two-way view suggests that there are no gender differences in childhood

immunization rates, but difference emerge when adding other stratifiers. For example, 98 percent of urban

boys were vaccinated against measles whereas` only 90 percent of urban girls were. Despite these urban

differences, no gender differences emerged among rural children. A gender bias in vaccination rates,

which exists slightly for families where the mother has no education, decreases as the mother’s education

increases, but not to the extent that females are favored as in Ethiopia. Results for urban/rural differences

are surprising: DPT1 is about the same for urban and rural residents, but rural children have a slight

advantage for DPT2 and DPT3, whereas urban children have a large advantage for measles. When

further stratified by mother’s education, the measles pattern is only significant among those with

secondary education or more. Although most ethnicity indicators fall out (due to small numbers), it can

be seen that the dominant group (Kikuyu) has 100 percent coverage in vaccinations and few underweight

children, while the non-dominant group has numbers worse than the national averages.

Child mortality: It appears that the difference between no maternal education and primary

education does not yield large differences in terms of NNMR, IMR and CMR (though significance is not

measured), but secondary education yields large benefits (mortality drops to half the previous rates).

Ethnic groups show very wide variation, particularly for under-five mortality where the Luo suffer rates

2.5 times higher than the national average and almost eight times that of the Kikuyu. The lowest wealth

quintile suffers mortality at least two times as high as the upper quintile.

GHANA

Education, Ethnicity, Residence, and Region are highly significant stratifiers for virtually all

indicators. Although Ghana has the lowest mortality in the region (Balk et al., 2004) there are large

regional differences, with the highest regions—in the northern interior districts—appearing to have

mortality much more like that of its landlocked neighbors in Mali than like of more southern, and coastal

districts of Ghana. Further, across all the maternal and child indicators, regional differences far outweigh

p. 21

the effects observed by other strata, except for ethnicity, with which region is closely aligned. Northern

region, in particular, stands out as consistently having the worst outcomes for the child health indicators

examined, as well as skilled birth attendants. Significant effects of education, urban residence, and gender

were also observed on immunizations and underweight prevalence, as well as on the reproductive and

maternal health indicators.

Additionally, the variable on health care access was not included in the Ghana DHS survey. And,

although a wealth quintile has been constructed for prior Ghana surveys, it was not constructed for the

survey that was the basis of our analysis here.

Expected Results

Contraceptive prevalence rate -modern method: This indicator is significantly stratified by

education, ethnicity, region and residence, in the ways that we might predict. Disaggregation by ethnic

group (p<.05) yields a clustering around the national mean for many ethnicities, but certain groups like

the Gruma and Mole-Dagbani fall well behind.

Skilled birth attendant: Ethnicity appears to dramatically affect delivery assistance by a skilled

birth attendant, with a near two-fold, statistically significant difference between dominant (63 percent)

and the not dominant groups (34 percent). Mole -Dagbani, Gruma, and Guan ethnic groups were

particularly disadvantaged in terms of access to skilled birth attendants. Education level shows a very

similar pattern with mothers with secondary or more education being attended by a SBA at a rate of 65

percent and those with no education at 25 percent. By region, SBA exhibits a wide range, from 11.2

percent in the Northern Region to between 40 and 48 percent in the Central, Western and Eastern Regions

to a high of 73 percent in Greater Accra.

AIDS knowledge: Both indicators of AIDS knowledge are stratified significantly by education,

ethnicity, region and residence, suggesting a rather unequal spread and uptake of critical information and

education about HIV/AIDS.

Underweight: The figure below shows the variation of the various strata on underweight

children. The red mark indicates the national average, or 25% of children. In terms of region, the Greater

Accra region stands out as having rates of underweight some one half to a third those of the other regions

(only 12 percent of children underweight). In terms of residence, rural children fare worse, as one might

expect, with underweight at 28 percent while for urban children it is 15.6%. Maternal education conforms

to the expected pattern with 30% of children of mothers with no education underweight, 27% of those

with primary education and 19% of those with secondary or more education. Ethnicity yields significant

differences, with groups such as the Gruma and Grussi enduring percentage of childhood underweight

double those of other groups such as the Akwapim and the Ga/Adangbe. (Note that our ethnicity recode

p. 22

category does not yield statistically significant results—in fact the range is quite narrow—suggesting that

categorical aggregations may sometimes be misleading.)

GHANA: Proportion of Underweight Children

0

5

10

15

20

25

30

35

40

45E

duca

tion

Eth

nici

ty

Gen

der

Reg

ion

Res

iden

ce

Source: Ghana DHS 1998

Child mortality rates: Inequality in childhood mortality is closely aligned with differences in

education and place of residence: more highly educated women and urban dwellers have much lower

mortality, and boys experience somewhat higher rates than do girls.

Unexpected results

Age at first marriage: The age at first marriage does not vary by educational level, but does vary

significantly by ethnic group and by region (ranging from 16 to 18 years of age for both strata).

