SOCIO-ECONOMIC DETERMINANTS OF CHILD GROWTH
IN FIVE DEVELOPING COUNTRIES: IMPLICATIONS FOR HEALTH
AND DEVELOPMENT

by
Katherine S. Mohindra
BSc Hons, Carleton University, 1996
THESIS SUBMITTED IN PARTIAL FULFILLMENT OF
THE REQUIREMENTS FOR THE DEGREE OF
MASTER OF SCIENCE
m
COMMUNITY HEALTH
©Katherine S. Mohindra
June 1999
All rights reserved. This work may not be
reproduced in whole or in part, by photocopy or
other means, without the permission of the author.

11

ABSTRACT
Malnutrition is one of the five major killers of children under five in the developing
world (Gove, 1997). Using DHS data, and anthropometric indicators, this study describes
changes in the nutritional status of children in Northeast Brazil, Guatemala, Mali, Uganda
and Zimbabwe between the 1980's and 1990' s. Results indicate that although there is a
significant decrease in malnutrition in Northeast Brazil, there has been little progress and
even deterioration of nutritional status in Guatemala, Mali, Uganda and Zimbabwe. There
was a notable increase in wasting in all five countries, predominantly in rural areas. This
study also examines several socio-economic determinants of growth. These determinants
are compared both over time and across countries. A series of logistic regressions were
performed based on a conceptual model of malnutrition. Results indicate that maternal
education is a key predictor of stunting and radio ownership is an important predictor of
underweight status. Other significant variables include place of residence, sanitation
facilities, source of drinking water, immunization status and diarrhea prevalence.
Reducing malnutrition is essential for social and economic development. Malnutrition
consists of a complex array of determinants that requires comprehensive strategies to
treat root causes. A concentrated effort is needed from national governments and the
international community to promote the implementation and practical use of primary
health care.

lll

TABLE OF CONTENTS
Abstract.. .... ......... .... .............. ......... .............. ..... ............ .... ................ ..... ....... ... ..... .... 11
Table of Contents.. ................................................................................................... . 111
List of Tables................. .. ..... ............................................................................. .... .... v
List ofFigures......... ......... ..... ...... ....... ........ ......... ..... .... ..... ............... ............. .......... .. v1
Acknowledgement......... .... ..... ................. .......... .... ............. ... ......... ..... ................. ..... v11

Chapter One: Introduction and Overview
1.0 Overview.............. ......... ........... ..... ................ ..... ................ ..... ........ ..... ......... ... .. 1
1.1 Health and Development. .... ........... .... ...... ............... ................... ............. ...... ..... 3
1.2 Maternal-Child Health ........................................................................................ 12
1.3 Malnutrition............ ....... .......................................... .. ........................................ . 13
1.4 Summary ... .. ... ... .. .. ... .. .. .. ... .. .. .. .. .. .. .. .. .. .. .. .. .. .. .. ... .. ... .. .. .. ... .. .. .. .. .. .. .. ... .. .. .. .. .. ... .. . 19
Chapter Two: Literature Review
2.0 Introduction ......................... ............ ......... .... ....... ....... ........... ............... .............. 21
2.1 Statistical Association between Nutritional Indicators.... ...... .............. .... .......... 21
2.2 Global Trends ofNutritional Indicators .... .......................... .... ........................... 21
2.3 Determinants of Growth ...... ....................................................... ..... ............. ...... 26
2.4 Summary .... ..... .. ................... .................. .... ................... ..... ......... .......... ....... ...... 37
Chapter Three: Methods
3.0 Source ofData.................................................................................................... 39
3.1 Sampling Procedures...... ... ....... ...... ..... ....... ............................. ... ................ ........ 40
3.2 Country Profiles... ...... ...... ....... ... ..................................... ...... .... ..... ........... ......... 41
3.3 Description of Sample.............................................. .......................... ..... ........... 46
3.4 Description ofVariables ............................................. ....... ..................... ............ 46
3.5 Data Analysis Procedures ................................................................................... 48
3.6 Limitations of Analysis........................ .......................................... ........ ............ 51
Chapter Four: Results
4.0 Introduction ........................................................................................................ 53
4.1 Compilation of Tables and Figures.................................... .......... ...................... 53
4.2 Descriptive Statistics.......................................... ........................ ........................ 54
4.3 Univariate Analysis......... ........... ........................................ ................. ......... ...... 59
4.4 Summary of Descriptive and Univariate Results ............................................... 65
4.5 Multivariate Results......... .. ...... .... .......................... .......... ....................... ........... 66
4.6 Summary of Multivariate Results .............. ...... .................... ............................ .. 75

IV

Chapter Five: Discussion and Conclusion
5.0 Introduction.... .. ... ... ... ......... ............................. ... ......... ...... ........... .... ............ .... ..... 77
5.1 Discussion ofResults. .... ................................................ .... ................ ......... ....... ... 77
5.2 Linking Study Results with Health and Development.. ................ ................ ......... 86
5.3 Recommendations for Policy.................... .... ......... ........................ ............ ............ 91
5.4 Recommendations for Future Research .. ..... .... ...... .............. .......... ......... 92
5.5 Conclusion....... ..... ... ....... ..... ...... ........ ................. ..... .... ..... ............ ........ .... .... ......... 93

References. ... ....... ..... .... ... .......... .... ....... ......... .......... ........ ............ ........................ .........

94

Appendix A: Description of DHS Surveys.. ............... ..... .. ...................... ........... .......... . 102
Appendix B: Logistic Regression Results......... ...... ......... .... ............. ....... .... 103

v

LIST OFT ABLES

Table 1.0: The Essential Components of Primary Health Care................... ....... ........ 5
Table 1.1: Five Major Principles of Health Promotion.... ............ ......... .... ..... ........... .. 6
Table 2.0: Median Prevalences ofMalnutrition and Estimated Under-five
Mortality Rates (U5MR), by Region. ................. ......... .......... ....... .......... ... 23
Table 2.1: Correlation of the Prevalences of Wasting on Stunting by Region..... ...... 23
Table 2.2: Global Estimates for the Prevalence and Number of Underweight,
Stunted, and Wasted Children in Developing Countries.......... ..... .... ..... .... 25
Table 2.3: Regional Estimates for the Prevalence of Underweight, Stunted, and
Wasted Children in Developing Countries.... .. ....... ...... .. .... .......... .............. 26
Table 3.0: Human Development Index (HDI) Ratings for all Five Countries ............ 43
Table 3.1: Selected Child Health Indicators ................... ......... ....... ........... ............... ... 45
Table 4.0: Anthropometric Indicators ... ...................... ......... ........ ............ .................... 55
Table 4.1: Combination of Anthropometric Indicators.............. ............ .... ............. .... 56
Table 4.2: T -test Results for Changes in Nutritional Status over Time ................ ....... 62
Table 4.3: T -test Results for Changes in Nutritional Status over Time:
Rural versus Urban Residence......................... ................. ............. ............ 63
Table 4.4: T -test Results for Changes in Nutritional Status over Time:
Male versus Female Child... ........ ................... ............................................ 64
Table 4.5: Correlation ofthe Prevalences of Anthropometric Indicators,
by Survey...... ................... .... .............. .. .............. ........ ....... ....... ....... ............ 65
Table 4.6: Predictor Variable Coding ......... ............. ..... .... ......... .... ........ .... .......... .... .... 72
Table 4.7: Odds Ratios for Stunting ......... ..... ........ .... .. ... ......... .... ............. ....... ......... ... 73
Table 4.8: Odds Ratios for Underweight Status .......... ..... .... .. ... ........ ...... ...... ........... ... 74

Vl

LIST OF FIGURES

Figure 1.0: Causes of Child Malnutrition........................... .......... ............... ........... .....

16

Figure 3.0: Conceptual Framework of Risk Factors for Malnutrition in
Developing Countries..................... ... .................. ............. ...... ............. ......

50

Figure 3.1: Logistic Regression Models. ..... ... ..... ... ...... ... ...... ... ................ .......... ... ......

50

Figure 4.0: Anthropometric Indicators for Urban versus Rural Residence.. .. ............

57

Figure 4.1 : Anthropometric Indicators for Sex of Child....... .................... .............. .....

58

Figure 4.2: T-test Results: Urban versus Rural Residence (Intra-Country
Comparisons). .. ... .................. ................. ............ ... ......... ......... .... ........ .......

60

Figure 4.3: T-test Results: Male versus Female Child (Intra-Country
Comparisons)........... ............ .... .... .... .. ........ ...... ... ........ ........ ...... ....... .... .....

61

Figure 4.4: Summary of the Significant Predictor Variables for each
Logistic Regression Model: Stunting Models......... .... ................... ...... ......

69

Figure 4.5: Summary of the Significant Predictor Variables for each
Logistic Regression Model: Underweight Models. ... .............. .... ...............

70

Vll

ACKNOWLEGEMENT
This thesis involved the collaborative efforts of an excellent thesis committee
spanning three UNBC departments, social work, economics and psychology. This multidisciplinary approach illustrates the importance of cooperative work in international
community health issues. I am deeply indebted to my supervisor, Dr. Kwong Leung Tang
who taught me the value of applied quantitative research and how to incorporate it into
the broader development context. I am grateful for the warmth and support of
Dr. Tomson Ogwang and his guidance during statistical analysis. I also appreciate Dr.
Han Li's useful comments and her assistance in helping me integrate important health
theories into my thesis.
I would like to thank Evaon C. Wong-Kim for her time and expertly advice as my
external examiner.
Millione grazie per Dr. Bruno Zumbo, una excellente professore di statistica e una
persona molto simpatico.
I would also like to thank Annette Browne, who expanded my thinking with
respect to health issues and Dr. Fiona MacPhail for increasing my understanding of
development issues.
Special thanks to Dr. Chris Opio for his strength and words of wisdom.
Finally, many thanks to Joanne, Joy, and Adrienne and to my fellow Hooch
drinkers. You are all very star!

1

Chapter 1: Introduction and Overview
The 'food problem' has slowly yielded to the 'public health problem'
on the agenda of international discourse.
-Partha Dasgupta (1993)

1.0 Overview
Purpose of Study
The purpose of this thesis is to describe patterns and trends in the nutritional
status of children in developing countries throughout the past decade using a multicountry approach. In addition, this study provides insight into the socio-economic
determinants of growth from a global perspective.
The primary reason for using a comparative analysis in this thesis is that
nutritional patterns vary among global regions, suggesting various causes of malnutrition
(Victora, 1992). To gain a better understanding of the socio-economic factors influencing
nutrition it is necessary to examine different regions of the world.
Approximately 12 million children under the age of 5 die each year, 99% of these
deaths occur in developing countries (Gove, 1997). Malnutrition, one ofthe 5 major
killers of children, is associated with 54% of these deaths. Malnutrition prevents one in
three children in the developing world from attaining full physical and mental potential
(Madzingira, 1995). The World Summit for Children set specific goals for improvements
in maternal and child health for the decade 1990-2000 (United Nations, 1990). One of the
major goals was to reduce severe and moderate malnutrition by half of 1990 levels
among children under 5. As the year 2000 approaches, it is important to examine the
changes and progress that have occurred throughout the last decade. Close monitoring of

2
each country is important to determine the influencing socio-economic determinants of
health. This information can help to shape health policies and develop effective
community health programs.

Objectives
There are three main objectives of this study:
•

To describe changes in nutritional status of children under 5 in developing countries
between the 1980's and the 1990's.

•

To determine if socio-economic factors are affecting malnutrition over time.

•

To compare socio-economic factors across countries.

Methods
This study used data from the Demographic and Health Surveys (DHS). Five
countries were selected. Each country had one survey conducted in the 1980's and
another survey in the 1990' s. Descriptive and univariate statistics were utilized to
describe changes in the nutritional status over time for all five countries. Logistic
regression techniques were subsequently used to examine patterns and identify key
determinants of malnutrition.

Organization of Study
The structure of the thesis is as follows. The first and present chapter consists of
three sections. The first section outlines and defines key concepts in the field of health
and development and develops a theoretical framework for the thesis. The need for

3
quantitative health research in developing countries is also addressed. The second section
examines maternal-child health with a focus on the role of women in development. The
final section examines the epidemiology of malnutrition, the relationship between health,
development and malnutrition, the causes of malnutrition, and methods of measuring
nutritional status.
Chapter Two is the literature review. The chapter discusses the statistical
association between anthropometric indicators and outlines global trends in the
nutritional status of children. The literature pertaining to the specific socio-economic
determinants of growth used for this study is also reviewed.
Chapter Three explains the methods that were utilized for the study including a
description of the source of data, sampling procedures, description of sample, description
of the variables, and data analysis procedures. The countries selected for the study are
profiled and the limitations of the study are also discussed in this chapter.
Chapter Four reports the findings of the study. The results are presented in several
tables and figures and the main patterns and trends are summarized.
Chapter Five discusses the findings and places them in the context of other
research. In addition this chapter integrates the findings of the study with the role of
health in development. Recommendations are made for future research and policy
directions.

1.1 Health and Development
Relationship between Health and Development
The relationship between health and development is complicated by the evasive
definitions of both concepts. Definitions for health range from narrow medical views to

4

an 'ideal state' (Beaglehole & Bonita, 1997). The more narrow definition of health stems
predominantly from medical professionals, who see health purely as 'the absence of
disease' and should be controlled solely by those in the health field. The more widely
cited definition understands health as an ideal state. This is the definition that is written in
the preamble of the constitution of the World Health Organization (WHO, 1946): "health
is a state of complete mental, physical and social well-being and not merely the absence
of disease or infirmity." In the context of the WHO definition, health of human
populations involves multiple interactions of social and economic factors, and the
physical and biological environments (Phillips & Verhasselt, 1994). The nature in which
health is defined is important for it often directs the type of health policy that is adopted.
For instance, multi-sectoral approaches that address comprehensive lists of health
determinants often stem from wide range definitions ofhealth, while, using specific
technologies to target selected diseases stems from narrow medical definitions (Rifkin &
Walt, 1986).
A new era ofhealth began in the 1970's when health became an important issue in
development. The WHO and United Nations Children's Fund (UNICEF) noted that the
promotion of health required the recognition of social, economic and environmental
determinants of health, and therefore changes in health status could not be accomplished
through the health care system alone (Asthana, 1994b). Primary health care (PHC), a
strategy for health development, arose in 1978 at the Alma Ata conference. Primary
health care is defined as:
essential health care made universally accessible to individuals and
families in the community by means acceptable to them, through their

5

full participation and at a cost that the community and country can
afford. It forms an integral part both of the country's health system of
which it is the nucleus and of the overall social and economic
development of the community (WHO, 1978, p.2).
The slogan 'Health for all by the year 2000' was subsequently coined by WHO,
based on the principles of PHC and guided the philosophy in health care development
throughout the 1980's and 1990's. This approach was designed to move away from the
previous curative medical strategies to a community-oriented approach emphasizing
health promotion and prevention. The essential components ofPHC are summarized in
Table 1.0.

Table 1.0 The Essential Components of Primary Health Care
1. Health Education
2. Environmental sanitation, especially of food and
water
3. The employment of community or village health
workers
4. Maternal and child health programs, including
immunization and family planning
5. Prevention of local endemic diseases
6.

Appropriate treatment of common diseases and

injuries
7. Provision of essential drugs
8. Promotion of nutrition
9. Traditional medicine

Source: Alma-Ata Conference documents (1978)

Health promotion emerged in the 1990's as a unifying concept bringing together
various fields of study (Macdonald & Bunton, 1992). The idea of health promotion is
based on numerous developments in the health field including the strategies of primary

6
health care. Similar to the ideas ofPHC, health promotion rejects the traditional medical
view of health for a broader definition, which includes social and economic aspects.
Health promotion also seeks to empower and increase community participation. The five
principles of health promotion are summarized in Table 1.1.

Table 1.1 Five Major Principles of Health Promotion
1. Health promotion involves the population as a whole in the
context of every day lives, rather than focusing on people at
risk for specific diseases.
2. Health promotion is directed towards action on the
determinants or causes of health.
3. Health promotion combines diverse, but complementary,
methods or approaches.
4. Health promotion aims particularly at effective and concrete
public participation.
5. Health professionals-particularly in primary health care-have
an important role in nurturing and enabling health
promotion.
Source: Kickbusch (1986)

One of the primary criticisms surrounding health promotion is that it is based
predominantly on western ideology. Most of the work pertaining to health promotion
evolved out of Canada 1 and Great Britain. It is therefore questionable whether this theory
can be transferred to developing countries. There are two main arguments presented by
Kickbusch (1986) which suggest that health promotion can be applied worldwide.
Kickbusch argues that the philosophical underpinnings are relevant to both developed
and developing countries. The difference lies in the 'entry point' of health promotion.

1

Health promotion first appeared as a term and a concept in A New Persepctive on the Health of
Canadians, a report published in 1975 by Marc Lalonde, the Canadian Minister of National Health and
Welfare. In 1986, Jake Epp, Canadian Minister of National Health and Welfare published Achieving Health
for All: A Framework for Health Promotion. The goal of this report was to attempt to translate theory into
practice.