Immunization: The four immunization indicators are also stratified by ethnicity, with the

dominant group consistently exhibiting higher rates than the not dominant group. Educational level of the

mother also stratifies all four immunization indicators (DPT 1, 2, 3 and measles) with the expected pattern

of no education conferring greater risk of no immunization, primary education, slightly better and

secondary or more education yielding the highest rates of immunization. It’s important to note that the

range of values for the immunization indicators is between 60 percent and 95 percent-- a tighter, and

higher range of values as compared to delivery attendance by SBA (25 percent to 65 percent). In urban

areas, girls were more privileged in immunizations than boys. To know whether this is related to

education we would need a further dimension of analysis.

p. 23

DOMINICAN REPUBLIC

The Dominican Republic DHS did not collect information on ethnic or racial differences, nor was

a wealth index available for analysis.

Expected Results

Access to a health facility: Access to a health facility is significantly stratified by the level of

maternal education, with women with no education having lower access (53 percent) than those with

primary (64 percent) or secondary or more (66 percent). In addition, this indicator is stratified by region

with a range from 55 to 73 percent.

AIDS Knowledge: AIDS knowledge questions were significantly stratified by education, as well,

and residence. Knowledge of AIDS varies by region as well, with a range from 78 to 96 percent for the

first AIDS knowledge indicator. This finding is interesting for it shows that even in a country with a very

high national average (95 percent), knowledge indicators (and other health indicators) can still be quite

inequitably distributed.

Unexpected Results

CPR (any and modern) – Surprisingly, the CPR modern method and any method indicators

actually decline as education increases, and the differences are statistically significant. Among women

with no education, CPR is significantly higher in urban areas, but among women with primary education,

use rates are slightly but significantly higher in rural areas [P -value: 0.03]. CPR decreases significantly

with education at all levels in urban areas, and from primary to secondary in rural areas.

Skilled birth attendant: Despite relative equity in delivery by SBA in terms of maternal

education and the urban versus rural residence, region reveals gaps in SBA, ranging from Valverde (70

percent) to Monsenor Nouel (89 percent). Still, it is worth noting that SBA is at a relatively high absolute

level in the Dominican Republic, and certainly more equitably utilized than is the case in many other

countries.

Immunization: Among women with no education, DPT3 is higher for sons, but measles is higher

for girls. For secondary-schooled women, there is a weak DPT3 preference for daughters. For the primary

educated, there is little difference.

Underweight: Of the percentage of underweight children, the difference between primary and

secondary schooling of mothers is much greater than that between no education and primary. In

vaccinations, the differences between each level of maternal education are comparable.

p. 24

TAJIKISTAN

Note that this country is the only country for which a MICS dataset was used. Unfortunately this

survey did not allow disaggregation by ethnic group.

Expected Results

Age at first marriage (and first intercourse): Education makes a great difference in age at first

marriage: 16 years of age for women with no education and 20 years of age for women with secondary or

more education.

AIDS knowledge: The dramatic differences in AIDS knowledge occur at the level of tertiary

education. In addition, the Duchanbe region stands out as having levels of AIDS knowledge far above the

national level. Rural populations have much lower levels of AIDS knowledge and wealth by quintile

does not make a very big difference except if you are in the richest group in which case knowledge is

greater.

Skilled birth attendance: There are clear differences between secondary and tertiary educational

levels for both ‘delivery assistance by doctor’ and ‘delivery assistance by skilled attendant’ (no data for

those with no education). Leninabad stands out as having delivery assistance at 90 percent by skilled

birth attendant (national average is 71 percent). Wealth does make a difference whereby 55 percent of the

lowest quintile and 87 percent of the highest use SBAs. And the rural/urban differential is 68 versus 84

percent.

Immunization: The regional differences in immunization are the most striking in Tajikistan. For

DPT1, 2 and 3 as well as measles, Leninabad has coverage levels above ninety percent, while in Khatlon

district these levels are between 78 and 88 percent and in RRP they are lower—between 66 and 79

percent.

CPR: Overall, CPR-condom is zero. For CPR-modern method (MM), there is a clear educational

gradient, with those with no education at 16 percent CPR with a modern method (insufficient numbers for

primary education), 26 percent for secondary education and 41 percent for tertiary. GBAO region stands

out at 55 percent CPR coverage with a modern method while the other regions range between 22 and 33

percent. By quintile, CPR-MM is relatively equitably distributed though the richer groups have ten

percentage points higher CPR-MM.

p. 25

Unexpected Results

Skilled birth attendant: Surprisingly, delivery by a skilled birth attendant is quite equitable in

terms of level of education (74 percent for those with no education, 77 percent for those with secondary

or more). There is no difference between rural and urban levels of SBA.

Immunization: For immunization by wealth quintile, no clear pattern emerges from the Tajikistan

data. In fact, the second quintile (or second poorest group) seems to have higher levels of DPT coverage

than the third, fourth and fifth quintiles. While a clear stepwise pattern ‘up the wealth ladder’ is missing,

the poorest group does have immunization levels slightly below the national level. For education, there is

not enough data on those with lower than secondary education to make an assessment. For gender, girls

have very slightly higher levels of immunization.

What is striking is that in fact rural areas appear to have very slightly higher levels of DPT

coverage—and increasingly higher as you move from DPT1 to 2 to 3 (though it is not clear whether this

is statistically significant).