7

Developed countries require health promotion to encourage preventative participation
and healthy public policies to decrease reliance on the medical establishment. In the
context of developing countries, health promotion can be utilized to formulate health
policies and primary health care. The second argument for using the theory of health
promotion in developing countries is that it is important to unify the experiences of
developed nations. Encouraging international health promotion can help to promote
partnerships and to link global health strategies.
A second major criticism ofhealth promotion is the inability of translating theory
into practice. Becker (1986) suggests that health promotion in practice has led to
exploitation of the examination of risk factors, provision ofunreliable and inconsistent
'health' advice to the public and the encouragement of victim-blaming, while ignoring
important social, economic and environmental factors related to health. It has also been
suggested that health promotion fails to empower populations and "effectively constructs
the individual subject as a "health consumer" in accordance with the model of consumer
capitalism" (Grace, 1991, p.329). It is believed, however, that if the focus can remain on
the core philosophy of health promotion and realistic goals can be set, the application of
health promotion will have significant improvements on population health for all nations.
PHC was criticized for being too idealistic and not appropriate during time of
economic crisis 2 (Asthana, 1994b). Therefore, shortly after Alma-Ata, the concept of
selective primary health care (SPHC) was introduced. In contrast to the comprehensive
nature ofPHC, SPHC strives to use specific and economically beneficial technologies to

2

The economic recession during the late 1970 's and 1980's led to economic crisis in many developing
countries. The World Bank and the International Monetary Fund imposed structural adjustment programs
in heavily indebted countries which led to drastic cuts in social and health expenditures (Asthana, 1994a).

8
reduce specific diseases in targeted populations (Rifkin & Walt, 1986). SPCH maintains
the traditional medical view that improvements in health come primarily from medical
interventions without the collaboration of professionals outside the health field.
SPHC uses low-cost measures to improve the health of targeted groups, children
in particular. Although proponents of SPHC believe that it is the most progressive way to
achieve significant improvements in health, others have argued that SPHC is a 'step
backwards' in international health policy (Asthana, 1994; Rifkin & Walt, 1986). Critics
argue that SPHC attempts to improve health by controlling disease in isolation, while
failing to address root causes of ill health, such as poverty and inequality. SPHC uses a
technological and cost-oriented approach that prevents community participation, one of
the crucial ideas put forth by PHC. Since the 1980's the ideals ofPHC have not been put
to practical use and have been generally neglected for more selective national health
policies (Asthana, 1994b).
The definitions of development are diverse and far ranging. The manner in which
development is defined is reflective of political and inherent beliefs of the individual.
Esteva (1992) traces the meaning of development from the post World War II era to the
present, associating development with growth, evolution, and maturation. He criticizes
the misuse and misinterpretation of development, particularly with respect to the
predominant Western hegemony surrounding development issues. Although traditionally
viewed as economically based, development in its broad definition includes elements
such as distribution of resources, access to opportunities and political and human rights
(Phillips & Verhasselt, 1994). Development has also been described as a process that
increases human choices (Haq, 1995), and aspects ofhuman dignity (Seers, 1969).

9

Although there continues to be wide debate surrounding the meaning of development,
there is a general consensus that development is a process that improves the quality of
human life (WHO, 1992, as cited in Phillips & Verhasselt, 1994).
One of the most controversial issues in development is the longstanding debate
between economic growth and human basic needs (King, 1998). Should developing
nations strive for swift economic growth, which may deepen inequalities, or should
governments attempt to secure basic needs for the population at the risk of increasing
inefficient economic strategies? Conventional economic theory proposes that inequalities
are a necessary consequence for economic growth. Midgley (1997), however, suggests
that 'distorted development' will follow economic growth in the absence of social
development. Distorted development, a current problem of many developing countries,
occurs when economic growth fails to reach the whole population.
Social development is "a process of planned social change designed to promote
the well-being of the population as a whole in conjunction with a dynamic process of
economic development." (Midgley, 1995, p.25). The purpose of social development is to
integrate both economic and social development to enhance human welfare (Midgley,
1997). Without such integration those individuals that are the most deprived may not
receive the benefits of economic progress. Investing in the health of people is one of the
greatest resources for socioeconomic development (Abel-Smith, 1990; World Bank,
1993). It is imperative, to make health central to development (Herrell & Mulholland,
1998).
There is a complex interrelationship between health and development involving
synergistic and reciprocal elements (Phillips & Verhasselt, 1994). Development is greatly

10
influenced by the health of the population. Poor health restricts economic growth by
reducing the availability of labour, reducing the productivity of workers, creating
dependency on others, and by wasting resources (Abel-Smith, 1990). Economic growth,
on the other hand, can improve health by providing the necessary funds to ensure such
things as safe drinking water, adequate sanitation facilities, immunization programs and
health promotion activities. The relationship between health and economic development
is difficult to generalize. Although increases in economic indicators often lead to
improved health this is not always the case. Problems arise when there is unequal
distribution of resources and the whole population does not benefit from economic
improvements. Moreover, after a minimum per capita income is reached social and
political factors become more influential in determining health status than national wealth
(Beaglehole & Bonita, 1997). Social development is therefore critical to improvements in
health and nutrition (Phillips & Verhasselt, 1994).
In recent years, international organizations such as the World Bank and the
United Nations have recognized the importance of a healthy population. These
organizations have incorporated health in economic policies and development strategies.
In 1993, the World Bank dedicated an entire report to the relationship between human
health, health policy and economic development. Good health was determined to be "a
fundamental goal of development as well as a means of accelerating it" (World Bank,
1993, p.21).
As the year 2000 approaches, it is clear that the goals for Health for All have not
been achieved. Discrepant social and economic development and various forms of
persistent discrimination prevented equitable access to health care and resources (Herrell

11
& Mulholland, 1998). This thesis examines one indicator of development, nutritional

status of children, to determine if there has been progress during the past decade. This
study illustrates the importance of socio-economic factors in relation to child health and
the need to incorporate broad base social changes in conjunction with economic
development in order to improve health conditions.

Social Statistics in Developing Countries
It is now recognized that economic statistics are insufficient for monitoring the

progress of development. Economic statistics should be complemented with social
statistics for a more complete picture. Moreover, indicators typically used in developed
countries are often not appropriate for poor countries (Cortinois, Vella & Ndiku, 1993).
Cortinois and his associates (1993) argue that it is more relevant to collect information of
the family, such as demographic features, literacy, living conditions ofparents, financial
standing and nutrition items.
There is a lack of current social statistics for many of the developing countries.
Certain statistics are more readily attainable such as the percent of children completing
primary schools, while others are more difficult to collect, such as the percent of children
malnourished (UNICEF, 1994). In particular, there is a lack of data over time that is able
to detect substantial change (Phillips & Verhasselt, 1994). The "data gap" in many ofthe
poorer countries is contributed to lack of resources and systems (Lerer et al. , 1998). A
large proportion of developing countries continue to use social statistics from the
eighties, many other statistics are manipulated through computerized techniques to
present more up to date values (UNICEF, 1994). Studies involving comparative analyses

12
are often problematic due to lack of consistent definitions and operational terms of the
objects of interest (Midgley, 1997).

1.2 Maternal-Child Health
Maternal-child health is one of the main components of primary health care.
Mothers and children have been recognized as part of a more vulnerable population in
many settings as they are the least powerful members of society (Price, 1993). Improving
the health of women and children are an important part of social and economic
development in communities. In the 1980's UNICEF launched a ' child survival and
development revolution' (CSDR). UNICEF aims to use a ' targeted PHC strategy' with
specific economically efficient health promotion strategies for children and mothers. It is
widely recognized that the promotion of the health of women is an important aspect of
improving the health of children (Price, 1994). The health ofmothers and children are
influenced by multiple factors including biological inheritance, individual circumstances
and the physical, socio-economic and cultural environment in which they live.
Women generally face numerous challenges in their lives, with the greatest
burden stemming from the status of women and the discrimination against them (Price,
1994). Women are often expected to both care and nurture her family and to also have a
productive role, which may involve various types of work, usually extremely labour
intensive. Women are more likely to be living in poverty, control fewer productive assets
and work longer hours at lower wages than men (Beaglehole & Bonita, 1997). These
factors all influence the health of the women, and thereby the health ofher children. It is
therefore, imperative to consider the role of women when examining the health and
nutritional status of children. It should be noted that although this thesis examines the

13
role of women in relation to the health of children, this is only one aspect of women in
development.

1.3 Malnutrition
What is malnutrition?
Malnutrition, is one of the most extensive and serious health problems in the
developing world. Malnutrition in developing countries is often labeled undernutrition, or
protein-energy malnutrition (PEM) because the child does not receive the necessary
protein or calories necessary for growth and development (Foster, 1992). Severe forms of
PEM can lead to nutritional disorders such as marasmus and kwashiorkor. Without
proper care and medical intervention children with these disorders may die within a short
period. Other forms of malnutrition involve micronutrient deficiencies such as lack of
iron, iodine and vitamin A. In this study malnutrition will refer specifically to PEM.

Epidemiology of Malnutrition
Approximately one-third of the children in the world are affected by PEM (WHO,
1998). Most of these children are from developing countries. The risk ofbeing an
undernourished child varies in different parts of the world. South Asia has the highest
levels of malnutrition, in both prevalence and absolute numbers. Children in Asia are 1.2
times more likely to be malnourished compared to African children, who in tum have a
risk 3 times higher than Latin American children. Future predictions expect that the
burden of malnutrition in terms of absolute numbers affected will shift from South Asia
to sub Saharan Africa (Thompson, 1996).

14

Mortality rates for children under five are 2.5 times higher in children who are
moderately underweight and 5 times higher in children who are severely underweight
(WHO, 1998). Over 50% of deaths in children under five in developing countries are
associated with malnutrition (Gove, 1997; WHO, 1998).

Relationship between Malnutrition, Health and Development
There is a strong synergistic relationship between nutrition and health; "good
nutrition leads to good health and good health leads to good nutrition" (Foster, 1992,
p.193). Good nutrition is important to the health development of individuals, families,
and communities (UNICEF, 1998). At the individual level, malnutrition effects physical
growth and development, health, and intellectual and social development. Poor physical
and cognitive development rooted in undernutrition exacerbates the cycle of poverty and
deprivation (Lerer, et al., 1998). These effects of childhood nutrition can extend into
adulthood. Poor health and nutrition in female children can lead to a vicious cycle of
malnutrition (Merchant & Kurz, 1993). Childhood PEM can lead to poor nutrition during
reproductive years. An undernourished mother is more likely to have a low birth weight
baby; low birth weight increases the chances of malnutrition during childhood. To break
this cycle, it is important for female children to start life with adequate nutrition.
At community and national levels, there can be losses in human potential leading
to great social and economic costs (Thompson, 1996). Dasgupta (1993) provides
evidence illustrating a strong link between nutritional status, physical work capacity,
endurance and physical productivity. "The nutritional well-being of people is a precondition for the development of societies. Governments will be unsuccessful in their

15
efforts to accelerate economic development in any significant long term sense until
optimal child growth and development are ensured for the majority." (WHO, 1994 as
cited in UNICEF, 1995) The importance of good nutrition is therefore not only of
importance to present health but also to future well being and productivity of the child,
and society.

The Determinants of Malnutrition
The determinants of malnutrition (direct or indirect causes) are multi-factorial and
complex. Numerous models and diagrams have been produced to attempt to explain the
relationship between the causes of undernutrition. This section provides an overview
using a conceptual framework and a description of the causes defined by UNICEF
(1998).
The causes of malnutrition can be classified as immediate causes, underlying
causes and basic causes (UNICEF, 1998). Good nutrition requires three important
components: food, health and care. These factors are broken down further in Figure 1.0 in
a general model of undernutrition. Determining the relationship among these factors is
important for finding solutions and developing programs to combat malnutrition.
Of particular concern in developing countries is the synergistic relationship
between malnutrition and infection, sometimes referred to as the malnutrition-infectioncomplex (Foster, 1992). Inadequate food intake coupled with illness are the two most
influential immediate causes of malnutrition (UNICEF, 1998). Children who are
malnourished have weakened immune systems increasing their susceptibility to infection;
an infection exacerbates malnutrition through various biological processes such as

16

Figure 1.0 Causes of child malnutrition, conceptual framework developed in 1990 as
part of the UNICEF Nutrition Strategy (UNICEF, 1998, p. 24)

Child malnutrition,
death and disability

Inadequate dietary
intake

c:::==~::::::::=-- Disease

Insufficient access
to food

._C
_____.r

=

Inadequate
inappropriate knowledge
discriminatory attitudes
household access to
resource

I....._
__

__.JI

Inadequate
maternal and childcare practices

/

Poor water/
sanitation and
inadequate health
services

Quantity and quality of actual
resources-human, economic
and organizational-and the
way they are controlled

Potential resources: environment, technology, people

Political, cultural,
economic and social
including women's status,
the utilization of
resource

17
reducing the body's absorption ofnutrients (Foster, 1992). Most childhood deaths are the
result of malnutrition combined with repeated infections.
Underlying causes contributing to malnutrition and disease include three main
categories: household food security, access to health services and environmental health,
and caring practices (UNICEF, 1998). Household food security is defined as sustainable
access to resources to acquire the necessary food requirements for adequate nutrition. The
accessibility to food includes financial, physical and social elements. Health services
(preventative and curative) are extremely important for maintaining good health. It is
essential to have health services that are both affordable and accessible. Environmental
health determinants such as access to safe water, good sanitation facilities and safe food
handling practices are important for good nutrition. These factors are extremely important
in the malnutrition-infection-complex. Caring practices influencing the nutritional status
of children include feeding practices, protection of the child's health, support and
cognitive stimulation for children and care and support for the mother.
Poverty continues to be the most pressing issue in public health; poverty is the
most basic cause of malnutrition and ill health (Lerer et al., 1998; WHO, 1998;
Beaglehole & Bonita, 1997; Phillips & Verhasselt, 1994). Poor people lack the income
for basic requirements such as food, clean water, and shelter needed for a minimum level
of health and nutrition. In the world as a whole, people with low incomes tend to
underconsume food, while people with a high income tend to overconsume food (Foster,
1992). Without progress towards reducing poverty there will continue to be inequality
with respect to health and nutritional status. There are other important socio-economic

18
causes of malnutrition, these are ofkey interest in this study and are discussed in the
following chapter.

Measuring Nutritional Status
Anthropometry, the science of measuring the human body, is the most popular
epidemiological tool for assessing the health and nutritional status of children (Foster,
1992). Anthropometric indicators are not as accurate as other biological and biochemical
methods, and cannot be considered a precise measurement of nutritional status,
particularly at the individual level (Gorstein & Akre, 1988). This method, however, is
relatively accurate at determining protein and calorie nutrition at the population level and
can be achieved at a low cost. Moreover, anthropometric data is considered one of the
most revealing statistics with respect to health (Dasgupta, 1993).
Three common anthropometric measurements used for this study include heightfor-age (HA), weight-for-height (WH), and weight-for-age (WA) (Waterlow, 1992).
Height-for-age measures stunting, which is an indication of chronic malnutrition. This
measure is particularly useful for evaluating the effects of social and economic changes.
Weight-for-height is a measure ofwasting, which indicates present nutritional deficiency.
Weight-for-age measures if a child is underweight, indicating both past and present
nutritional status. Undernutrition refers to a child who is classified as stunted, wasted,
and/or underweight (Sommerfelt & Stewart, 1994).
Anthropometric indicators require a reference population to compare an
individual's growth status (Dibley et al., 1987). The reference population that is most
commonly used comes from the World Health Organization/National Center for Health

19
Statistics/Centers for Disease Control (McMurray, 1996). The data were collected from
two samples ofhealthy American children during the 1970's. The use ofthis American
reference population in developing countries has been criticized for the assumption that
children of different origins share the same potential for growth. Studies have
demonstrated, however, that ethnic variations in growth are either small or non-existent
during early childhood; instead poor growth stems from malnutrition and infectious
diseases (Habitch et al., 1974). Ethnic differences in growth attainment are established
during adolescence and not during early childhood (Falkner, 1986).
The most commonly used index for anthropometric indicators are Z-scores, which
are standard deviations from the reference median (Dibley et al., 1987). Standardization
has made it possible to evaluate nutritional changes and trends over time in nationally
representative cross-sectional studies that have been conducted at various time frames
(Gerstein & Akre, 1988). Cross-sectional data can be used to study differences between
subgroups such as sex and can illustrate associations with socio-economic, environmental
and demographic factors (McMurray, 1996).

1.4 Summary
Malnutrition is a serious problem in the developing world. Good nutritional status
is important for normal growth and development. There are numerous biological, social,
economic and environmental determinants of child growth. Targeting certain key
pediatric diseases, therefore, is most likely not the answer to reducing malnutrition.
Although PHC was initiated to take a multi-sectoral approach to health development, it
has been pushed aside by more narrow medical technologies and cost-effective strategies.
This study involves the examination of some of the socio-economic factors influencing

20
the nutritional status of children. It is believed that broad social changes coupled with
economic development will have the greatest impact on the promotion of health. Such
action will require support and dedication from national governments and international
organizations. Anthropometric indicators are used to both examine global changes in the
nutritional status of children and socio-economic influences of malnutrition.

21

Chapter Two: Literature Review
2.0 Introduction
The literature review is organized into four sections. The first section discusses the
statistical association between the three anthropometric indicators; height-for-age, weightfor-height and weight-for-age. The second section outlines the global trends of the
nutritional indicators that have been observed. The third section discusses the socioeconomic determinants of growth. The determinants are reviewed with respect to the
choice of variables chosen for this study. The final section summarizes the key fmdings in
the literature.

2.1 Statistical Association between Nutritional Indicators
The three most commonly used anthropometric indicators, described in the previous
chapter, include height-for-age (stunting), weight-for-height (wasting) and weight-for-age
(underweight). Waterlow (1992), a prominent researcher in this field, examined the
statistical associations between indicators in a number of studies. He determined that most
populations have a weight-for-age fairly well correlated with height-for-age. Waterlow did
not comment on the association between weight-for-age and weight-for-height. Keller
(1983, as cited in Waterlow, 1992) found no correlation between weight-for-height and
height-for-age. These two indicators, however, explain more than 95% of the variance in
weight-for-age. This field remains unsettled.

2.2 Global Trends of Nutritional Indicators
Victora (1992) noted that there was a low prevalence ofwasting in Latin American
children. This is surprising because these children were living in impoverished areas, with
high levels of mortality and morbidity, and high levels of stunting. Using data from the

22
WHO, Victora tested the distribution of weight-for-height, height-for-age and weight-forage in different continents and the associations between the indicators. Cross-sectional
data were collected from 175 studies from the 1960's, 1970's and 1980's. The studies are
broken down by region in Table 2.0. Using individual studies as the units of analysis,
Victora calculated median prevalences of the indicators for each region and used
Pearson's correlation coefficient and linear regression to analyze the association between
indicators. The analysis was restricted to children between the ages of 12 to 23 months
because this is the age range when wasting is most prevalent.
The median prevalences of malnutrition are presented in Table 2.0. Stunting had the
highest prevalence (except in Asia) and wasting had the lowest prevalence of the three
indicators. Asia had the highest prevalence of all three indicators and Latin America had
the lowest prevalences. Stunting demonstrated little variability in prevalence among
regions, while wasting varied sevenfold. The correlation results are presented in Table 2.1.
There was a strong correlation between wasting and stunting prevalences for Asia and the
Mediterranean, a weak correlation for Latin America and no association for Africa.
The results of this study do not explain the association between stunting and wasting.
It is interesting that correlations between wasting and stunting vary among regions.

Victora (1992) interprets the varying patterns of global nutrition as an indication of some
differences in causes of undernutrition. He recommends developing a better understanding
of these factors and the reasons why the associations of the indicators are different around
the world.

23
Table 2.0 Median Prevalences of Malnutrition and Estimated Under-five Mortality
Rates (USMR), by Region
Prevalence of malnutrition 1
Wasting
Stunting
Underweight U5MR
No. of
studies

Africa
Latin America
Asia
Eastern Mediterranean

10.1
2.7
18.8
5.8

41.1
33.8
47.0
38.6

35.7
22.8
52.9
25.4

146
85
72
99

Source: Victora (1992)

89
37
32
17

1

Percentage of children below the National Center for Health Statistics median minus two standard
deviations.

Table 2.1 Correlation of the prevalences of wasting on stunting by region

Region
Africa
Latin America
Asia
Eastern Mediterranean
1

2

0.044
0.311
0.700
0.652

p
0.7
0.06
<0.001
<0.001

Source: Victora (1992)

r = correlation coefficient
p = probability value

Trends in child malnutrition were examined by Sommerfelt & Stewart (1994). The
researchers published a Demographic and Health Surveys technical report comparing the
nutritional status of children under 5. There were 19 countries included in the study; 7
from sub-Saharan Africa, 3 from Near East/North Africa, 2 from Asia and 6 from Latin
America/Caribbean. The countries were all DHS-I studies ranging from 1985 to 1990. The
percentage of children stunted, wasted, and underweight were calculated using Z-scores
below minus 2 standard deviations according to the WHO/CDC/NCHS International
Reference Population.

24
Three main patterns emerged: (1) levels of stunting were much higher than levels
of underweight combined with a small proportion of children wasted (approximating
levels of the International Reference Population); (2) high levels of stunting with even
higher levels of children underweight, wasting had a higher prevalence than other
countries; (3) relatively equal levels of stunting and underweight, and wasting was more
common than expected. The first pattern was typical in Latin American, North African
countries and Uganda and Zimbabwe, the second pattern was seen in countries such as
Mali and Sri Lanka and the third pattern included most sub-Saharan African countries.
The patterns are explicitly stated and confirm the findings ofVictora (1992). There is,
however, no explanation given for possible reasons for the different nutritional patterns.
The worldwide distribution ofPEM was described using the WHO Global
Database on Child Growth (de Onis, Monteiro, Akre & Clugston, 1993). The database
consists of anthropometric data on children under 5 using the WHO/CDC/NCHS
international reference population. Cross-sectional data were used from nationally
representative surveys carried out in developing countries in Africa, Asia, Latin America
and Oceania between 1980 and 1992. The study determined differences in nutritional
status between regions and also differences within each region. Regional prevalences for
height-for-age, weight-for-height, and weight-for-age were calculated by weighting
national prevalences according to the population of children under five in each country in
1990. Global prevalences were then obtained by calculating the mean number of
undernourished children in each area.
The results of the global estimates for the prevalence of the anthropometric
indicators are summarized in Table 2.2. Stunting had the highest prevalence for all three

25
regions, while wasting had the lowest prevalence. The risk of being underweight was 1.5
times higher in Asia than in Africa, and 2.3 times higher in Africa than in Latin America.
There were similar differences for prevalences in stunting and wasting, except levels of
stunting in Africa were similar to those in Latin America and levels of wasting in Africa
were similar to levels in Asia.

Table 2.2 Global Estimates for the Prevalence3 and Number of Underweight,
Stunted, and Wasted Children in Developing Countries

Africa
Asia
Latin America
Oceania
All developing
Countries

% underweight

%stunted

%wasted

27.4(31.6)b
42.0 (154.1)
11.9(6.5)
29.1(0.3)

38.6 (44.6)
47.1 (172.8)
22.2 (172.8)
41.9 (0.4)

7.2 (8.3)
10.8 (39.6)
2.7 (1.5)
5.6 (0.1)

42.7 (229.9)

9.2_(49.5)

35.8 (192.5)

Source: deOnis, Monterro, Akre and Clugston (1993)

•% Below - 2 standard deviations of WHO/NCHS reference value
b Figures in parentheses are millions of children

This study found that there were regional differences in the prevalences of the
nutritional indicators (see Table 2.3). The highest levels of stunting, wasting and
underweight were in southern Asia. The lowest prevalences of these indicators were in
northern Africa and in all three Latin America regions.

26

Table 2.3 Regional Estimates for the Prevalencea of underweight, stunted and wasted
children in developing countries, ranked in descending order
% underweight
Southern Asia
South-eastern Asia
Western Africa
Eastern Africa
Melanesia
Eastern Asia
Caribbean
Central America
Northern Africa
Southern America

60.5
37.8
32.8
31.0
29.5
21.3
19.4
17.7
11.3
8.4

%stunted
Southern Asia
Eastern Africa
South-eastern Asia
Melanesia
Western Africa
Eastern Asia
Central America
Caribbean
Northern Africa
Southern America

60.3
47.0
43.2
42.2
37.9
32.1
29.8
25.9
25.4
18.1

%wasted
Southern Asia
Western Africa
South-eastern Asia
Eastern Africa
Northern Africa
Melanesia
Central America
Eastern Asia
Caribbean
Southern America

17.3
9.5
7.6
6.0
5.8
5.5
4.6
3.6
2.2
1.9

Source: deOnis, Monteiro, Akre and Clugston . (1993)
a % Below - 2 standard deviations of the WHO/NCHS reference value

The purpose of the study was to determine the worldwide magnitude of PEM. The
study also describes varying patterns of undernutrition, but the researchers do not
comment on why there are regional and global differences in anthropometric indicators.
They treat all three nutritional indicators as a general category of undernutrition and
conclude that growth retardation is rooted in poverty and lack of education.
Common trends of nutritional indicators have been described in the
aforementioned studies. The patterns that have emerged indicate that there are regional
and continental differences in nutritional status. A better understanding of the socioeconomic determinants of these anthropometric indicators may help to explain, in part, the
reasons for global differences in nutritional status.
2.3 Determinants of Growth
This section reviews the literature of the determinants of growth in children. The
determinants are first reviewed individually with respect to their relationship to
malnutrition. This is followed by a discussion of six studies that have attempted to

27
understand the relationships among numerous determinants of growth. The purpose of this
review is to describe methodologies and results of studies that have examined socioeconomic determinants and to provide a background for this study.

Specific examination of determinants ofgrowth
It is generally accepted that rural residents have poorer health than urban residents.

People living in rural areas tend to have less access to health care services and to cash with
which to buy food and other services (Phillips & Verhasselt, 1994). The flip side to this
view is that the provision of services is not necessarily better in cities. The urban poor may
be just as disadvantaged as their rural counterparts. There is evidence that nutrition, family
support and environmental factors such as clean water are better in rural areas for some
countries. There have been explosions of population in some cities that have lead to
squatter settlements or 'septic fringes' which have extremely poor sanitation conditions
leading to widespread of infectious diseases and other health problems (Basch, 1990).
Poor nutritional status is a reflection of poverty and impoverished conditions which can
exist both in rural and urban areas.
Distinguishing place of residence is important when attempting to determine progress
and changes of nutritional status that have occurred over time. Child growth was
examined in China before and after economic reforms to compare the effect of the reforms
on the growth of children in urban and rural areas (Shen, Habicht & Chang, 1996). The
authors used five large-scale cross-sectional surveys between 1975 and 1992 that
examined the growth of children in China between the ages of2 to 5. Using height-for-age
as the measurement of growth, analysis of variance was utilized to examine differences

28
between urban and rural growth and secular trends. In 1975, before economic reforms, the
results indicated that children in urban areas were only slightly shorter compared to
children in rural areas. Between 1987 and 1992 there were improvements in child growth,
but the changes were not equal. Children in urban areas had a greater increase in growth
compared to rural children. The authors conclude that although there have been great
improvements in the nutritional status of children since the reforms, the changes are not
the same in all areas of the country. Mean differences should be carefully considered and
are not necessarily indicative of overall improvement.
Gender differentials in child survival is an important and widely debated topic. Girls
have a biological advantage in child survival, but in many countries girls have higher
mortality rates than boys (Merchant & Kurz, 1993). There is evidence of a parental
preference for boys in many developing countries leading to inequalities in intrahousehold allocation of resources. Chen (1980) noted in Bangladesh that girl children
were fed less and suffered from greater malnutrition (Dasgupta, 1993). There were similar
infection rates but parents were more likely to buy medication for boys and there was a
tendency for boys to have quicker recoveries.
Particular regions of the world have demonstrated gender disparities such as parts of
Asia and North Africa (Hill & Upchurch, 1995). Latin America countries do not tend to
display gender differences with respect to undernutrition. Svedberg (1990, 1996) and
Klasen (1996) have debated whether there is a gender bias in sub Saharan Africa. Both
authors analyzed anthropometric indicators using various national data from sources such
as UNICEF and DHS. Svedberg (1990, 1996) determined that there is either no gender
bias or an anti-male bias in nutritional status. He concluded that his results do not imply an

29
overall female advantage in the quality of life but that undernutrition is more common in
boys than girls in sub Saharan Africa. Klassen (1996) re-analysed data from Svedberg's
(1990) study and provided additional evidence. Klassen found that there was an antifemale bias. Klassen concluded that more research is needed to confirm his conclusions
and suggested using time series for health and mortality indicators to conduct further
examinations of gender differences.
Hill and Upchurch (1995) explored gender bias in child health using data from the
DHS. Thirty-five countries were examined from the Americas, sub Saharan Africa,
Middle Eastern Crescent and other Asian countries. Results indicated that more boys were
stunted than girls. Girls were also less likely to be wasted, although these differences were
considerably smaller. The authors confirmed that there was a gender bias with respect to
mortality but concluded that nutritional differences were not the key to differences in
mortality rates.
There appears to be an association between maternal education and the health status of
children. Cleland and Ginneken (1988) determined that for every year of maternal
education there was an average of 7 to 9% decline in child mortality (Bicego & Boerma,
1993). Caldwell (1993) examined 11 countries in Asia, Africa and Latin America. The
countries had low per capita incomes but relatively good health achievements. He found
that education of females at childbearing age had the highest correlation with health
success. Increasing female education is a central factor in improving maternal and child
health (Beaglehole & Bonita, 1997).
In another study, Bicego & Boerma (1993) examined the statistical association of
maternal education and child health and survival. Using DHS data, 17 countries from

30
Asia, Africa, and Latin America were examined using multivariate logistic and hazard
regression with stunting and underweight status as the outcome variables. There was a
strong relationship between maternal education and stunting even after controlling for
socioeconomic variables. An even stronger relationship was found between maternal
education and underweight status. This relationship, however, was significantly decreased
once economic variables were included in the equation. This study illustrates that the
impact of maternal education is different according to nutritional indicator.
Victora and his associates (1992) conducted a prospective study to examine the role of
maternal education on various infant and child outcomes. The purpose of the study was to
determine the association between education of the mother and child health indicators
controlling for other socio-economic factors. The researchers followed a cohort of 6011
children born in a city in Southern Brazil. Information was collected by interviews with
the mother at the hospital, city-wide mortality surveillance system and a follow-up
examination of the children when they were 12-27 months. Data analysis included crosstabulations and chi-squared tests using length-for-age, weight-for-age and weight-forlength as continuous outcome variables. Multiple logistic regression was used to adjust for
several confounding variables (age, racial group, family income, education of
father/partner). After controlling for other socio-economic variables, the effect of maternal
schooling remained significant on the nutritional status of their children.
As more women join the labour force, the impact of a working mother on the health
and nutritional status of children has come into question. The two major effects of
maternal work status that have been hypothesized include the positive income effect and
the negative time effect (Leslie, 1988). More simply stated, women bringing in more

31
income increase household resources such as food, but will spend less time in the
household caring and nurturing children.
Leslie (1988) reviewed 50 studies that examined women's work and child nutritional
status. The findings of these studies were complex and contradictory. Some of the studies
found that there was a positive association between maternal work and child nutritional
status. The children had better nutrient intake and better physical growth, but these effects
were small and not consistent. Other studies found that there was a negative association,
but did not find a causal relationship. The author concluded that there is little evidence of
a negative relationship between maternal work status and nutritional status of children and
suggests that there should be country specific research to account for differing patterns in
female labour force participation.
The relationship among female employment, childcare strategies, and nutritional status
were examined in a study involving Nicaraguan children (Lamontagen, Engle & Zeitlin,
1997). Households were sampled by randomized block design in low income urban
communities. Data collection involved an interview, observations of feeding and childcare
practices and anthropometric measurements of the children. Using height-for-age, weightfor-age and weight-for-height as the outcome measurements, multiple regressions were
performed to determine the independent effects of women's employment and adequacy of
childcare on child nutritional status controlling for possible confounding variables. The
study found that children of working mothers had significantly higher weight-for-heights
than children of mothers who were not employed. This relationship did not change once
socioeconomic status, maternal education, paternal financial support, childcare adequacy
and the sex and age of the child were controlled. The variables that distinguished between

32
mothers who provided adequate care and mothers that did not include; the provision of a
larger variety of food, the increased likelihood of completing measles vaccinations and
better hygienic practices.
There are direct links between the health of a child and the adequacy of sanitation
facilities and water supplies. The 1980's, the International Drinking Water Supply and
Sanitation Decade, led to a greater proportion of the world achieving access to safe water
and sanitation facilities (Beaglehole & Bonita, 1997). The world population, however, also
increased during this time leading to a large number of people without safe water and
adequate sanitation. Currently there are more people with access to safe water than to
adequate sanitation facilities (Esrey, 1996). Safe water has been the target of many
development programs and there have been great improvements during the past few years.
Sanitation coverage in populations, however, is slipping.
Esrey (1996) examined the impact of incremental improvements ofwater and
sanitation on health status. DHS data was analyzed for 8 developing countries which
included Burundi, Ghana, Togo, Uganda, Sri Lanka, Morocco, Bolivia, and Guatemala.
Controlling for household, maternal and child-level variables, a series of multiple
regressions were performed with anthropometric measurements and diarrhea prevalence as
outcome variables. Esrey found that children had less bouts of diarrhea, and had a better
nutritional status with improved sanitation at each of three levels of water supply. The
health benefits from improved sanitation were greater than improvements in water
conditions. The benefits from improved water supply only occurred ifthere were
improvements in sanitation and when there were optimal water conditions. These results

33
indicate that there should be greater emphasis on improving community sanitation
facilities in conjunction with improved water supplies.
Diarrhea is the most frequently and consistently reported disease influencing
nutritional status (Black, 1991 ; Ulijasek, 1990) Childhood infectious disease reported to
influence growth include diarrhea, measles, respiratory diseases, tuberculosis, malaria and
intestinal parasitic. Black (1991) estimated 10-80% of growth retardations can be
attributed to diarrheal disease. Malnutrition can make a child more susceptible to diarrhea,
episodes of diarrhea can increase nutritional deficiency (Huffman & Martin, 1994). A
study in North India examined the impact of nutrition on diarrheal disease. The case
fatality rate was 24 times greater for severely undernourished children compared to
children with adequate nutrition. The impact that diarrheal disease has on the growth of
children has been questioned by some researchers. A study conducted in Bagladesh found
that the effects of diarrhea are only temporary and that controlling diarrhea may not
improve nutritional status (Briend et al., 1988). Lutter and his associates (1989) found that
diarrhea did not have a negative impact on growth in children who were adequately
nourished. Diarrhea, however, did impact negatively on growth when children were
malnourished.
Ulijasek (1990) examined 10 studies, which examined the effects ofPEM on
immunization outcome. Tetanus and diphtheria were not found to effect nutritional status,
while BCG, measles, typhoid and polio did show an effect. Measles is considered to be an
important infectious disease with respect to malnutrition because there is a high risk of
death upon infection (Sommerfelt & Stewart, 1994; Ulijasek, 1990). These diseases are
best prevented by vaccinations.