VI. Discussion

A. Key Findings

The first striking result from this analysis of a few select countries is that for most countries, the

majority of the social stratifiers chosen yield statistically significant differences within all of the child,

maternal and reproductive health indicators. This finding validates the importance of assessing health

equity across dimensions other than simply wealth quintiles. Wealth, ethnicity, educational level of the

mother, gender, region and urban versus rural residence are important dimensions of health inequity, with

different implications for policy. Our analysis strongly suggests that reliance on single indicators alone—

and certainly national level averages—would lead to limited, and perhaps even misguided,

recommendations for policy.

But the importance of these social stratifiers must be understood as more than a statistical

exercise in mapping out health gaps. The significance of social processes in determining who is healthy

and who is sick, and more profoundly who lives and who dies, is a fundamental challenge to health

policy. The current global focus on pro-poor health policies “limits intervention to the end of the social

production chain” leaving out many other core social processes that generate health inequities (Vega and

Irwin 2004). The global trend toward equating the ‘health equity agenda’ into ‘pro-poor’ health policies

is a misleading oversimplification. The causes of inequities in health are multifactorial, and the solutions

are intersectoral. Oversimplification will address only pockets of the problem. This analysis was

undertaken to reduce that oversimplification while not introducing complexities that make the analysis

unrepeatable where it is needed.

p. 26

The appeal of the Millennium Development Goals is that they bring many of these social

processes to the same table—education, gender, health, water and sanitation, environmental issues and

poverty are seen as interlinked. Many of the Task Forces addressing these issues have noted that

improvement in one MDG is dependent upon improvement in several of the others, but more attention

must be paid to the synergies between these important goals. In addition, certain issues like ethnic

differentials and ethnic discrimination are indeed core social processes which play a role in poverty,

health exclusion and social marginalization—must be fully explored in developing MDG-relevant and

equity-sensitive Poverty Reduction Strategies.

The other key, and related, point in this analysis is that inequities in health are complex and

interactive. Health exclusion is often a result of multiple and overlapping forms of social exclusion as

well as differences in health system infrastructure that varies by region. One cannot draw inferences

about the nature or extent of inequities in the health system from a single indicator, as illustrated in Figure

3. Nor can we assume that the groups that are disadvantaged in one indicator are necessarily the same

groups disadvantaged in another, as shown in Table 1. The full picture is complex, and resources must be

garnered to get an accurate picture of health exclusion in each country context.

Nevertheless, we are able to discern some common trends. We have seen that increases in levels

of maternal education consistently and positively influences maternal and reproductive health, as well as

child health, indicators. Ethnicity proves to be a very powerful stratifier of health access and outcomes,

particularly in Ethiopia and Kenya (though data on ethnicity was not available for all the countries). The

prevalence and extent of ethnic inequities in health suggests that this area should be prioritized in research

and in policy. Gender did not prove to be a major stratifier of child health outcomes (though we note that

had a South Asian country been included in this analysis the results would have been different). In

general, the urban advantage held true, though as noted above, important inequities within urban areas are

not revealed by the kind of data reported in DHS and MICS. In some cases, immunization levels were

higher in rural as compared to urban areas. Finally, regional differences were usually highly statistically

significant and often profound. Further explanation of the root causes of these differences are necessary,

for example whether these differences are confounded by ethnicity and poverty (and they most likely are)

and/or confounded or simultaneously caused by a lack of roads, landlocked or coastal isolation, arid

climate, and other region-specific causes.

Whatever the causal factors and the eventual policy choices made, the uncovering of severe

disadvantage in certain population groups necessitates a clear plan for mitigating or eliminating the

health exclusion suffered by certain population groups. Such a plan should be an integral part of Poverty

Reduction Strategies.

p. 27

B. Limitations

The purpose of this analysis is to make an argument about the importance of disaggregating data

when tracking the MDGs. From the analysis of these six countries, it is clear that national figures obscure

a very complex reality within countries, with regional, ethnic, residence-based, gender and wealth-based

inequalities in health stratifying health opportunity. Despite the richness of the data in the DHS and

MICS, we note that this brief snapshot of health inequalities is put forth as a starting point for discussion,

and is not intended to form the complete baselines for the countries we studied. There are at least four

important limitations which should be addressed: 1) The data sources consulted should certainly be

augmented (note especially the need for intra-urban inequities and refugee studies); 2) The stratifiers used

may need to be changed or augmented, particularly due to the need for occupational analysis in many

countries; 3) Measures of the health system are wholly inadequate; and 4) Perhaps most importantly,

tying the data to policy will involve debate and discussion as well as additional qualitative and

quantitative research within countries (see Section D below).

First, we used the data available to us, but reducing future data gaps will go a long way to

monitoring and reducing health disparities. As stated earlier in the paper, DHS and MICS were chosen as

data sources because, in practice, they may be the only reliable data source for some of these indicators in

very low resource or post-conflict settings. However, other (and additional) data sources may be more

appropriate to track all health indicators, or specific ones. For example, HIV/AIDS prevalence cannot be

ascertained from most of DHS and MICS datasets and yet will be an essential health indicator to monitor

in equity terms. Important additional population-based surveys include the Center for Disease Control’s

(CDC) International Reproductive Health Surveys (IRHS) as well as the World Bank’s Living Standards

Measurement Surveys (LSMS) (though it’s worth noting that the number of countries covered is much

smaller than either MICS or DHS). Importantly, countries have their own unique population based

surveys such as Indonesia’s Family Life Survey (IFLS) that provide a wealth of information on health and

other social indicators that is explicitly tailored to national circumstances.