34

General examination of determinants of growth
The study described previously by Sommerfelt and Stewart (1994) examined
differentials in levels of undernutrition according to bio-demographic characteristics,
socioeconomic characteristics, recent illness, and vaccination history. The authors
distinguished the determinants for each indicator, but discussed the results as a general
category of undernutrition. Results indicated that there were small differences in
nutritional status with respect to sex of the child, with the exception of the Dominican
Republic where boys were much more likely to be malnourished compared to girls.
Children were more likely to be undernourished if they lived in rural areas compared to
urban areas. There was a greater prevalence of undernutrition in mothers with no
education and the lowest prevalence among mothers with secondary or higher education.
The children of mothers who did not work were more likely to be undernourished. There
was a strong correlation between undernutrition and sanitation facilities, malnutrition is
more common in children where there is a lack of toilet facilities. Source of drinking
water was an important determinant, malnutrition was more prominent for children who
drank surface water. The study also found children who did not receive a vaccination for
measles, as well as children who had diarreal illness were more likely to be malnourished.
This study includes all the determinants of undernutrition previously discussed but the
relationship between the variables were not addressed.
The health and nutritional status of Guatemalan children were investigated in
relation to the effects of socio-economic determinants (Pebley & Goldman, 1995). The
study used data from the National Survey of Maternal and Child Health (ENSMI)

35
conducted in Guatemala in 1987. Using height-for-age as the outcome variable regression
techniques were performed to assess the effects of various covariates. The findings
concluded that poverty and poor living conditions lead to deficits in the height of children.
The growth of children was also influenced by ethnicity, father's occupation, land
distribution in the area where they live, and maternal education.
Malnutrition in children under 5 was examined in Zimbabwe for the effect of
socioeconomic and health factors (Madzingira, 1995). Using data from the 1988
Zimbabwe Demographic and Health Survey, separate forward stepwise logistic
regressions were performed using height-for-age, weight-for-age and weight-for-height as
the dependent variables. The independent variables included various socioeconomic,
environmental and health-related variables. The author determined that the most
statistically significant independent variables were the greatest predictors of nutritional
status. These variables included birth status of the child, residence, diarrhoeal status, birth
weight, maternal education and duration ofbreastfeeding. Separate analyses were
performed and results are reported for each anthropometric indicator. The findings,
however, are discussed in a general category of nutritional status. Such discussion
prevents a closer inspection of the reasons why the variables were different for each
nutritional indicator.
Vella and his associates (1994) examined the determinants of stunting in Northwest
Uganda. Length and height of 1072 children under 5 were measured in 30 villages using
random sampling. After 2 years the same children were measured again. Questionnaires
were administered to collect information on socioeconomic and health-related variables.
The data were analyzed using logistic regression with height-for-age as the outcome

36
variable. Of the variables measured there were three variables that were significant for
predicting stunting; father's lack of income besides agriculture, mothers lack of education
and the presence of stunting at baseline.
The prevalence and socio-economic determinants of malnutrition were
investigated in Sikasso, Mali (Bouvier et al., 1995). Surveys were implemented every year
for 5 years in a sample of 491 families . Logistic regressions were performed using wasting
and stunting as outcome variables. No strong associations were found between wasting
and socio-economic determinants. Stunting showed strong associations for education
levels of mother and father and with the family assets. This study is important for it shows
that in this sample the causes of malnutrition were different for stunting and wasting.
The associations between stunting and wasting, in children under five, were
examined globally to attempt to explain the variability among nations (Frongillo, de Onis,
& Hanson, 1997). The data came from the WHO Global Database on Child Growth. The

authors used nationally representative studies that were done after 1980. Seventy
countries were included from Africa, Asia and Latin America. Using a comprehensive
conceptual model, underlying causes and basic causes of malnutrition were analyzed.
Specific variables were selected for multiple linear regression models through correlation
and factor analyses. Results found that once social, economic, and demographic factors
were accounted for, differences between global patterns of wasting and stunting were
eliminated. Most of the variability among nations was explained by national factors. More
of the variability of stunting than wasting could be explained. Energy availability was the
only factor associated both with stunting and wasting. Stunting was associated with female
literacy, health expenditures and gross national product, while wasting was associated

37
with immunization. These results indicate that there are differences in causation in
stunting and wasting. These results have implications for intervention strategies. Solutions
should target those determinants that are causing the greatest burden of malnutrition.

2.4 Summary
The nutritional status of children varies in different parts of the world and
according to the choice of nutritional indicator. There are also differing statistical
associations of anthropometric indicators among regions . Victora (1992) noted that the
varying prevalences of anthropometric indicators and associations between these
indicators may be contributed to different causes of undernutrition. The work of Frongillo
and his associates ( 1997) further suggest that there are different causes for stunting and
wasting. There is a need of a greater understanding of the social and economic
determinants of malnutrition with respect to each indicator.
This section has linked various socio-economic factors to child growth. Some
factors such as maternal education and adequate sanitation have strong and positive
impacts on child nutrition, while the relationships between malnutrition and factors such
as maternal work status and sex of the child are not as clear. Most studies examine one
anthropometric indicator at a time, or discuss the determinants of growth as a general
category of undernutrition. The studies that have examined determinants according to
specific indicators have found that relationships are different according to the choice of
indicator (Bouvier et al., 1995; Bicego & Boerma, 1993). The relationship between
varying nutritional patterns and the socio-economic determinants of growth of each
indicator is explored. The present study adds to this body of research by examining these
relationships over time.

38
The studies reviewed have described trends in nutritional status of children over
time and others have described differentials of malnutrition. This study builds on this long
tradition of research by relating some of the key determinants of malnutrition to patterns
of nutritional status. Furthermore, most ofthe studies examining the determinants of
growth have focused on one country. This study compares nutritional and socio-economic
factors across countries. An attempt is made to determine common determinants for all
five countries.

39
Chapter Three: Methods
Monitoring the nutritional status of children and women is a fundamental instrument
for the goal of ending malnutrition.

~

Macpherson (1993)

3.0 Source of data
The data for this thesis came from Demographic and Health Surveys (DHS,
1998). Funded by the United States Agency for International Development (USAID),
DHS assists developing countries in the implementation of nationally representative
surveys. Based on multistage random sampling, these surveys contain demographic and
health data as well as background information. Originating in 1984, DHS has
implemented three phases of surveys, DHS-1 (1984-1989), DHS-11 (1988-1993) and
DHS-111 (1992-1997). DHS surveys are a powerful source for comparing data between
countries (Caulfield, Bentley & Ahmed, 1996; Esrey, 1995; Fosu, 1994; Bicego &
Boerma, 1993; Forsberg, Van Ginneken & Nagelkerke, 1993).
While there is a shortage of surveys measuring social indicators of development
in most developing countries, DHS has been acclaimed as one of the best sources of data
and is used by many researchers in various fields. There is a wide body of literature that
discusses the quality ofDHS data (Le & Verma, 1997; Sommerfelt & Stewart, 1994;
Sommerfelt & Stewart, 1993; Boerma & Sommerfelt, 1993). A few key points will be
summarized that are relevant to this particular study.
DHS surveys have not been conducted at regular intervals. At this time, it is not
possible to examine each country after a ten year span, with the exception of Brazil. Each
country, however, has had the first survey conducted in the late 1980's and the second

'

40
survey in the mid 1990's. This allows for a comparison before and after the World
Summit for Children. Furthermore, as the year 2000 approaches, this study will describe
changes and progress that occurred throughout the 1990's.
DHS surveys are cross-sectional, therefore it is not possible to evaluate nutritional
and health status of children at an individual level. Cross-sectional data are useful for
depicting national patterns and can show strong associations with certain socio-economic,
environmental and demographic factors (McMurray, 1996).
Like other large sample surveys, information in DHS surveys are highly subjected
to reporting and recall biases (Boerma & Sommerfelt, 1993). Most ofthe health
information such as diarrhea prevalence and age of the child is obtained from maternal
reporting which inevitably leads to recall bias. Weight and height measurements are
measured by the researchers, so the measurements are not influenced by reporting bias
but there will contain some measurement errors. Misclassification biases have also been
noted, and are extremely difficult to correct. Boerma & Sommerfelt (1993) suggested that
these biases are not too problematic if they are fairly random, but caution should be
extended when interpreting individual level data.

3.1 Sampling Procedures
Purposive sampling was utilized to select countries for analysis. Countries were
selected on the following criteria: the country must have been surveyed during the
mid- to-late 1980's and then surveyed five to ten years later, and the survey must contain
a child anthropometry segment. Only five countries met the stipulated criteria: Brazil,
Guatemala, Uganda, Mali and Zimbabwe. A further description of the surveys can be
seen in Appendix A.

41
3.2 Country Profiles
Brazil is the largest country in South America (total area= 8,511,965 square
kilometers) (Gale, 1998). The population (1996) in Brazil is approximately 160 million
with an annual growth rate of 1.7%. The adult literacy rate is 81%. Of the five countries
in this study, Brazil ranks the highest on the human development index (HDI) (see Table
3.1). The Gross Domestic Product (GDP) per capita (1994) in US dollars is 5,362 (United
Nations, 1997). Urban growth has been extremely fast and as of 1991, 75% of the
population were living in urban areas (Gale, 1998). Brazil has a highly diversified
economy and large variations in levels of development. Appearing to lead many other
developing countries in economic development, Brazil also has one of the most
inequitable distributions of incomes in the world. The Northeast region is the poorest part
of Brazil, and displays lower social development than other regions.
Guatemala, located in Central America (total area= 108,890 square kilometers)
has a population (1996) estimated to be 10.5 million with an annual population growth
rate of2.9% (Gale, 1998) Guatemala, similarly to Brazil, has a medium ranking on the
HDI, but falls below Brazil by .21. The literacy rate is considerably lower than Brazil's at
52%. The GDP per capita in US dollars (1994) is 3,208 (UNDP, 1997). The economy has
been growing steadily in the 1990's, but the developmental process has been problematic
due to natural hazards such as volcanoes, tropical storms and frequent earthquakes.
Guatemala also has poor distribution of income and wealth; the wealthiest 10% of the
population receive close to 50% of all income. More than half the population lives in
poverty of which two thirds live in extreme poverty.

42
Mali, located in Western Africa (total area= 1.24 million square kilometers), has
a population (1997) of9.8 million with an annual growth rate of2.9% (Gale, 1998). Mali
ranks extremely low on the HDI, among the lowest five countries in the world. The
literacy rate is 15%. Mali is one of the ten poorest countries in the world with a per capita
income in US dollars (1994) of543 (UNDP, 1997). With limited land available to
agriculture, most economic activity, fishing and farming, is concentrated in one area
(Gale, 1998). Furthermore, Mali is burdened by recurring droughts. There was economic
progress in the early 1980's after a series of adjustment and stabilization programs but
progress slowed considerably in 1987. Mali is extremely dependent on foreign aid and is
vulnerable to fluctuations in world prices.
Uganda, located in Eastern Africa (total area= 236,040 square kilometers), has a
population (1995) of approximately 19 million people with an annual growth rate of2.9%
(Gale, 1998). Uganda has a low rating on the HDI, slightly higher than Mali. The literacy
rate is considerably higher than Mali's at 62%. Blessed with significant natural resources
the Ugandan economy has great potential, however, political instability and inconsistent
economic management have made Uganda one of the poorest countries in the world with
a GDP per capita in US dollars (1994) of 1370 (UNDP, 1997). After periods of economic
setbacks throughout the 1980's there has been economic progress including significant
decreases in inflation after 1987 (Gale, 1998).
Zimbabwe is situated in Southern Africa (total area= 390,759 square kilometers)
with a population (1991) of 11.5 million with an annual growth rate of3.1 %. Zimbabwe
rates low on the HDI, but has the highest score of the three African countries. The
literacy rate for Zimbabwe is the highest among the selected countries at 85%. Zimbabwe

43
is a lower-middle income country with a GDP per capita in US dollars (1994) of2,196
(UNDP, 1997). Throughout the 1980's, Zimbabwe experienced periods of economic
growth along with foreign exchange crisis and droughts. Economic reform began in 1991
but there have been persistent problems with budget deficits.

Table 3.0 Human Development Index ratings for all five countries (HDI) (Human
Development Report, 1997)

Country

HDI

HDI Ranking

HDI Classification

(of 175 countries)
Brazil

.783

68

Medium

Guatemala

.572

117

Medium

Mali

.229

171

Low

Uganda

.328

159

Low

Zimbabwe

.513

129

Low

The human development index is a measure of development composed of three indicators: life
expectancy, literacy and income. The index is measured on a scale from zero-to-one with higher values
signifying higher levels of human development.

44
Latin America has seen tremendous progress in child health in the past fifty years.
The under 5 mortality rate (U5MR) dropped from 15% in 1960 to 5% in 1993, the lowest
rate in the developing world (UNICEF, 1998). Brazil and Guatemala have similar U5MR
and infant mortality rates (IMR), as seen in Table 3.2. Guatemala, however, has a much
higher percent of children underweight (identical to the percent of children underweight
in Mali), and the social indicators of Guatemala are among the worst in Latin America
(Gale, 1998).
Sub-Saharan Africa has displayed the slowest progress for child health in the
world (UNICEF, 1998). Mali demonstrates the poorest child health of the selected
countries, ranking seventh for the U5MR, a rate approximately four times higher than
Brazil and Guatemala. Uganda has a low U5MR; there is no available data for the percent
of children underweight, but Uganda displays the lowest percent of low birthweight
babies born between 1990 and 1994. Zimbabwe falls in the middle of the Latin American
countries and Uganda for all selected health indicators.

45

Table 3.1 Selected child health indicators

Under-5

Under-5

Infant

Underweight

% ofinfants

Mortality

Mortality

Mortality

children under

with low birth

Rate

Rank

Rate

age 5 (%)

weight

1996

1996

1996

1990-1996

1990-1994

Brazil

52

79

44

6

11

Guatemala

56

74

43

27

15

Mali

220

7

134

27

17

Uganda

141

30

88

Zimbabwe

73

65

49

16

14

World

88

175

60

30

17

Country

26

Source: UNICEF (1998)

Definitions of the indicators:
Under-5 mortality rate-Probability of dying between birth and exactly five years of age
expressed per 1,000 live births
Under-5 mortality rank-Rank of country in descending order of their estimated 1996
under-5 mortality rate out of 175 countries.
Infant mortality rate-Probability of dying between birth and exactly one year of age
expressed per 1,000 live births.
Underweight-Moderate below two standard deviations from median weight for age of
reference population
Low birthweight-Less than 2500 grams.
- data not available

46
3.3 Description of Sample

The surveys were cross-sectional household surveys administered to women of
childbearing age, regardless ofmarital status. The respondents were women 15-44 years
of age in Brazil and Guatemala and 15-49 in Mali, Uganda and Zimbabwe.
Anthropometric measurements of children were assessed in a nationally representative
sample with the exception ofBrazil1986, which recorded anthropometry in the Northeast
region. Only children from the Northeast region were, therefore, selected from the Brazil
1996 survey to maintain consistency. Children were measured at specific ages varying
among surveys between 0 to 66 months. To increase comparability between surveys
children were selected iftheir age fell between 3-35 months. The data were further
reduced to remove subjects with missing data and flagged cases. Flagged cases were
extremely improbable Z-scores determined by DHS.

3.4 Description of Variables
Outcome measures. The three anthropometric measurements described in
Chapter one (height-for-age, weight-for-height, and weight-for-age) were calculated by
dividing the provided percent Z-score by 100. DHS calculated percent Z-scores from the
median of the International Reference Population developed by the United States Centers
for Disease Control and recommended by the World Health Organization (Sommerfelt &
Stewart, 1994). Weight-for-height, was removed from multivariate analyses, however,
due to the small percent of children in the surveys that were classified as wasted.

Predictor variables. Nine predictor variables were chosen for the multivariate
analysis. These variables were adopted from a DHS comparative study, which used
numerous differentials to compare the nutritional status of children across countries

47

(Sommerfelt & Stewart, 1994). These researchers have extensive knowledge of
determinants of child health and an excellent working knowledge ofDHS data. The
variables were clear and could be easily adapted to this study. The variables were
classified into three categories: socio-economic, environmental and health. Each category
represented a level in the hierarchial relationship of the determinants of malnutrition.
The socio-economic variables included place of residence, maternal education,
maternal work status, radio ownership, and sex of child. Current place of residence was
defined as rural or urban according to the individual country's national statistical office.
Maternal education was defined as the highest level of schooling attended, regardless of
whether or not the woman completed the level. Secondary and higher education was
collapsed into one level because of the small number of women achieving higher
education in these surveys. Mother's work status was defined as whether or not the
woman was working at the time of the survey and if she was working for cash wages.
Radio ownership has been included as a proxy of household income (Sommerfelt &
Stewart, 1994). A family owning a radio is more likely to have greater financial security
than a family that does not own a radio. Furthermore, radio ownership increases exposure
to health education and information. Sex of the child was included to control for both for
the biological and the social importance.
The environmental variables included source of drinking water and sanitation
facilities. Both variables were important in addressing the synergistic relationship of
malnutrition and disease (Esrey, 1996). The variables were collapsed into three levels to
maintain simplicity. Moreover, surveys differed in the manner that the variables were
defined (eg. whether or not the wells were differentiated between protected or not

48
protected); these levels helped to maintain consistency between surveys. Source of
drinking water was categorized into piped water, well/pump water or surface water.
Sanitation facilities were categorized as flush toilets, latrines or no facilities.
Immunization status and diarrhea prevalence were included as indicators of
general health and also to address the disease-nutrition cycle. Immunization status was
classified into three categories; measles vaccination, other vaccination or no vaccination.
Measles was included as a separate level for immunization status due to the importance
of this vaccination with respect to nutritional status (Sommerfelt & Stewart, 1994;
Ulijasek, 1990). The surveys classified diarrheal illness according to whether or not the
child was ill during the preceding 24 hours or in the past two weeks. This study
dichotomized the variable as either diarrhea or no diarrhea.

3.5 Data Analysis Procedures
All analyses were performed using the Statistical Package for Social Sciences
(SPSS) version 8.0. The data analysis was conducted in two stages. In the first stage,
univariate techniques were used to describe patterns and trends of nutritional status over
time and to compare these trends between countries. Multivariate techniques were used in
the second phase to produce models with a set of predictor variables for malnutrition.
These models were used to compare patterns of key factors over time. An attempt was
made to determine if there are common determinants for all five developing countries.

Univariate Analysis:
The percent of children that were stunted, wasted and underweight were
calculated for each survey using the cut-off points of minus 2 standard deviations. These

49
measurements provided a general description of trends and nutritional changes. T -tests
were then performed to determine if there were significant changes in each
anthropometric measurement between the first and second survey of each country.
Separate T -tests were also performed to determine ifthere were changes in nutritional
status with respect to place of residence and sex of the child within each country as well
as over time. The statistical associations between all three anthopometric indicators were
examined by Pearson's correlations.

Multivariate Analysis:
Data analysis was guided by a conceptual framework of malnutrition (see Figure
3.0). Conceptual frameworks can be important in aiding multivariate analyses that study
complex hierarchial interelationships of determinants of disease (Victora, Huttly, Fuchs&
Olinto. 1997). The choice of factors and the ordering of variables were therefore not
based purely on statistical relationships. Relying on systematic, mechanical and
traditional algorithms (such as stepwise regression) fails to account for theoretical or
statistical relationships and can lead to various sources of error (Rothman & Greenland,
1998). The regression models were built by entering the variables in three steps following
the categories of socio-economic, environmental and health (see Figure 3.1). This method
illustrated how the factors influence each other.

50

Figure 3.0 Conceptual framework of risk factors for malnutrition in
developing countries
SOCIOECONOMIC
ENVIRONMENTAL

MATERNAL REPRODUCTIVE

GESTATIONAL
PERINATAL

BIRTHWEIGHT
CHILD CARE

DIET

INFECTIOUS DISEASE

PREVIOUS MORBIDITY

MALNUTRITION
Source: Adapted from V1ctora, Huttly, Fuchs & Olmto, 1997

Figure 3.1 Logistic Regression Models

Modell

Model2

Model3

Sex of child
Urban-rural residence
Maternal education
Mother's work status
Radio ownership

Sex of child
Urban-rural residence
Maternal education
Mother's work status
Radio ownership
Source of drinking water
Sanitation facilities

Sex of child
Urban-rural residence
Maternal education
Mother's work status
Radio ownership
Source of drinking water
Sanitation facilities
Immunization status
Diarrhea prevalence

Socioeconomic

Socioeconomic and
Environmental

Socioeconomic,
Environmental
and Health

A series oflogistic regressions were performed for each survey. The goal of this
analysis was to attempt to produce models that encompassed a set of predictor variables
that influence nutrition. Through these models, comparisons were made to examine
changes in nutritional status over time and among countries.

51
Logistic regression is an extremely popular and useful method in public health
research (Hosmer, Taber, & Lemeshow, 1991). This statistical method was chosen for
this analysis since the predictor variables are all discrete. Logistic regression allows the
use of discrete, continuous or a combination ofboth types of variables {Tabachnick &
Fidel, 1997). Moreover, logistic regression allows for the easy interpretation of odds
ratios, a key role in understanding the risk factors of malnutrition.
Hosmer, Taber, & Lemeshow (1991) stress the importance of assessing the fit of
logistic regression models before attempting to interpret results. Two elements were used
as measures of goodness-of-fit: chi-square of the model, and the classification table for
predicted versus observed. The chi-squares were significant for all of the stunting and
underweight models. Three surveys for the wasting models had chi-squares that were not
statistically significant. In most of the models classification was over 70% and many
were greater than 90%. The concern of these models, however, was the poor
classification of cases (stunted, underweight, wasted). The wasted models had extremely
poor classification of cases, this is most likely due to the small prevalence of wasting in
the samples. Due to the poor model fits, the wasting models were not analyzed in the
study.

3.6 Limitations of the Study
The present study had several limitations that should be noted. First of all, the
presentation of mean averages of each nutritional indicator at a national level was not
able to represent regional disparities within countries. Furthermore, the averages of the
anthropometric indicators were not able to account for reasons behind changes in these

52
indicators; the measurements were only able to describe changes and patterns of
nutritional status.
The goal of this study was not to produce an exhaustive list of determinants, but
rather to examine trends in a few key predictor variables. Many important factors are
therefore not examined in this study such as calorie intake, seasonal variation, and access
to health care. Moreover, these factors are not available in DHS surveys. The analysis is
also not able to address factors such as the impact of AIDS and structural adjustment
programs that have wiped out some of the progress made in public health throughout the
past decade.
The selected countries cover two regions in the developing world, Latin America
and sub-Saharan Africa, providing an interesting comparison for analysis. The small
sample size will, however, prevent generalizations to these regions and to other
developing nations. In particular, Asian countries where malnutrition has the highest
prevalence, were not available for analysis.
The small prevalence of wasting in the study samples led to inadequate logistic
regression models. This prevented an examination of the determinants of wasting. The
stunting and underweight models were also problematic. The underweight models, in
particular, contain some unexplainably high odds ratios. Data considerations for logistic
regression preceded the analysis; these included ratio of cases to variables, adequacy of
expected frequencies, multicollineariy and outliers (Tabachnick & Fidel, 1997). These
assumptions were adequately met. The chi-square tests indicated that the models were
appropriate for analysis. The only explanation that can be extended to rationalize the
problematic logistic regression results is the 'nature of the data'.

53
Chapter Four: Research Findings
4.0 Introduction
This chapter presents the results of the data analyses described in chapter Three.
The structure of the chapter is as follows . The first section is an overview ofthe tables
and figures , which present the main results of the study. The second section summarizes
the descriptive statistics. The third section summarizes the univariate results. The fourth
section is a summary ofthe descriptive and univariate results. The fifth section presents
the multivariate results followed by a summary of these results. All univariate and
multivariate results were considered significant at the .05 level, unless otherwise
specified.
4.1 Compilation of Tables and Figures
The results of the analyses are summarized in several tables and figures. Table 4.0
presents the percent of children undernourished for each anthropometric indicator. Table
4.1 is a more complex description of the anthropometric indicators. Figure 4.0
distinguishes anthropometric indicators according to residence and Figure 4.1 according
to sex of the child. Figures 4.2 and 4.3 summarize T-test results of intra-country
comparisons for place of residence and sex of the child. Table 4.2 summarizes T-test
results of the changes in nutritional status over time; similar tables are presented with
respect to place of residence (Table 4.3) and sex ofthe child (Table 4.4). Figure 4.5
shows the predictor variables that were significant for each logistic regression model with
stunting as the outcome variable and a similar presentation is seen in Figure 4.6 using
underweight as the outcome variable. Odds ratios for full logistic regression models can
be seen in Table 4.7 for stunting and in Table 4.8 for underweight children.

54

4.2 Descriptive Statistics
Results indicate several marked patterns in nutritional status (Table 4.0). The
countries display similar patterns of undernutrition, with the exception ofMali. During
the 1980's and 1990's, Brazil, Guatemala, Uganda and Zimbabwe have a larger percent of
stunted than underweight children combined with a low percent of wasted children. Over
time, three trends occur. First, the countries show a decrease in the prevalence of
stunting. Second, the countries show an increase in the prevalence of wasting. Third,
Latin American countries show a decrease in the prevalence of underweight children,
while African countries show increases in the percent of children who are underweight.
Mali displays a higher prevalence of children underweight than children who are stunted
combined with high rates of wasting. Over time, this pattern is maintained but there is an
increase in all three indicators.
Further categorizing as suggested by W aterlow and his associates ( 1977) leads to
a richer description of nutritional status (Table 4.2). This categorization illustrates that the
percent of children stunted and wasted increases over time (except Northeast Brazil). The
percent of children that are underweight but are not stunted or wasted also increases
(except Mali). The percent of children, who are not stunted, wasted or underweight is an
indicator of good nutritional status. Northeast Brazil displays quite a large increase in this
indicator, while Guatamala, Uganda and Zimbabwe show only a slight increase. Mali
shows an almost 20% decrease.
Place of residence clearly indicates differences in the levels ofundernutrition
(Figure 4.0). The percent of children stunted and underweight are greater in rural areas
than in urban areas. The levels of wasting vary according to country and time. Over time,

55
the percent of children who were wasted increases in both rural and urban areas. The
changes in prevalence of children stunted and underweight varied between surveys.
The nutritional indicators are compared with respect to the sex of the child in
Figure 4.1. Generally there is a higher prevalence of undernourishment in male compared
to female children. This gender gap, however, is small. Over time, male children show
decreases in stunting (except in Mali) and increases in wasting. The percentage ofboys
underweight decreases in the Latin America countries but increases in African countries.
Female children also demonstrate increases in wasting but showed no consistent pattern
changes for the other two indicators.

Table 4.0 Anthropometric Indicators
Country

Year

Brazil (NE)
Brazil (NE)
Guatemala
Guatemala
Mali
Mali
Uganda
Uganda
Zimbabwe
Zimbabwe

1986
1996
1987
1995
1987
1996
1988
1995
1988
1994

%
Stunted
29.2
17.2
57.7
53.7
22.9
33.0
42.5
34.8
29.8
24.0

%
Wasted
0.9
2.8
1.3
4.5
11.3
24.8
2.0
6.2
1.2
5.9

%
Underweight
14.0
8.0
33.2
32.3
29.9
43.6
23 .7
26.2
12.6
17.5

Sample
Size
644
1306
2207
4908
948
4263
2297
3350
1496
1907

22.7

1994

*Total % all approximate 100.0.

1988

29.3

32.4
4.6

0.7
3.7
0.7

41.1

1995

1988

9.1

15.7

7.2

23.9

1996

1987

2.7

1.8

18.8

51.0

1.3

0.5

2.5

1.4

4.1

0.7

0.4

0.8

%
Stunted and
Wasted

2.5

0.6

57.0

1995

1987

16.9

0.2

28.4

1986

Brazil (NE)
Brazil (NE)
Guatemala
Guatemala
Mali
Mali
Uganda
Uganda
Zimbabwe
Zimbabwe

1996

%
Wasted
Only

%
Stunted
Only

Year

Country

3.4

1.9

3.4

2.0

5.7

6.3

2.2

1.4

1.6

0.9

%
Underweight
(not stunted
or wasted)

67.8

67.3

57.4

54.6

45.2

63 .3

42.1

40.1

78.5

69.7

%
Not stunted,
wasted or
underweight

Table 4.1 Combination of Anthropometric Indicators

56

99.8

99.7

99.4

99.8

99.6

99.7

99.8

99.8

99.9

100.0

Total*

1306

1907

1496

3350

2297

4263

948

4908

2207

644

Sample
Size

57
Figure 4.0 Anthropometric Indicators for Urban versus Rural Residence
BRAZIL
Year
Residence
Sample Size
%stunted
%wasted
% underweight
GUATEMALA
Year
Residence
Sample Size
%stunted
%wasted
% underweight
MALI
Year
Residence
Sample Size
%stunted
%wasted
% underweight
UGANDA
Year
Residence
Sample Size
%stunted
%wasted
% underweight
ZIMBABWE
Year
Residence
Sample Size
%stunted
%wasted
% underweight

1986

Urban

313
22.7
1.6
11.2
1987

Urban

615
46.8
1.3
25.2
1987

Urban

493
19.1
10.3
25 .8
1988

Urban

378
26.7
1.3
15.6
1988

Urban

350
16.0
1.4
6.9

1986
Rural
331
35.3
0.3
16.6
1987
Rural
1592
61.9
1.3
36.2
1987
Rural
455
27.0
12.3
34.3
1988
Rural
1919
45.6
2.2
25.3
1988
Rural
1146
34.0
1.1
14.4

1996

Urban

890
13.6
2.9
7.1
1995

Urban

1126
39.3
3.4
20.7
1996

Urban

1306
26.1
26.5
37.7

1995

Urban

917
24.3
5.5
18.0
1994

Urban

424
17.5
5.4
11.3

1996
Rural
416
25.0
2.6
10.1
1995
Rural
3782
58.0
4.8
35.6
1996
Rural
2957
36.0
24.1
46.2
1995
Rural
2433
38.8
6.5
29.3
1994
Rural
1483
25.8
6.0
19.6

58

BRAZIL
Year
Sex of child
Sample Size
%stunted
%wasted
% underweight

Figure 4.1 Anthropo metric Indicators for Sex of Child

GUATEMALA
Year
Sex of child
~
Size
%stunted
%wasted
% underweight
MALI
Year
Sex of child
Sample Size
%stunted
%wasted
% underweight
UGANDA
Year
Sex of child
Sample Size
%stunted
%wasted
% underweight
ZIMBABWE
Year
Sex of child
Sample Size
%stunted
%wasted
% underweight

1986

Male

322
31.7
1.2
14.6
1987

Male

1091
58.7
1.1
32.4

1987

Male

497
22.7
14.3
30.2
1988

Male

1124
44.7
2.0
24.6

1988

Male

729
31.0
1.5
13.6

1986
Female
322
26.7
0.3
13.4
1987
Female
1116
56.7
1.5
33.9
1987
Female
451
23.1
8.0
29.5
1988
Female
1173
40.4
2.0
22.8
1988
Female
767
28.7
0.9
11.7

1996

Male

651
20.0
2.8
9.1
1995

Male

2457
55.0
5.3
32.6
1996

Male

2101
33.6
25.8
42.7
1995

Male

1623
37.4
7.0
28.8

1994

Male

936
24.6
7.4
19.8

1996
Female
655
14.5
2.9
7.0
1995
Female
2451
52.4
3.6
31.9
1996
Female
2162
32.4
23.8
44.4
1995
Female
1727
32.4
5.5
23.7
1994
Female
971
23.4
4.4
15.2

59

4.3 Univariate Analysis
Intra-country comparisons indicate statistically significant differences between urban and
rural residence for stunting and underweight status (Figure 4.2). Place of residence for wasting,
however, is significantly different only in Uganda and Zimbabwe. Examining differences according
to sex of the child indicates that in most cases there are no statistically significant differences
(Figure 4.3). Uganda and Zimbabwe display the highest rates of sex differences.
There are significant changes in nutritional status over time for all countries (Table 4.2).
When distinguishing place of residence there are changes in anthropometric indicators over time for
urban and rural residence (Table 4.3). There were, however, three exceptions: wasting in rural
Brazil (NE), wasting in urban Guatemala and stunting in urban Zimbabwe. There are also statistical
differences in nutritional status when examining sex of the child over time (Table 4.4). The cases
that are not statistically significant include wasted females in Brazil (NE) and Guatemala, wasted
males in Brazil (NE) and underweight males in Uganda.
Table 4.5 shows the results of the Pearson's correlations between the prevalences of stunting
and wasting, underweight and wasting, and stunting and underweight for each country during the
1980's and 1990's. There are strong direct correlations between the prevalences of stunting and
underweight, weaker correlations between wasting and underweight and no association between
stunting and wasting. There were increases in the magnitude of the correlations between wasting
and underweight for each country during the 1990's. The correlations between stunting and
underweight decreased over time, except in Zimbabwe.