As important as population based surveys are, even in settings with very incomplete vital

registration systems, sub-sampling from the vital registration system can provide valid results and would

be an important complement to population based surveys (Vega 2004). In data poor countries,

demographic surveillance system (DSS) data collect information on a continuous basis on health and its

determinants, as a vital registration-like proxy for field research stations. Unlike vital registration data,

DSS data have the intent of evaluating health interventions aimed at reducing socio-economic

differentials in mortality and morbidity (Ngom et al 2003). Lastly, facility-based surveys, reviews of

national health accounts and health workforce assessments will be important added elements to complete

the picture of how resources are allocated and whether they are allocated based on need.

p. 28

In addition, the quantitative results reported in this paper may not be able to fully elucidate the

nuances of complexity of health systems and health-related behaviors. More in-depth and multivariate

quantitative analysis would assist in the clarification of causal pathways that lead certain groups to be

disadvantaged relative to others. Qualitative studies will play an important complementary role.

Behavioral research, quality of care, levels of trust and cultural norms are obvious aspects of a complete

picture of a health system

Certain population groups may be completely overlooked in analysis that relies about either

population-based surveys or, in some cases, even vital registration. For example, refugee populations fall

outside the purview of these data sources, and yet, in many countries represent sizable populations with

poor health outcomes and inadequate access to health care. It is incumbent upon host governments and

donor agencies to ensure that refugees’ health outcomes and access to health care is tracked—as a matter

of equity and of human rights. Other groups like orphans and ethnic or linguistic minorities are often not

well represented in population-based surveys. Surveys must overcome long-standing shortcomings in the

sampling frame such that vulnerable populations are fully assessed. The efforts of MICS to include

orphans and of DHS to include family structure and unmarried women are important to capturing all

population groups.

In addition, even disaggregating data by region or by urban versus rural is often far too coarse a

level of analysis to highlight the disadvantages suffered by particular population groups—including urban

slum dwellers. Increasingly, some of the most profoundly dire circumstances coexist side by side with

great advantage. For important data on these populations, rapid urban assessments may provide snapshots

of gross inequities in health. In addition, ongoing surveillance or monitoring that is linked to policy, such

as is evidenced in the ‘Equity Gauges’ will provide city-specific data geared toward policymakers

(McCoy et al 2003).

Reducing inequality in health will almost certainly require some geographic targeting of policies.

This analysis showed how important regional differences, but could not ask more thorough-going

questions about those geographic areas or spatial relationship, because not all of the surveys under study

included geographic identifiers. Even in the absence of information—including geospatial information—

about health clinics, being able to georeference survey information facilities offers much more flexibility

than is otherwise possible. Thus, geographic identifiers should be added to all surveys, including all the

MICS and the DHS that have not yet been geocoded.

Second, a word about the stratifiers and indicators used. For certain indicators, stratification by

age may be another important aspect of an assessment of inequalities in health. Particularly in an era

when the adolescent cohort in the developing world is the largest ever, the specific vulnerabilities of

adolescents and the unique needs of vulnerable adolescents will need special consideration. For example,

p. 29

our analysis of contraceptive prevalence rates (CPR) may have benefited from disaggregation by age to

examine the patterns of access to contraception for adolescents and others, particularly because of

adolescents’ greater chances of maternal mortality.

In addition, occupation would have been a very important stratifier for many of the indicators

examined but was not included. The type of data collected by DHS and MICS results in information that

is not particularly useful for equity analysis; professional titles or types such as ‘sales’ could refer

to an informal street vendor or a manager of a large department store. Instead, clarity about the actual

type of work and conditions associated with them it would be useful to classify occupations differently.

Issues of greater relevance include: the continuity of work (seasonal, permanent, temporary), whether the

job is in the formal or informal sector, whether the job is menial or skilled jobs, and quite importantly in

the agricultural sector, whether the laborer is working on his/her own land or someone else’s.

Third, we note that measures of the health system are wholly inadequate. We use a measure of

access to health services that comes from a survey module in which questions about access to family

planning are being asked, an imperfect indicator. Evaluations of the health system will most likely need

to come from a variety of data sources. In an evaluation of constraints to expanding access to health

interventions, Ransom et al (2003) use several indicators to estimate health services delivery and health

systems management including the following: DPT3 coverage, nurses per 100,000 population, access to

health services (a UNICEF measure) and percent case detection of new smear positive cases to assess

management. Population based surveys may be able to get certain aspects of the health system such as

household or individuals’ distance from the health service, expenditures on health systems or judgments

about the health system, but assessments of the level of the health worker contacted or the clinical

preparedness of the worker or facility will obviously need to be gathered elsewhere. For example, Task

Force 4 of the UN Millennium Project has called for the inclusion of ‘Access to emergency obstetric care

(EmOC)’ as a necessary complement to the maternal mortality indicator. It is critical to note that DHS

and MICS do not allow an assessment of EmOC but that it must be measured through other means, for

the distribution of this element of the health system is critical to reducing maternal mortality.