60
Figure 4.2 T -tese Results: Urban versus Rural Residence (Intra-Country Comparisons)
Year

Indicator

1986
1996
1986
1996
1986
1996

stunted
stunted
underweight
underweight
wasted
wasted

Year

Indicator

1987
1995
1987
1995
1987
1995

stunted
stunted
underweight
underweight
wasted
wasted

Year

Indicator

1987
1996
1987
1996
1987
1996

stunted
stunted
underweight
underweight
wasted
wasted

Year

Indicator

1988
1995
1988
1995
1988
1995

stunted
stunted
underweight
underweight
wasted
wasted

Year

Indicator

1988
stunted
stunted
1994
1988
underweight
underweight
1994
1988
wasted
1994
wasted
1
two-tailed independent T-test

BRAZIL
t(df)
4.50(642)
6.50(1304)
3.56(642)
4.78(1304
0.63(642)
~

GUATEMALA
t(df)
6.97(2205)
13.3(4906)
5.96(2205)
10.4(4906)
1.51(2205)
~

~

MALI
t(df)
2.83(946)
6.83(4261)
2.69(946)
4.69(4261)
1.09(946)
-

~

UGANDA
t(df)
0.92(2295)
9.23(3348)
0.89(2295)
8.24(3348)
0.30(2295)
2.94{3348l
ZIMBABWE
t(df)
8.31(1494)
3.66(1905)
8.71(1494)
5.69(1905)
9.81(1494)
~

p-value
.000
.000
.000
.000
.266
.496
p-value
.000
.000
.000
.000
.066
.026
p-value
.003
.000
.004
.000
.137
.309
p-value
.000
.000
.000
.000
.000
.002
p-value
.000
.000
.000
.000
.000
.000

61
Figure 4.3 T-test 1 Results: Male versus Female Child (Intra-Country Comparisons)
Year

Indicator

1986
1996
1986
1996
1986
1996

stunted
stunted
underweight
underweight
wasted
wasted

Year

Indicator

1987
1995
1987
1995
1987
1995

stunted
stunted
underweight
underweight
wasted
wasted

Year

Indicator

1987
1996
1987
1996
1987
1996

stunted
stunted
underweight
underweight
wasted
wasted

Year

Indicator

1988
1995
1988
1995
1988
1995

stunted
stunted
underweight
underweight
wasted
wasted

Year

Indicator

stunted
1988
stunted
1994
underweight
1988
1994
underweight
wasted
1988
1994
wasted
1
two-tailed independent T-test

BRAZIL
t(df)

p-value

-.515(642)
-2.39(1304)
-.113(642)
-1.14(1304)
0.10(642)
0.21(1304)

.304
.009
.455
.129
.459
.418

GUATEMALA
t(df)

p-value

-0.21(2205)
-2.41(4906)
-0.14(2205)
-2.27( 4906)
-0.53(2205)
-1.67(4906)

.419
.008
.445
.012
.298
.047

MALI
t(df)

p-value

-.386(946)
-1.14(4261)
-1.40(946)
0.09(4261)
-2.42(946)
-

~

~

UGANDA
t(df)
-3 .29(2295)
-3.62(3348)
-2.26(2295)
-2.45(3348)
-0.90(2295)
-

~

.350
.127
.081
.465
.008
.440
p-value
.001
.000
.012
.008
.184
.211

ZIMBABWE
t(df)

p-value

-2 .06(1494)
-1.20(1905)
-0.18(1494)
-2.83(1905)
1.19(1494)
-3.47{1905}

.020
.115
.428
.003
.118
.001

62

Table 4.2 T-Test 1 Results for Changes in Nutritional Status over Time
Country

Indicator

*t(df)

p-value

Brazil (NE)
Brazil (NE)
Brazil (NE)
Guatemala
Guatemala
Guatemala
Mali
Mali
Mali
Uganda
Uganda
Uganda
Zimbabwe
Zimbabwe
Zimbabwe

Stunted
Underweight
Wasted
Stunted
Underweight
Wasted
Stunted
Underweight
Wasted
Stunted
Underweight
Wasted
Stunted
Underweight
Wasted

-8.95(1320.1)
-4.52(1200.1)
1.90(1443.7)
-4.58(4248.1)
-2.42(4486.8)
2.44(5185.5)
5.73(1561.3)
8.65(1477.9)
8.38(1631.0)
-6.24(4952.2)
2.30(51 04.4)
38.5(5584.1)
-6.09(3289 .8)
4.56(3277.5)
11.7 (3339.7)

.000
.000
.029
.000
.008
.008
.000
.000
.000
.000
.011
.000
.000
.000
.000

* equal variances were not assumed
1

two-tailed independent T-test

63

Table 4.3 T -test 1 Results in Changes of Nutritional Status over Time:
Rural versus Urban Residence
Country

Residence

rural
Brazil (NE)
Brazil (NE)
rural
Brazil (NE)
rural
urban
Brazil (NE)
Brazil (NE)
urban
Brazil (NE)
urban
rural
Guatemala
rural
Guatemala
Guatemala
rural
Guatemala
urban
urban
Guatemala
Guatemala
urban
Mali
rural
Mali
rural
Mali
rural
Mali
urban
Mali
urban
Mali
urban
Uganda
rural
Uganda
rural
Uganda
rural
Uganda
urban
Uganda
urban
Uganda
urban
Zimbabwe
rural
Zimbabwe
rural
Zimbabwe
rural
Zimbabwe
urban
Zimbabwe
urban
Zimbabwe
urban
*equal variances not assumed
1
two-tailed independent T-test

Indicator

*t(df)

p-value

Stunted
Underweight
Wasted
Stunted
Underweight
Wasted
Stunted
Underweight
Wasted
Stunted
Underweight
Wasted
Stunted
Underweight
Wasted
Stunted
Underweight
Wasted
Stunted
Underweight
Wasted
Stunted
Underweight
Wasted
Stunted
Underweight
Wasted
Stunted
Underweight
Wasted

-4.68(723 .0)
-2.51(684.6)
0.97(742 .5)
-5 .70(549 .8)
-2.30(503 .3)
1.68(589.4)
-3.67(3049.3)
-1.89(3180.1)
2.06(3713.3)
-4.40(1171.0)
-2.82(1293.2)
1.05(1442.3)
3.67(645.7)
5.24(609.0)
5.31(680.8)
2.27(999 .0)
5.37(1014.4)
6.29(1082.6)
-4.41(4126.4)
2.89(4213 .0)
9.88(4304.0)
-1.89(692.4)
2.49(726.1)
5 .97(751.0)
-7.22(2568.4)
2.83(2546.0)
10.5(2586.0)
-.03(741. 7)
4.23(748.3)
5.24(756.1)

.000
.006
.166
.000
.011
.05
.000
.030
.020
.000
.003
.147
.000
.000
.000
.012
.000
.000
.000
.002
.000
.030
.007
.000
.000
.003
.000
.490
.000
.000

64
Table 4.4 T-test 1 Results for Changes in Nutritional Status over Time:
Male versus Female
Country

Sex of child

Brazil (NE)
Male
Brazil (NE)
male
Brazil (NE)
male
Brazil (NE)
female
Brazil (NE)
female
Brazil (NE)
female
Guatemala
male
Guatemala
male
Guatemala
male
Guatemala
female
Guatemala
female
Guatemala
female
Mali
male
Mali
male
Mali
male
Mali
female
Mali
female
Mali
female
Uganda
male
Uganda
male
Uganda
male
Uganda
female
Uganda
female
Uganda
female
Zimbabwe
male
Zimbabwe
male
Zimbabwe
male
Zimbabwe
Female
Zimbabwe
Female
Zimbabwe
Female
*equal variances not assumed
1
two-tailed independent T-test

Indicator

t(df)

p-value

Stunted
Underweight
Wasted
Stunted
Underweight
Wasted
Stunted
Underweight
Wasted
Stunted
Underweight
Wasted
Stunted
Underweight
Wasted
Stunted
Underweight
Wasted
Stunted
Underweight
Wasted
Stunted
Underweight
Wasted
Stunted
Underweight
Wasted
Stunted
Underweight
Wasted

-5.57(665.4)
-2 .76(577.3)
1.30( 699 .5)
-7.10(649 .7)
-3 .66(623 .0)
1.38(742.3)
-2.45(2103.8)
-0.86(2227.6)
2.18(2589.6)
-4.00(2141.8)
-2.57(2254.6)
1.26(2591.2)
4.25(822.1)
5.34(795 .6)
4.55(868 .7)
3.88(737.3)
6.94( 684.1)
7.49(767.9)
-4.58(2458 .1)
1.55(2551 .6)
7.66(2674.4)
-4.24(2492.9)
1.72(2544.3)
7 .95(2584.7)
-4.78(1622.9)
4.46(1592.3)
10.5(1625.1)
-3.87(1664.8)
2.00(1680.0)
6.08(1709.5)

.000
.003
.098
.000
.000
.084
.007
.195
.015
.000
.005
.103
.000
.000
.000
.000
.000
.000
.000
.061
.000
.000
.043
.000
.000
.000
.000
.000
.023
.000

65

Table 4.5 Correlation of the Prevalences of Anthropometric Indicators, by Survey
Survey
Brazil1986
Brazil 1996
Guatemala 19 87
Guatemala 1995
Mali 1987
Mali 1996
Uganda 1988
Uganda 1995
Zimbabwe 1988
Zimbabwe 1994

Stunted!Was ted
.115*
-.017
-.014
.024
.115*
.044
.075*
.027
.022
-.010

Underweight/Wasted
.241 *
.272*
.122*
.286*
.423*
.461 *
.201 *
.345*
.216*
.337*

Stunted/Underweight
.559*
.380*
.526*
.480*
.583*
.506*
.539*
.517*
.430*
.444*

*Correlation is significant at the 0.01 level (2-tailed).

4.4 Summary of Descriptive and Univariate Results
Results indicate that there were significant changes in nutritional status between the 1980's
and 1990's. The type of change varies depending on the choice of indicator. There was a decrease
in stunting (except Mali) but an increase in wasting. There was a decline in the percent of children
underweight in the Latin American countries but there were increases in the African countries.
Place of residence is an important differential in nutritional status, there are higher rates of children
undernourished living in rural areas than in urban areas. Over time, nutritional changes occur in
both urban and rural areas. Boys are more likely to be undernourished compared to girls. These
differences, however, are small and in most cases are not statistically significant.

66
4.5 Multivariate Results

Mode/fits:
The original logistic regression models had relatively poor fits; specifically there was poor
classification of the models. An inspection of the standardized residuals indicated 0-5% of the
sample were outliers (data points that were outside 2 standard deviations), these were subsequently
removed to determine their influence. Although outliers have been discussed widely in the
literature, there is little information and guidelines concerning multivariate outliers (Chatterjee &
Yilmaz, 1992). The removal of the outliers changed the results and improved the overall fit of the
models. It was therefore deemed appropriate to remove the outliers, which reflect the results
presented here.

Logistic Regression Results:
The logistic regression results were analyzed in two parts: (1) All three models for each
survey were examined to determine patterns of variables that were statistically significant in the
models. These were compared over time and across countries. (2) The full model was selected for
all surveys and odds ratios were compared to examine the quantitative contribution of each
predictor variable.

Part 1

Stunting
The results of the logistic regression models using stunting as the predictor
variable are summarized in Figure 4.4. The models display various patterns across
country and over time.

67
Two variables, which are common to all countries for Modell include place of
residence, and maternal education. Sex of the child is included for all countries except
Mali. Radio ownership is included for Guatemala, and the African countries. In Model 2
sanitation facilities is significant except in Mali. Source of drinking water is included for
Brazil and Uganda. In Model3, immunization status is included for all countries and
diarrhea prevalence is included for all African countries. Although each sequential model
typically maintains the variables from the previous models, place of residence displays
two different patterns. In Brazil and Zimbabwe, residence is included in Model 1, but is
no longer significant in Models 2 and 3. Mali and Uganda maintain place of residence for
all three models. Guatemala displays the first pattern during the 1980's and the second
pattern during the 1990's.
Over time the variables in Model 1 did not change, except with some differences
with respect to sex of the child and radio ownership. In Model2, source of drinking
water, which is not included in the 1980's models appears in the 1990's models. Type of
sanitation facilities is more prominent during the 1980's than the 1990's.

Underweight

The results of the logistic regression models using stunting as the predictor
variable are summarized in Figure 4.5. These models are not as comparable as the
stunting models. Brazil, Mali and Zimbabwe have very few variables included in the
1980's models. There are, however, a few observations that can be made.
Radio ownership and place of residence are common variables in Model 1.
Maternal education is significant, except in Brazil. Maternal work status is significant for

68
the Latin American countries and Uganda. Model 2 includes type of sanitation facilities
for Guatemala, Uganda and Zimbabwe, and source of drinking water for Brazil, Uganda
and Zimbabwe. Model 3 includes immunization status for all countries, although the
measles vaccination appears more frequently. Diarrhea prevalence is also common to all
countries.
In general, the models contain more variables during the 1990's compared to the
1980' s. In particular, place of residence and maternal education become more prominent.
Unlike the stunting models, the addition of the environmental variables did not alter place
of residence. Environmental variables are not significant during the 1980's (except in
Guatemala). During the 1990' s type of sanitation facilities and source of drinking water
are included in Models 2 and 3. The health variables remain, relatively unchanged.

69

Figure 4.4 Summary of the Significant Predictor Variables for each
Logistic Regression Model: Stunting Models
~

~

Modell

Model2

-

Res
Edu(l)
Edu(2)
Sex
Edu(l)
Sex
Sanit(l)
Sanit(2)

Model3
Edu(l)
Sex
San it(!)
Sanit(2)
lmrnun(2)

GUATEMALA-87
Model I
Res
Edu(l)
Edu(2)
Radio

MALI-87
Modell
Res
Edu(l)
Edu(2)
Radio

Model2

Model2
Res
Edu(l)
Edu(2)
Radio

UGANDA-88
Modell
Res
Edu(l)
Edu(2)
Radio
Sex
Model2
Res
Radio
Sex
Sanit(2)

Model3

Model3

Model3

Edu(l)
Edu(2)
Radio
San it( !)
Sanit(2)

Edu(l)
Edu(2)
Radio
Imrnun(l)
Imrnun(2)
Diarrhea

Edu(l)
Edu(2)
Radio
San it(!)
Sanit(2)
lmrnun(l)
lmrnun(2)
GUATEMALA-95
Modell
Res
Edu(l)
Edu(2)
Sex

MALI-96
Modell
Res
Edu(l)
Edu(2)
Radio

Mode\2
Edu(l)
Edu (2)
Sex
Sanit(2)
H20( I)
H20(2)

Model 2

Model2

Model3

Model3

~

-

Model I
Res
Edu(l)
Edu(2)
Sex

Edu(l)
Edu(2)
Work
Sex
Sanit(2)
H20(!)
H20(2)
lmrnun(l)
Imrnun(2)
Diarrhea

Res
Edu(l)
Edu(2)
Sex
San it(!)
Sanit(2)
Res
Edu(l)
Edu(2)
Sex
San it(!)
Sanit(2)
Imrnun(l)
Imrnun(2)

Model3

Res
Edu(l)
Edu(2)

Res
Edu(l)
Edu(2)
Imrnun(l)
Immun(2)
Diarrhea

Res
Radio
Sex
Sanit(2)
Imrnun(l)
Imrnun(2)
Diarrhea
UGANDA-95
Modell
Res
Edu(l)
Edu(2)
Radio
Sex
Model2
Res
Edu(l)
Edu(2)
Radio
Sex
H20(!)
H20(2)
Model 3
Res
Edu(l)
Edu(2)
Radio
Sex
H20(2)
Imrnun(l)
Imrnun(2)
Diarrhea

ZIMBABWE-88
Modell
Res
Edu(l)
Edu(2)
Radio
Sex
Model2
Edu(l)
Edu(2)
Radio
Sex
Sanit(l)
Model3
Edu(l)
Edu(2)
Radio
Sex
San it(!)
Imrnun(2)
Diarrhea
ZIMBABWE-94
Modell
Res
Edu(l)
Edu(2)
Model2

Model3

Edu(l)
Edu(2)

Edu(l)
Edu(2)

70
Figure 4.5: Summary of the Significant Predictor Variables for each
Logistic Regression Model: Underweight Models
(NE)BRAZIL-86
Modell

GUATEMALA-87
Modell
Edu(l)
Edu(2)
Radio

MALI-87
Modell
Radio

Mode12

Model2

Model2

-

Mode13

-

Immun(l)
Diarrhea

(NE)BRAZIL-96
Modell
Work
Res
Sex
Model2

Model3

Res
Work
Radio
Sex
H20(2)

Res
Work
Radio
Sex
H20(2)
Immun(l)
Diarrhea

Model 3

Edu(l)
Edu(2)
Radio
San it(!)
Sanit(2)

Edu(l)
Edu(2)
Radio
San it(!)
Sanit(2)
Immun(l)
Immun(2)
GUATEMALA-95
Modell
Res
Edu(l)
Edu(2)
Work
Radio
Model2
Res
Edu(l)
Edu(2)
Work
Radio
Sanit(l)
Sanit(2)
Model3

Res
Edu(l)
Edu(2)
Radio
San it(!)
Sanit(2)
Immun(l)
Immun(2)
Diarrhea

Model3

Radio

Radio
Immun(2)
Diarrhea

MALI-96
Modell
Res
Edu(l)
Edu(2)
Radio
Mode12

Mode13

Res
Edu(l)
Edu(2)
Radio

Res
Edu(l)
Edu(2)
Work
Radio
Immun(l)
Immun(2)
Diarrhea

UGANDA-88
Modell
Res
Edu(l)
Edu(2)
Work
Radio
Model2
Res
Edu(l)
Work
Radio
Sex
Model3
Res
Edu(l)
Work
Radio
Immun(l)
Immun(2)
Diarrhea
UGANDA-95
Modell
Res
Edu(l)
Edu(2)
Radio
Sex
Model2
Res
Edu(l)
Edu(2)
Radio
Sex
H20( I)
H20(2)
San it(!)
Sanit(2)
Mode13
Res
Edu(l)
Edu(2)
Radio
Sex
H20(!)
H20(2)
San it(!)
Sanit(2)
Immun(2)
Diarrhea

ZIMBABWE-88
Modell
Radio

Model2

Mode13

Res
Radio

Res
Radio
Immun(2)
Diarrhea

ZIMBABWE-94
Modell
Res
Edu(l)
Edu(2)
Radio
Sex
Model2
Res
Edu(l)
Edu(2)
Radio
Sex
H20(!)
H20(2)
San it( I)
Sanit(2)
Model3
Edu(l)
Edu(2)
Radio
Sex
H20( I)
H20(2)
San it(!)
Sanit(2)
Immun(2)
Diarrhea

71

Part II
To keep the results clear and simple, only the results of the full logistic regression
models are presented. The full model allows for the examination of the socioeconomic
variables controlling for environmental and health variables, which is the main interest of
this study. Moreover, the results of the first two models do not contribute any further
meaning to the study.
The logistic results are presented in the form of odds ratios, an approximation of
relative risk. For simplicity, only the odds ratios for the significant variables are
presented. (see Table 4.7 for stunting models and Table 4.8 for underweight models). For
greater details concerning logistic regression results, refer to Appendix B. There is great
variation in the odds ratios across countries and over time. There are, however, a couple
of patterns that emerge.

Interpreting Odds Ratios
An odds ratio that is greater than 1 indicates that the child for the given
determinant is more likely to be considered stunted or underweight. An odds ratio that is
less than 1 indicates that the variable is decreasing the probability of
stunting/underweight. A value of 1 indicates that there is no effect. Table 4.6 presents the
coding utilized for the analyses.

72

Table 4.6 Predictor Variable Coding
Residence (O=urban, 1=rural)
Maternal Education (O=no education, 1=primary, 2=secondary or higher)
Maternal Work Status (O=no, 2=yes)
Radio ownership (O=no, 1=yes)
Sex of child (O=rnale, 1=female)
Sanitation facilities (O=no, 1=latrine, 2=flush)
Source of drinking water (O=surface, 1=well, 2=piped)
Immunization status (O=no vaccinations, 1=other vaccinations, 2=rneasles)
Diarrhea (O=no, 1-yes)

Stunting
Generally, there is an increase in the odds ratios for both primary and secondary
maternal education between the 1980's and 1990's. Across countries, maternal education
consistently has the highest odds ratios, with primary education having higher odds ratios
than secondary education. The other variables do not show significant changes in the
magnitude ofthe odds ratios.

Underweight
There appears to be problems with the data because some of the odds ratios are
extremely high, interpretation of the models should be done with extreme caution.
Maternal education displays a similar pattern to the stunting models in that the odds ratios
tend to increase between the first and second survey. Radio ownership, a common
variable across countries (except NE Brazil 1986) demonstrates a slight decrease in the
magnitude ofthe odds ratios between the 1980's and 1990's.

73

Table 4. 7 Odds ratios for Stunting
Stunting

Year

Brazil(NE)

Guatemala

Mali

Uganda

Zimbabwe

Residence

80's
90's

-

0.79

0.61

0.63
0.57

-

Education
(primary)

80's
90's

3.30
11.96

5.22
7.97

0.11
7.87

2.86

2.69
2.37

Education
(secondary)

80's
90's

-

2.90
6.03

0.10
5.98

2.35

2.07
2.35

Work

80's
90's

-

1.73

-

-

-

-

Radio

80's
90's

-

1.34
-

2.23
-

1.33
1.20

1.48
-

Sex

80's
90's

1.63
2.49

1.14

-

1.21
-

1.29

Sanitation
(latrine)

80's
90's

2.37
-

2.62
2.30

-

-

3.51

Sanitation
(flush)

80's
90's

0.03
3.46

2.48
2.35

-

0.17
-

-

Water source
(well)

80's
90's

6.18

-

-

-

-

Water source
(piped)

80's
90's

6.93

-

Immunization
(other)

80's
90's

3.00

Immunization
(measles)

80's
90's

Diarrhea

80's
90's

- variable not significant

6.28

-

-

-

-

-

-

0.49

-

0.52
0.64

2.11
0.66

0.65
0.64

-

0.53
0.26

0.40
0.40

0.37
0.40

0.69
0.53

0.50

0.63

0.63

0.57
0.61

0.81
0.72

0.64

-

74

Table 4.8 Odds Ratios for Underweight Status
Underweight

Year

Brazil(NE)

Guatemala

Mali

Uganda

Zimbabwe

Residence

80's
90's

4.53

0.56

0.82

0.34
0.47

38.01
-

Education
(primary)

80's
90's

-

4.81
59.72

2.16

-

3.04
3.77

9.24

Education
(secondary)

80's
90's

-

2.58
43.03

1.52

3.01

-

4.36

Work

80's
90's

9.24

-

1.14

0.50
-

-

Radio

80's
90's

19.39

1.26
1.18

1.52
1.28

1.90
1.47

10.19
0.20

Sex

80's
90's

6.41

-

-

1.47

2.42

Sanitation
(latrine)

80's
90's

-

8.62
2.40

-

0.32

168.2

Sanitation
(flush)

80's
90's

-

7.47
2.23

-

-

0.23

90.01

Water source
(well}

80's
90's

-

-

-

-

4.73

0.11

Water source
(piped}

80's
90's

9.56

-

-

4.11

0.21

Immunization
(other)

80's
90's

9.59
7.46

0.71
0.69

-

0.62

0.63
-

-

Immunization
(measles)

80's
90's

-

0.41
0.43

0.50
0.61

0.55
0.73

0.09
0.01

Diarrhea

80's
90's

0.38
0.10

0.70

0.43
0.55

0.50
0.53

0.20
0.35

- Variable not significant

-

-

-

75

4.6 Summary of Multivariate Results
Findings of the stunting models are more consistent than the underweight models.
There appears to be some data problems with the underweight models because some of
the odds ratios are extremely high. Both models are examined, but caution should be
extended to the interpretation of the underweight models.
Controlling for environmental and health variables, maternal education displays
the most consistent pattern of all the socioeconomic variables for stunting, both over time
and across countries. Maternal education appears more frequently than any other variable
in the models and also has the highest odds ratios among all the statistically significant
variables. Although both levels of education are important, primary education is the
'stronger' predictor. These patterns are similar, but not as prominent for the underweight
models.
Model 1 for stunting consistently contains place of residence as a significant
variable. Place of residence changes, however, with the addition ofvariables. In some
cases residence is removed with the addition of environmental and health variables and in
other cases residence remains significant. For the underweight models, place of residence
is not displaced with the addition of other variables.
Radio ownership proves to be an important predictor for the stunting models and
even more important for the underweight models. There is no meaningful pattern for sex
of the child or work status of the mother. The addition of environmental variables
indicates that sanitation is a more common predictor than source of drinking water.
Source of drinking water, however, becomes significant for some countries during the
1990's. Diarrhea is an important predictor, particularly for the African countries.

76

Immunization status is often included in the models, with measles vaccinations appearing
more frequently than other vaccinations.

77

Chapter Five: Discussion and Conclusion
In the House of Hunger diseases were the strange
irruptions of a disturbed universe

- Dambudzo Marechera, The House of Hunger (1975)
5.0 Introduction
The structure of the final chapter is as follows. The first section discusses the
results of the data analyses and places them in context with other studies. The second
section integrates the results of the study in the context of health development and
examines the role of national governments and the international health community. The
final sections provide recommendations for future research and policy directions.
5.1 Discussion of Results
Changes in the Nutritional Status of Children

The first objective of the analysis of the DHS data was to examine the changes in
the nutritional status of children over the past decade, and to compare these changes
across countries. The aim of the analysis was to determine ifthere has been progress in
decreasing child malnutrition in developing countries. The progress was considered with
respect to the goals of the 1990 World Summit for Children which sought a 50 %
worldwide reduction ofthe 1990 PEM levels by the year 2000 (United Nations, 1990). In
this study, progress was examined between the mid 1980's to the mid 1990's. UNICEF
(1995) states that for nations to achieve the goal of the World Summit, a minimum 20%
decrease in malnutrition is required by 1995.
The results of the study indicate that there are significant changes in the
nutritional status of children between the 1980's and 1990's for all five countries. The

78
improvements in child nutrition, however, are modest at best. The one exception is
Northeast Brazil which demonstrates an almost 10% decrease in the percent of children
undernourished. There is an approximate 50% decrease in children who are both stunted
and underweight. This is an important reduction because the Northeast is the most
impoverished region in Brazil. It is surmised that national figures would show an even
greater reduction in malnutrition.
In Guatemala, there are decreases in the percent of undernourished children, but
these changes are very small. This lack of progress is extremely troublesome in
Guatemala due to their initially high rates of malnourishment. Over 50% of the children
under five sampled during both the 1980's and 1990's are stunted. High rates of
undernutrition in Guatemalan children are noted elsewhere (Pebley & Goldman, 1995; de
Orris, Monteiro & Clugston, 1993). In Mali, there are obvious signs of deterioration in
nutritional status. There is an almost 20% increase in children who are malnourished.
There are large increases in all three nutritional indicators. Uganda and Zimbabwe
display mixed results. There is a slight decrease in the overall percentage of children
undernourished; there is a significant reduction in stunting but an increase in children
who are underweight. These results suggest that there has not been adequate progress in
the reduction of malnutrition but rather stagnation and even deterioration.
The deterioration of health in Africa has been noted elsewhere (UNICEF, 1998;
Thompson, 1996; Stamps, 1993). Three trends of malnutrition have been observed in
Africa: (1) static or a slight increase in prevalence; (2) trend appears static but there has
been a recent peak in incidence (majority); and (3) clear evidence of increasing
prevalence of underweight (Thompson, 1996). The increase in malnutrition follows the

79
economic decline of the 1980's. Social and health spending have been cut in many
government budgets in accordance to the structural adjustment programs that were put in
place. There have also been declines in per capita incomes, reducing household food
expenditure. Stamps (1993), the Minister ofHealth in Zimbabwe, has attributed the
increase in chronic malnutrition in Zimbabwe to the drought that occurred during the
early 1990's, and the explosion ofthe AIDS virus.
Children under 5 may develop AIDS either by perinatal transmission or through
contaminated breast-milk. Malnutrition is linked with infectious disease, including AIDS.
One of the main clinical manifestations of AIDS is wasting (Ball, 1998). Patterns of
malnutrition cannot be distinguished from patterns of weight loss associated with AIDS .
Studies conducted in sub-Saharan Africa found that 25% of children with malnutrition
were inflicted with the human immunodeficiency virus (HIV) virus (Ball, 1998).
Paediatric AIDS is a growing concern in developing countries. In 1997, over 500 000
children under the age of 15 became infected with the AIDS virus; 95% of the children
were under 5 (WHO, 1998).
Many previous studies use one anthropometric indicator to disseminate
information with respect to the nutritional status of children at a community, regional or
national level. The nutritional status of children, however, depends on the choice of
indicator utilized. The results of this study indicate that there are national differences in
the prevalence of each of the three nutritional indicators. The results also confirm that
there are common patterns of nutritional status among countries. This is consistent with
the studies demonstrating global (Sommerfelt & Stewart, 1994; Victora, 1992) and
regional variation (de Onis, et al., 1993). Northeast Brazil, Guatemala, Uganda and

80
Zimbabwe have higher levels of stunted children compared to underweight children with
relatively low levels of wasting. Mali has a higher prevalence of children who are
underweight than children who are stunted combined with relatively high levels of
wasting. These patterns remain constant over time.
There is an increase in wasting for all five countries. This change in wasting is
both interesting and troublesome. Wasting is an indication of present, short term
nutrition. An increase in wasting may indicate present economic hardships or
environmental disasters such as droughts. Increases in this indicator may be an indication
of an overall deterioration of nutritional status. There is a possibility that increases in
wasting may lead to further long term nutritional deficiencies, which may ultimately lead
to increases in the prevalence of stunting and underweight children. The statistical
associations between the indicators indicate that there is no correlation between stunting
and wasting, which is consistent with the findings ofKeller (1983). Increases in levels of
wasting, therefore, may not lead to increases in stunting. There is, however, a moderate
association between wasting and underweight status. This suggests that perhaps the
increase in wasting in these populations may lead to increases in underweight children.
Another possible explanation of the increase in wasting may be the link between
malnutrition and AIDS . As noted earlier, wasting is one of the side effects of AIDS (Ball,
1998). DHS surveys did not examine children for the possibility ofHIV infection.
Although this study is not able to directly link the increase in wasting with an increase in
AIDS, there may be a correlation between the two factors .
Changes in nutritional status over time are not necessarily indicative of overall
improvement (Shen, Habicht & Chang, 1996). Nutritional status of children was therefore

81
examined according to place of residence. Rural children have significantly poorer
nutritional status compared to urban children. This is true both during the 1980's and
1990's in all countries. The prevalence of wasting, increases more in rural areas
compared to urban areas. It appears that the increasing prevalence of wasting in all five
countries is predominantly a rural phenomenon.
The debate between sex differences in the nutritional status of children under five
is longstanding. The results indicate that there is a slight male bias with respect to
nutritional status, these results are similar to other studies (Hill & Upchurch, 1995 ;
Svedberg, 1990). Over time, gender differences are maintained for stunting and
underweight status. Wasting increases slightly more in male children compared to female
children. Anthropometric indicators, however, are not necessarily capable of determining
sex differences (Hill & Upchurch, 1995). The analyses that were performed did not
account for the initial female advantage. Girls are biologically less susceptible to
infectious diseases and nutritional disorders. Since girls have this advantage, boys should
have higher rates of malnutrition. The disparities between sexes are small to insignificant.
The results cannot conclude that there is a gender bias in nutritional status in this sample,
further study is needed for clarification.

Socio-economic Determinants of Growth
The second objective of the data analysis was to examine changes in the socioeconomic determinants of growth over time and to determine if there were common
determinants for the five developing countries. Victora (1992) suggests that there are
possibly different causes for the various indicators, which would explain the variability of

82
these indicators around the world. Logistic regression models were developed for each
indicator to determine if there were differences between stunting and underweight status.
The importance of each predictor was determined by their frequencies in the models and
by the odds ratios in the full logistic regression model.
The results of the multivariate analyses indicate different patterns in the
determinants for stunting compared to underweight models. Frongillo and his associates
(1997) found different causes for stunting and wasting. This study indicates that there are
also differences in causes of stunting and underweight status in children. Further
exploration of the causes of each indicator will illustrate more detailed patterns. This may
be determined by using a larger number of countries in the study and including a greater
number of variables in analysis.
Maternal education is the most prominent and consistent variable for stunting.
After controlling for environmental and health factors, maternal education remains
significant. A mother with primary education, in particular, decreases the likelihood of
her child being stunted. This supports the notion that maternal education is an important
predictor of child health and nutrition (Bicego & Boerma, 1993; Caldwell, 1993; Victora
et al., 1992; Cleland & Ginneken, 1988). Over time, all the countries demonstrate an
increase in the value of the odds ratios. This suggests that maternal education has become
an increasingly important predictor of child nutritional status. A similar relationship is
found with respect to underweight status, except in Northeast Brazil where there is no
statistical association between maternal education and child nutritional status.
Education is a necessary component of development. Haq (1995) specifically
notes that "Education is the true essence ofhuman development. Without education,

83
development can be neither broad-based nor sustained" (UNICEF, 1999). Of the 9
variables analyzed, this study finds maternal education to be the most important
determinant of child stunting, an indicator reflecting social and economic conditions.
Education is important for increasing the transferability ofhealth education. Household
nutrition is promoted through knowledge of nutritional food, food handling procedures
and infant feeding practices. Mata (1988) has labeled this knowledge "maternal
technology." Mothers may be raised without the acquisition of this technology, hindering
child survival and growth. Women who possess effective maternal technologies often
have children who thrive in impoverished areas and poor conditions. The promotion of
maternal technologies is necessary for development.
The education of women is repeatedly shown to not only benefit the health ofher
children, but also to improve family planning choices, decrease violence and abuses
against women and increase empowerment of women. UNICEF (1999) states that of 130
million school age children in the developing world without access to primary education,
two-thirds are girls. The education of girls is one of the most widely stated strategies for
development, cited often by the United Nations and the World Banlc These
organizations, however, fail to provide ideas and strategies to promote education of
females. The recognition of the importance of female education is not sufficient. There
needs to be specific strategies that recognize the social, cultural and economic reasons
why female children are not attending school and to provide practical solutions. For
example, how does one educate the girl who stays at home to take care ofher younger
siblings while her parents are both working, or the girl that stays home to care for sick
family members afflicted with AIDS? There have been some approaches to these issues

84
that have proven effective. Leslie and her associates (1988) cite the following examples:
increasing the number of schools to reduce traveling distance which often alleviates
safety concerns of girls, creating flexible timing of education programs, particularly
during agricultural periods, shortening the periods of class on a daily basis and
introducing small scholarships and/or stipends to increase attendance. Further
examination of these issues and incorporating such strategies in the communities may be
one of the greatest contributions to community health.
Radio ownership was used as a proxy for household economic status. The use of
radio ownership is not a universally accepted method of measuring socio-economic status,
but is appropriate for this particular study. The socio-economic classifications in
developed countries typically use income, occupation or literacy, which are inappropriate
to use in developing countries due to the "structural and economic heterogeneity between
both developing and industrialized countries" (Cortinois, Vella & Ndiku, 1993, p.1087).
Cortinois and his associates ( 1993) constructed a socio-economic index which indicated
that radio ownership is one of the most useful variables in the classification of socioeconomic status. This study demonstrates an association between owning a radio and
nutritional status. Radio ownership is the most consistent variable in the models for
underweight status, both during the 1980's and 1990's. Households owning a radio are less
likely to have an underweight child than households with no radio. It may be that
households with a radio have greater financial and food security. Owning a radio may also
be an important medium for health education, which could promote good nutrition
(Sommerfelt & Stewart, 1994).