The analysis presented herein is in no way meant to substitute for in-depth country-level work on

researching inequities in health using local data sources, including vital registration and census data

where it is available. In addition, as a final caveat, we note that the selection of indicators reviewed was

based upon Task Force 4 priorities (including reproductive health, child health and maternal health) and

time limitations. Clearly, having some sense of key child health conditions and care-seeking such as

diarrhea, care-seeking for diarrahea, oral rehydration therapy, etc. would be valuable additional

information. And, for example, in malarious regions, the inclusion of malaria indicators would also be an

p. 30

a priori condition of any sort of equity analysis. Put simply, equity analysis must be situated within a

clear assessment of the epidemiological and health profile of the country.

C. Health information asymmetry

Putting aside the technical aspects of measuring and monitoring inequalities in health, we must

address a broader question of health information asymmetry. Some information on population health

exists in all countries, and in many cases, donors have injected considerable resources for population-

based surveys such as the DHS and MICS. The broader question of who owns that data and to what

purposes it is put is not the primary subject of this paper, but it is certainly a highly relevant and important

aspect of health equity.

Information about disparities in health is a potent tool, and it is critical that an adequate share of

donor resources be allocated to ensure country level involvement in the surveys—from their design and

implementation, to their use and interpretation. In particular, we note that donors should fund NGOs,

academic institutions and Parliamentary groups for the policy-relevant interpretation of this critical data

on population health. Citizens (and their representatives) must be empowered with information about

their health status—and in particular, their relative health status—how well they fare as compared to what

is possible. Holding the government (and donors) accountable for wide, unfair disparities in health is

only possible with sound data and an understanding of its interpretation. Furthermore, in addition to

injecting resources into population-based surveys, donor support must be given to developing complete

vital registration systems as a fundamental part of health systems while simultaneously developing

country capacity to conduct country-owned population based surveys.

D. Setting the Stage: Adapting the MDGs for equity-sensitive monitoring

As mentioned above, this analysis does not purport to advise countries on specific amendments to

the MDGs. Rather, the paper attempts to drive home the importance of disaggregating data—and the

importance of setting equity-sensitive targets and indicators. The development of national ‘Poverty

Reduction Strategies’ should tackle the issue of inequalities in health (and other aspects of well-being)

and prioritize the ‘gaps’ that need to be closed. This paper sets forth important quantitative findings, but

the necessary work of choosing indicators and prioritizing gaps must be done in a specific country context

with the relevant country-level actors.

In a country like Ethiopia, and most under study here, regional and ethnic inequities in health are

profound. A plan to reach the MDGs should tackle these gaps head on, and set out to track health

outcomes by region and by ethnicity. Further, regions are often co-terminus with ethnic divisions or

poverty profiles, although this codetermination is only revealed by the trivariate analysis undertaken here.

p. 31

For example, measles vaccination rates seem to vary considerably significantly by wealth, but when

regions are added as substrata it becomes clear that three districts alone represent the bottom three

quintiles of the population. While wealth is an important focus, the geographic elements of poverty would

be overlooked and perhaps misdirected without the disaggregation. In keeping with the global focus on

pro-poor interventions, understanding the correlates of poverty will be an important element in reducing

it. We also note that in many countries, ethnicity is a profoundly inflammatory subject and it may not be

politically feasible to officially track health outcomes by ethnic group; often, however, geography is a

proxy for ethnic group and could be used as such.

In addition, educational attainment of mothers is a critical social determinant of most health

indicators (for both children and mothers). Thus, investments in education must be seen as having a dual

positive effect—both within the education sector as well as in health—for the education and health

MDGs. And concomitantly, health messages and programs should be tailored so as to reach less educated

mothers and their children.

Another pattern we found was that different health indicators yield different patterns of inequity.

In Ghana, health exclusion is conditioned by level of mother’s education, ethnicity, region and residence

(and quite likely wealth as well). For example, in order to create an efficient, equitable plan for access to

skilled birth attendants, one of the MDG indicators, recognition of the nature and extent of inequities in

access to this critical aspect of the health system is necessary. The Northern, Upper East and Upper West

regions have access to skilled birth attendants that is as low as one-fourth the national average and as low

as one-fifth that of other regions. Likewise, certain ethnic groups can be identified as having dramatically

lower levels of access to skilled birth attendants, and women with no education and in rural areas are

similarly disadvantaged. Yet in this same country, neonatal mortality is higher than average in the Brong

Ahafo and Central Regions. In this example, and many others, it is clear that different health indicators

yield different patterns of inequity.

We suggest that a baseline assessment of inequities across multiple indicators and stratifiers

should ultimately lead to participatory dialogue which sets the MDGs in terms which prioritize (either

specifically or generally) the disadvantaged groups. Once a baseline of these differences has been

established, the difficult work begins. What is the appropriate framing of an indicator which will allow

each country to measure, whether in five years, these different regional, ethnic and educational gaps have

closed? And what are the policies and programs that will address these critical issues? In order to

monitor the progress toward greater coverage of skilled birth attendants, for example, the MDG itself

ought to be framed so as to put priority on the groups that are lagging so far behind. Whether the groups

or regions are explicitly named in the MDG itself is up to the country, but certainly detailed tracking and

monitoring should be clear about the specific gaps that need to be closed.

p. 32

In many countries, interim-PRSPs and PRSPs have mentioned inequalities in health—often

noting regional inequalities and sometimes noting the particular vulnerabilities of the poor. However,

Poverty Reduction strategies must become far more nuanced and comprehensive on two fronts: 1) First,

their evaluation of inequalities in health—across multiple indicators and several stratifiers—must be more

thorough in order to piece together an accurate picture of health in the country and 2) Second, the linking

of specific policy initiatives (and monitoring mechanisms to assess the failure or success of these policies)

to redress health gaps must be made explicit.

Practically, our recommendations mean that countries must prioritize the worse-off groups or

ensure that they improve at the same or faster rate as the better-off groups. This does not boil down to a

‘pro-poor’ intervention here or there. Nor does it imply a set of targeting programs. Rather, we advocate

for steps toward universal health system that is designed to encompass disadvantaged groups. An

egalitarian approach is faster and more efficient to reach the MDGs than the current top-down model

where the poorest and most marginalized are relegated to the end of the queue. And, as documented in the

literature, some countries may not reach the MDGs without addressing their current levels of inequality

(Minujin and Delamonica 2003). The set of recommendations in the next section summarizes critical

steps toward setting the baseline for equity-sensitive MDGs with the goal of reducing inequities in health.

VII. Recommendations

A. For Donors, WHO, The Millennium Project and Country-level Actors

§ It is essential to measure and monitor inequities in health, across multiple indicators and different social stratifiers (i.e. not just wealth but education, gender, ethnicity and region as well). This effort must be part of poverty reduction strategies and efforts to meet the MDGs. Inequities do exist even in the poorest countries and as this analysis has shown, creating a baseline from which to monitor inequities is feasible in these countries.

§ Health exclusion and health gaps are ‘a drag’ on national health indicators but they also represent

grossly unfair circumstances, and often, violations of human rights. Monitoring health gaps—and being explicit about which policies are being implemented to reduce them—is the responsibility of both governments and donors alike. And monitoring health gaps (both in terms of gathering and analyzing data), in turn, is crucial to be able to design national health policies

§ The targets and indicators which buttress the Millennium Development Goals should be stated in

equity-sensitive terms, prioritizing progress amongst the disadvantaged groups in a particular country. And the MDG priorities should be directly linked to Poverty Reduction Strategies that outline policies and programs to close health gaps.

§ Beyond measurement of the gaps, more country-specific work must complement the analysis of

the quantitative data (such as displayed in the two-way and three-way tables here). To provide policy-relevant input, a range of standard behavioral and social science methods must be used to explain and augment the numbers. The social determinants of health gaps are complex, and solutions to close these gaps must be intersectoral.

p. 33

§ To overcome information asymmetry and to help close health gaps, others, including civil society

groups and Parliaments should be empowered to act as ‘checks’ on governments and Ministries of health. Being empowered means access to resources, facility with simple data and information and voice to bring unfair health gaps to the fore.

§ In terms of designing national health policies, these results confirm that there are many

vulnerable groups (not just the 'poor') and their different forms of vulnerability interact. Rather than having a patchwork of targeting programs, the results suggest that a universal health system that pays attention to inclusion of all population groups will likely be more efficient and equitable.

§ Population based surveys such as DHS and MICS are important, but long-term investments in

health information systems must also begin now and must be part of PRSPs and MTEFs. Donors and WHO must invest in country capacity to oversee population based surveys and health information systems.

§ Expansion of local and regional health information systems will be as important than national

health information systems. In, part this is due to the higher efficiency of gathering information at the local level. Also, in order to compare across municipalities and region, this local level system needs to be part of a national system.

B. On adaptations of population-based surveys surveys (or for new surveys)

Key data on socioeconomic status must be included and expanded to the following:

§ some measure of accumulated economic assets/wealth § ethnicity § gender § occupation/ nature of work § region (note including georeferencing) § residence

There are diverse obstacles to capturing the health status of urban slum dwellers, refugees, and orphans. Household surveys, even large ones, may not yield adequate numbers of people in certain particularly vulnerable subgroups, and special approaches may need to be designed for the following groups: § ethnic or linguistic minorities § urban slum dwellers § refugees § orphans

C. On measuring the health system

Reliable, valid metric(s) for assessing the health system must be developed and implemented to include the following: § the private sector § distribution (georeferenced clinics), capacity, etc. § human resources § access to Emergency obstetric care (EmOC) must be added in order to have adequate

information to reduce maternal mortality

p. 34

References AbouZahr, C., Wardlaw, T., et al. (2004) Antenatal Care in Developing Countries: Promises,

achievements and missed opportunities: An analysis of trends, levels and differentials, 1990-2001. WHO/UNICEF.

Braveman, P. (1998). Monitoring Inequities in Health: A Policy-Oriented Approach in Low- and

Middle-Income Countries. WHO/CHS/HSS/98.1. Geneva: World Health Organization. http://whqlibdoc.who.int/hq/1998/WHO_CHS_HSS_98.1.pdf

Braveman, P. and Gruskin, S. (2003). Poverty, equity, human rights and health. Bulletin of the World

Health Organization 81(7):539-545. Braveman, P. and S. Gruskin (2003). Defining equity in health. Journal of Epidemiology and Community

Health 57: 254-258. Castro-Leal, F., Dayton, J. et al. (2000). Public Spending on Health Care in Africa: Do the Poor

Benefit? Bulletin of the World Health Organization. 78(1): 66-74. Central Statistical Authority [Ethiopia] and ORC Macro. (2001). Ethiopia Demographic and Health

Survey 2000. Addis Ababa, Ethiopia and Calverton, Maryland, USA: Central Statistical Authority and ORC Macro.

Centro de Estudios Sociales y Demográficos (CESDEM) and MEASURE DHS+/ORC Macro (2003).

República Dominicana Encuesta de Demografía y Salud 2002. Calverton, Maryland, USA, Centro de Estudios Sociales y Demográficos, CESDEM and MEASURE DHS+/ORC Macro.

Evans, T. and Stansfield, S. (2003). Health Information in the New Millennium: A Gathering Storm?

Bulletin of the World Health Organization. 81(12): 856. Evans, T., Whitehead, M, et al. (2001). Challenging Inequities in Health: From Ethics to Action. New

York: Oxford University Press. Filmer, D. and Pritchett L. (2001). Estimating wealth effects without expenditure data--or tears: an

application to educational enrollments in states of India. Demography. 38(1):115-32. Freedman, L., Wirth, M. et al. (2004). Interim Report. Task Force 4: Child health and maternal health.

New York: UN Millennium Project. Freedman et al. (2003). Background Paper on Child health and Maternal Health . Task Force 4. New

York: UN Millennium Project. Ghana Statistical Service (GSS) and Macro International Inc. (MI). (1999). Ghana Demographic and

Health Survey 1998. Calverton, Maryland: GSS and MI. Gwatkin, D.R. (2002) Who would gain most from efforts to reach the Millennium Development Goals for

health?: An inquiry into the possibility of progress that fails to reach the poor. HNP Discussion Paper. Washington, DC: World Bank.

p. 35

Gwatkin, D., S. Rutstein, et al. (2003). Initial country-level information about socio-economic differences in health, nutrition, and population. Washington, DC, The World Bank.

Kunst and Houweling (2001). A global picture of poor-rich differences in the utilisation of delivery care.

Safe motherhood strategies: a review of the evidence. W. Van Lerberghe and V. De Brouwere: 297-316.

McCoy, D., Bambas, A., et al. (2003).Global Equity Gauge Alliance: Reflections and Early Experiences.

Journal of Health Population and Nutrition. 21(3): 273-287. Minujin, A. and Delamonica, E. (2003a). Mind the Gap! Widening Child Mortality Disparities. Journal

of Human Development. 4(3): 397-418. Minujin, A. and Delamonica, E. (2003b). Equality Matters for a World Fit for Children: Lessons from the

90s. UNICEF Staff Working Papers, Division of Policy and Planning Series (Draft, Number DPP-03).

Moyo, S. (2004) “Socio-Economic Dominance of Ethnic and Racial Groups – The African Experience.”

Human Development Report Office (Occasional Paper, Background Paper for HDR 2004). United Nations Development Programme, 2004/8.

National Council for Population and Development (NCPD), Central Bureau of Statistics (CBS) (Office of

the Vice President and Ministry of Planning and National Development) [Kenya], and Macro International Inc. (MI). (1999). Kenya Demographic and Health Survey 1998. Calverton, Maryland: NDPD, CBS, and MI.

National Institute of Statistics, Directorate General for Health [Cambodia], and ORC Macro. (2001).

Cambodia Demographic and Health Survey 2000. Phnom Penh, Cambodia, and Calverton, Maryland, USA: National Institute of Statistics, Directorate General for Health, and ORC Macro.

Ngom, P, Binka, F et al. (2003). Demographic Surveillance and Health Equity in Sub-Saharan Africa:

Past and Present Efforts. African Population and Health Research Center. Working Paper Series. Nairobi: APHRC.

Ransom, M.K., Hanson, K., Oliveira-Cruz, V., Mills, A. (2003). Constraints to expanding access to

health interventions: an empirical analysis and country topology. Journal of International Development. 15:15-39.

Sachs, J. et al. (2004). Millennium Development Goals Needs Assessments: Country Case Studies of

Bangladesh, Cambodia, Ghana, Tanzania and Uganda. Working Paper. New York: UN Millennium Project.

Segone, M. (2001). Child Poverty in Niger: Longitudinal, Gender, Spatial and Wealth Analysis Based on

DHS 1992, MICS 1996, DHS 1998 and MICS 2000. (Draft). UNICEF, October 2001. Swinkels, R. and R. Turk (2002). Achieving the Vietnam Development Goals: An Overview of Progress

and Challenges. World Bank. United Nations Children’s Fund (UNICEF). (2000). Multiple Indicator Cluster Survey (MICS):

Tajikistan. Dushanbe: UNICEF.

p. 36

United Nations Children’s Fund (UNICEF), Division of Evaluation, Policy and Planning, Programme

Division. (2000). End-Decade Multiple Indicator Survey Manual: Monitoring Progress Toward the Goals of the 1990 World Summit for Children. New York, NY, USA: UNICEF.

United Nations Development Programme (UNDP). (2003). Human Development Report 2003. New

York: UNDP. Vandemoortele, J. (2000) Absorbing Social Shocks, Protecting Children and Reducing Poverty: The Role

of Basic Social Services, UNICEF staff working papers, Evaluation, Policy and Planning Series no. 00-001. New York: UNICEF.

Vega, J. and A. Irwin. (2004). Tackling health inequalities: new approaches in public policy. Bulletin of

the World Health Organization. 82(7): 482. Vega, J. (2004). Personal communication. Victora, CG, Vaughan, P. Barros, FC, Silva, AC, Tomasi, E. (2000). Explaining trends in inequities:

evidence from Brazilian child health studies. The Lancet. 356(9235). Wagstaff, A., Bustreo, F., Bryce, J., Claeson, M., Axelsson, H. (2003). Inequalities in Child Health: Are We

Narrowing the Gap? (2003) HNP Discussion Paper. Washington D.C: The World Bank. Wagstaff, A. (2002). Poverty and health sector inequalities. Bulletin of the World Health Organization.

80(2): 97-105. Wagstaff, A. and Watanabe, N. (2000). Socioeconomic inequalities in child malnutrition in the

developing world . Policy Research Working Papers. Washington D.C.: The World Bank.

p. 37

Appendix 1. Definition of indicators used

Closest Related MDG

Variable Goal Target Indicator Indicator Definition

U5MR 4 5 13 Children under five mortality rate per 1,000 live births (UNICEF estimates)

Underweight 1 2 4 Nutrition; moderately or severely underweight, aged under 5 years.

Moderately or severely underweight is below two standard deviations below median weight for age of reference population; severe is below minus three standard deviations from median weight for age of reference population.

KNOWLEDGE OF AIDS

6 18 19b. Percentage of population aged 15-24 years with comprehensive correct knowledge of HIV/AIDS (UNICEF-WHO)

HIV knowledge, women aged 15-24 who know that a healthy-looking person can transmit HIV, percent (UNICEF-UNAIDS-WHO)

HIV knowledge, women aged 15-24 who know that a person can protect herself from HIV infection by consistent condom use, per cent (UNICEF-UNAIDS-WHO)

CPR 6 18 19c. Contraceptive prevalence rate (UN Population Division)

Contraceptive use among currently married women aged 15-49, any method, per cent (UN Population Division)

Contraceptive use among currently married women aged 15-49, condom, per cent (UN Population Division)

Contraceptive use among currently married women aged 15-49, modern methods, per cent (UN Population Division)

MEASLES 4 5 15 Proportion of 1 year-old children immunized against measles (UNICEF-WHO)

DPT Proportion of 1 year-old children immunized against Diphtheria, Pertussis and Tetanus.

DELIVERY CARE

5 6 17 Proportion of births attended by skilled health personnel (UNICEF-WHO)

Refers exclusively to people with midwifery skills (for example, doctors, midwives, nurses) who have been trained to proficiency in the skills necessary to manage normal deliveries and diagnose or refer obstetric complications.

AGE AT FIRST INTERCOURSE

Average age at first intercourse.

AGE AT FIRST MARRIAGE

Average age at first union.

HEALTH SYSTEM USE

Visited a health facility in the past 12 months.

p. 38

Appendix 2. Definition of Stratifiers

Stratifier Definition Adjustments

GENDER Gender of child

EDUCATION Mother's highest level of education Grouped into None, Primary and Secondary. Non-formal curricula and strictly religious education excluded

RESIDENCE Urban or Rural ETHNICITY Country-specific Uses standard DHS recodes (not available in

MICS) Ethnicity (recoded by group dominance)

Country-specific Divided into dominant, non-dominant, and secondary dominant (where available; see p. 15)

WEALTH Quintiles of wealth (country-

specific)

Ranges from 1 = "poorest" to 5 = "richest"

POVERTY

LINE

Above or below national poverty

line

Data from World Development Indicators 2003 applied to wealth index

REGION Country-specific

p. 39

Appendix 3. Two major population based surveys: DHS and MICS4

DHS MICS

Oversight agency: USAID UNICEF

Key features: Household questionnaire and women’s questionnaire Focus on nutritional status and anemia, mother’s health, fertility, health of the children and family planning Some DHS include service availability questionnaires (infrastructure) Some include biomarker information (HIV, syphilis, etc,

More child-labor and the same education information Several modules which can be omitted or included (gives countries flexibility) Includes a number of very poor countries not in DHS Based on human rights principles Includes the same specific attention to malnutrition and mortality disaggregated by age, etc.

Number of countries: 70 countries 65 countries (MICS2)

Sample sizes: 5000-30,000 hh, nationally representative

4000-5000 hh, nationally representative (some are larger)

Strengths: -Geocoding available for many current surveys -Qualitative data, biomarker data, service provider assessments

-Lower cost, faster turnaround time -UNICEF provides less direct technical assistance, it carries out capacity building and training

Weaknesses High cost ($.5M) and long turnaround time

No geocoding yet. Information about HIV/AIDS is not as complete as in DHS. Information on ethnicity is going to be included in the next round

4 Adapted from Komarecki (2003).