85
Univariate and multivariate analyses indicate that place of residence is an
important determinant of growth. Generally, children living in rural areas are more likely
to be undernourished than children in urban areas. In the stunting models, place of
residence is modified in the cases ofNortheast Brazil, Zimbabwe and Guatemala (1987).
Once the environmental variables were controlled, place of residence is no longer
significant. This suggests that in these countries, rural residents are disadvantaged
through poor sanitation and lack of safe drinking water. For Uganda, Mali and Guatemala
(1995) residence remains significant for all three models, indicating there may be other
factors accounting for the rural disadvantage such as poor access to health services.
Maternal work status is a significant determinant of underweight status in three
countries; Northeast Brazil, Guatemala and Uganda. Over time, the working status of the
mother is not consistent for these countries. Work is a significant predictor of
underweight status during the 1990's for Northeast Brazil and Guatemala and during the
1980's for Uganda. In the cases ofNortheast Brazil and Guatemala, there is a greater
likelihood of a child being malnourished if the mother was working, while in Uganda this
relationship was reversed.
Although it is possible to ascertain that there are both intra and inter-country
differences with respect to maternal work status, no conclusive statements can be made.
This is due primarily to the manner in which working status was defined in the study.
Women's employment, as defined by DHS, was whether the woman was working at the
time of the survey and was working for cash wages. This is an extremely narrow
definition of employment. There are many women in the developing world that work
outside of the cash economy and would not have been included in the definition of

86
'working'. In addition, this definition may not include women who have seasonal or
casual employment. Dasgupta (1993) claims that the casual work force in developing is
predominantly women, ranging between 60 and 90 %. Moreover, the DHS definition
does not account for other variables which may affect the nutritional status of the child,
such as the number of hours worked by the woman, whether or not she worked at home
or away from the household, the effect of maternal income or the quality or type of
childcare that was in place (Leslie,1988). DHS took notice of these types of issues and
developed more detailed questions concerning maternal work status in recent surveys.
A second problem with using maternal employment as a predictor of child
survival involves the income side of the employment equation, the positive income effect.
The positive income effect is based on the assumption that acquired resources will benefit
the child's health and well being. Intra-household distribution, however, is dependent on
the person who controls the income. There is a growing body of evidence that suggests
that there are gender differences with respect to the allocation of resources in a household
(Dasgupta, 1993). For example, women tend to allocate more money on feeding children
than men do. An increase in maternal income, therefore, cannot necessarily be interpreted
as an increase in the distribution of food and other items for the well being of her
children.
5.2 Linking Study Results with Health and Development
Health is both an essential objective and a condition for development (WHO,
1998). Nutrition is an essential link to social development (MacPherson, 1990). The
prevalence of undernourished children in a community or nation is one of the most
important indicators of development (UNICEF, 1998). The stagnation of progress of

87
reducing childhood malnutrition shown in four of these countries is an indication of
inadequate development and insufficient health practices during the past decade.
The results of this study illustrates that malnutrition is a complex disorder
associated with various socio-economic, environmental and health factors. These issues
confirm health promotion ideology, which stresses the importance of examining a
multitude of determinants with respect to health status. Strategies should, therefore, focus
on integrating all such aspects.
Health researchers often argue whether strategies to combat malnutrition should
be promoted through improved nutrition or by the reduction of disease. The results of this
study indicate that socio-economic factors are crucial to the nutritional well being of the
child. Non-medical factors may in fact have a greater impact on health status than
medical factors (Beaglehole & Bonita, 1997). Maternal education and radio ownership
(as a proxy for income level), are key determinants of growth even after controlling for
environmental and health conditions. Improvements in malnutrition will be greatest if
strategies are integrated into social and economic development as delineated by PHC.
A crucial component ofPHC, initiated at Alma Ata, was community participation
(Asthana, 1994b). The influx of SPHC strategies, however, created confusion and
ambiguity, and limited the practical use of community participation. Participation implies
"a struggle for the redistribution of power and resources in society." (Asthana, 1994b,
p.185) This is closely linked to the thinking of the Brazilian educator Paolo Freire who
argues that society is based on inequality and governed by oppressors. The manner to rise
above oppression is to "critically recognize its causes, so that through transforming action
they can create a new situation." (Freire, 1970, p.29)

88
The importance of participation is that it serves as a source of empowerment. In
the field of health, empowerment is the notion that people have the capacity to increase
control over their own health by other people in the community (Grace, 1991). A
healthier society can be attained by involving people as a collective group to identify
problems, and to develop strategies through influencing 'personal and social spheres'
(Wallerstein & Bernstein, 1988). Research, particularly in the psychological field,
indicates that gaining greater control and power has a crucial impact on health status. A
focus on community participation will help to empower people to gain greater perceived
control. Such a focus will also help to create suitable local programs and to encourage
people to actively partake in the development process (Ashtana, 1994b).
National governments and the international community should promote and invest
in PHC. With limited finances for health expenditures, national governments should
incorporate broad-based strategies and refrain from high cost medical interventions.
Strategies will vary among nations depending on social and political situations (AbelSmith, 1990). Abel-Smith (1990) compiled a list of failings in the health services of
developing countries, such as a disproportionate amount of money spent for urban health
care, concentrated resources on hospitals, rather than PHC and a lack of spending on
preventative health as opposed to curative health. The researcher stresses that countries
may not be in a position to immediately follow low cost services for the whole
population, but moving in this direction will lead to progressive changes.
The ideals ofPHC sought to not only distribute health resources nationally, but
also internationally. Article II in the Declaration of Alma Ata (1978) includes the
following statement: "The existing gross inequality in the health status of the people

89
particularly between developed and developing countries is politically, socially, and
economically unacceptable and is, therefore, of common concern to all countries."
Foreign aid agencies should refrain from implementing individual, project type work but
rather work jointly with nations to implement wide ranging PHC programs. After
structural adjustment in Mozambique, health care policies became increasingly
dominated by donor and loan agencies (Cliff, 1991, as cited in Buchmann, 1996). The
policies led to a "fragmentation of the primary health care approach" due to the
implementation ofregional "projects". Action should be taken to help share health
resources with developing countries. There should also be action to prevent abuses such
as the selling of overpriced, unnecessary, and often out of date pharmaceuticals or the
marketing of commercial breast milk substitutes. Without international collectivity, and
increases in support of developed countries the strategies ofPHC will be futile.
Progress cannot be made through the health sector alone. There needs to be
changes in the global economy to improve the situation that many governments in the
developing world find themselves in today, namely excessive external debt. The United
Nations estimated that Africa requires approximately 9 billion dollars a year to meet
basic human needs (Buchmann, 1996). This price is less than the money being spent to
repay foreign debts. There are great concerns regarding financial debts voiced by various
leaders in the developing world. Mavis Muyunda, Zambian Minister of State for Foreign
Affairs, states that this debt "translates into a dehumanizing life of poverty, malnutrition,
infant mortality and moral degradation" for people living in developing countries (United
Nations, 1990, as cited in Buchman, 1996). Stamps (1992), attributes the general
decrease in the quality of life of the people of Zimbabwe to "the tendency of those with

90
economic power to move the goal posts with depressing regularity." Haq (1995) rejects
charitable actions and one-way transfers from developed countries and urges for new
partnerships between the North and the South based on justice, equitable sharing and
mutual cooperation.
An expansion of thought is needed in the field of health, in particular with respect
to maternal-child health strategies. Medical and public health researchers concerned with
child survival often view women solely as an 'instrument' to child health (Leslie, 1988).
This view trivializes the roles of women in society. Using the example of maternal
working status, it is clear that this factor cannot be used as a generalized predictor of
child health. Doing so ignores other factors such as the necessity that many low-income
families rely on the employment opportunities of women. The status and roles of women
need to be clearly integrated into the socio-economic views of health promotion.
Dasgupta (1993) acknowledges the importance ofhealth in development, but
cautions solutions based predominantly on the implementation of public health measures.
In particular, infants and young children have caught the attention of international
organizations and national governments, such as the 'child survival and development
revolution' (CSDR) by UNICEF. The narrow focus of the international health community
on early child survival and the implementation of specific medical techniques, such as
oral rehydration therapy, fail to address underlying causes of ill health, namely poverty
and inequality. The health of the child is inextricably bound to the mother and other
members of the family. The child will not fare well if the economic providers and
nurturers of the family are ill. Sometimes families must make decisions that have a
negative impact on child health and development for long term needs of the family,

91
especially in times of crisis (Thompson, 1996). For example, during distress migration, as
seen in Rwanda in 1994, children may be abandoned or placed in orphanages in an
attempt to secure a better chance of survival. Furthermore targeting child survival reduces
the right to health for other members of society. Researchers in the field of health must be
prepared to step back and take a wider lens and to examine development in the entire
social constructs instead of using narrow and simplistic (even if it is with good intentions)
approaches.

5.3 Policy Recommendations
There are two dominant views in developing nutritional policies (Foster, 1992).
One type of policy aims to treat the symptoms of malnutrition, while the other type of
policy aims to treat causes. Treating the symptoms involves various nutrition
interventions ranging from international food aid, to food-rationing schemes to vitamin
supplementation programs. Policies that aim to treat symptoms may be effective at
reducing malnutrition, but are more often than not, unsustainable. Treating the causes of
malnutrition involves devoting attention to a broad range of social, economic,
environmental and health factors. This may help to decrease malnutrition in the long run,
but effects may not be reached as quickly as treating the symptoms. Due to limited
resources, it may come to a choice between policies. The dilemma, thus, is do we attempt
to reduce the immediate suffering of hunger or to alleviate malnutrition over time? The
results of this study in conjunction with the emerging field of health promotion indicate
that our goal should be for long term, sustained levels of nutrition obtained by addressing
the root causes of malnutrition.

92

The results of this study, in collaboration with other research (Frongillo et al.,
1997; Victora, 1992), suggest that nutritional indicators vary with respect to country and
region, and there are different socio-economic causes of each indicator. These have
implications for nutritional policies, which can be applied to a community, region or
nation (Frongillo et. al, 1997). These recommendations include: (1) identifying which
nutritional problems are prominent (stunting, wasting, underweight or a combination of
these); (2) identifying the factors that are causing the deficits in nutritional status; (3)
determining if these factors can be changed and if these changes will lead to
improvements in nutritional status. MacPherson (1990) stresses that nutritional
surveillance, particularly at the national level, is essential in organizing strategies to
combat poverty and to reduce social inequality. Longitudinal studies are also important to
observe the factors influencing changes in nutritional status and to assess the functioning
of different strategies and policies.

5.4 Recommendations for Future Research
This study illustrates that there are differences in nutritional patterns across
countries and over time. Future research should continue to examine global trends in
child malnutrition, specifically to examine the potential increase in wasting worldwide.
The relationship between wasting and other indicators needs to be investigated to
determine if increases in wasting in a region leads to increases in other anthropometric
indicators.
Maternal work status should be studied in a more in depth manner that allows an
exploration of specific elements such as time, social and cultural factors. This type of
study requires more detailed information, which often does not appear in international

93
health surveys and may require different methodologies, such as qualitative research and
case studies. If researchers choose to implement maternal work status as a predictor of
child survival, they should be aware of definitional problems and should refrain from
making conclusive or causal statements.
Logistic regression is the analysis of choice among public health researchers, due
mainly to the simplicity of the procedure and easy interpretation of the odds ratios for
determining risk factors. The results of this study are limited by problems attributed to
inadequate logistic regression model fits, specifically poor classification of cases.
Researchers typically do not report the adequacy of the logistic regression models which
raises the question ofthe appropriateness of models used in analysis. The use ofthe chisquare as a measure of goodness-of-fit has also been scrutinized and has various
disadvantages (Hosmer, Hosmer, Cessie & Lemeshow, 1997). The extensive use of
logistic regression should be questioned in the health field. There is a multitude of
multivariate techniques available. As the fields of public and community health expand,
researchers should explore new statistical techniques.

5.5 Conclusion
This study has demonstrated that, in spite of the goals of the World Summit for
Children, there has not been adequate progress in the reduction of childhood malnutrition
in 4 of the 5 countries selected in this study. The continuation of widespread malnutrition
is a major constraint in the development process. Various economic and political setbacks
have led to the implementation of targeted strategies, increased reliance on medical
interventions and project type solutions. Malnutrition constitutes a complex array of
socio-economic factors, environmental and health determinants. It is therefore imperative

94
to use comprehensive strategies for the improvement of nutritional status. Such strategies
will require national and international cooperation to engage in social change. The
essence of international cooperation for social progress and improvements in child health
is captured by Stamps (1992, p. 36) "We have to abolish the things which divide us and
promote mutual support on the basis of a common destiny ... Only uplifting the future
prospects for all our children can promote a harmoniously progressive healthy world."

95

References
Abel-Smith, B. (1990). The economics ofhealth care. In T.A. Lambo & S.B. Day
(Eds.), Issues in contemporary international health. New York: Plenum Press.
Asthana, S. (1994a). Economic crisis, adjustment and the impact on health. In
D.R. Philips & Y. Verhasselt (Eds.), Health and development. London: Routledge.
Asthana, S. (1994b). Primary health care and selective PHC. Community
participation in health and development. In D.R. Philips & Y. Verhasselt (Eds.), Health
and development. London: Routledge.
Ball, C.S. (1998). Global issues in pediatric nutrition: AIDS. [On-line]. Nutrition,
14 (10), 767-770. Abstract from : WebSPIRS File: HealthSTAR Item: 0899-9007.
Press.

Basch, P. (1990). Textbook of international health. New York: Oxford University

Beaglehole, R. & Bonita, R. (1997). Public health at the crossroads. Cambridge,
UK: Cambridge University Press.
Becker, M.H. (1986). The tyranny ofhealth promotion. Public Health Reviews,
1±. 15-25.
Bicego, G.T. & Boerma, J.T. (1993). Maternal education and child survival: A
comparative study of survey data from 17 countries. Social Science & Medicine,
36, (9), 1207-1227.
Black, R.E. (1991 ). Would control of childhood infectious diseases reduce
malnutrition? Acta Paediatr Scan Suppl. 374, 133-140.
Boerma, J.T. & Sommerfelt, A.E. (1993). Demographic and health surveys
(DHS): Contributions and limitations. World Health Statistics Quarterly Report, 46, 222226.
Bouvier, P., Papart, JP., Wanner, P., Picquet, M., Rougemont, A. (1995).
Malnutrition of children in Sikasso (Mali): Prevalence and socio-economic determinants.
[On-line] . Soz Praventivmed, 40 (1): 27-34. Abstract from: FirstSearch File: Medline
Item: 0303-8408
Buchman, C. (1996). The debt crisis, structural adjustment and women's
education: Implications for status and social development. International Journal of
Comparative Sociology, 375-30.

96
Caldwell, J.C. (1993). Health transition: The cultural, social and behavioural
determinants ofhealth in the third world. Social Science and Medicine. 36 (2), 125-135.
Caulfield, L.E., Bentley, M.E. & Ahmed, S. (1996). Is prolonged breastfeeding
associated with malnutrition? Evidence from nineteen demographic and health surveys.
American Journal ofEpidemiology.25 (4), 693-703.
Chatterjee, S. & Yilmaz, M. (1992). A review of regression diagnostics for
behavioral research. Applied Psychological Measurement. 16 (3), 209-227.
Cortinovis, 1., Vella, V. & Ndiku, J. (1993). Construction of a socio-economic
index to facilitate analysis of health data in developing countries. Social Science and
Medicine. 36 (8), 1087-1097.
Dasgupta, P. (1993). An inquiry into well-being and destitution. New York:
Oxford University Press.
Demographic and Health Surveys (1998). DHS Homepage [On-line]. Available:
http :www .macroint.com/dhs/.
de Onis, M, Monteiro, C., Akre, J. & Clugston, G. (1993). The worldwide
magnitude of protein-energy malnutrition: An overview from the WHO Global Database
on Child Growth. Bulletin of the World Health Organization. 71 (6), 703-712.
Dibley, M.J. , Staehling, N., Nieburg, P. & Trowbridge, F.L. (1987). Interpretation
of Z-score anthropometric indicators derived from the international growth reference.
American Journal of Clinical Nutrition. 46, 749-762.
Esrey, S.A. (1996). Water, waste, and well-being: A multicountry study.
American Journal ofEpidemiology. 143 (6), 608-23.
Esteva, G. (1992). Development. In Sach & Wolfgang (Eds.) The development
dictionary. A guide to knowledge as power. London and Atlantic Highlands: N.J. Zed
Books.
Forsberg, B.C., Van Ginneken, J.K. & Nagelkerke, J.D. (1993). Cross-sectional
household surveys of diarrhoeal disease-A comparison of data from the control of
diarrhoeal disease and Demographic and Health Surveys Programmes. International
Journal ofEpidemiology. 22 (6), 1137-1145.
Foster, P. (1992). The world food problem. Tackling the causes of undernutrition
in the Third World. Boulder, Colorado: Lynne Rienner Publishers Inc.
Fosu, G.B. (1994). Childhood morbididty and health services utilization: Crossnational comparisons of user-related factors from DHS data. Social Science and
Medicine. 38(9), 1209-1220.

97

Freire, P. (1970).Pedagogy of the oppressor. New York: The Continuum
Publishing Company.
Frongillo Jr., E.A. de Onis, M. & Hanson, M.P. (1997). Socioeconomic and
demographic factors are associated with worldwide patterns of stunting and wasting of
children. Journal ofNutrition. 127, 2302-2309.
Gale Research Staff ( 1998). Countries of the world and their leaders yearbook
1999. B. Rajewski (Ed.) Cleveland, Ohio: Eastwood Publications Development Inc.
Gove, S. (1997). Integrated management of childhood illness by outpatient health
workers: Technical basis and overview. Bulletin of the World Health Organization. 75
(Supplement 1), 7-24.
Grace, V .M. (1991 ). The marketing of empowerment and the construction of the
health consumer: A critique of health promotion. International Journal of Health
Services. 21 (2), 329-343.
Gorstein, J. & Akre, J (1988). The use of anthropometry to assess nutritional
status. World Health Statistics Quarterly. 41, 48-57.
Habitcht, J., Martorell, R. Yarbrough, C. Malina, R.M. & Klein, R.E. (1974).
Height and weight standards for preschool children: How relevant are ethnic differences
in growth potential? Lancet. 1 (858), 611-615.
Press.

Haq, M. (1995). Reflection ofhuman development. New York: Oxford University

Herrell, I.C. & Mulholland, C.A. (1998). Reflections on health in development
and human rights. World Health Statistics Quarterly Report, 51 (1), 88-92.
Hill, K. & Upchurch, D.M. (1995). Gender differences in child health: Evidence
from the Demographic and Health Surveys. Population and Development Review. 21 (1),
127-151.
Hosmer, D.W., Hosmer, T., Le Cessie, S. & Lemeshow, S. (1997). A comparison
of goodness-of-fit tests for the logistic regression models. Statistics in Medicine. 16. 965980.
Hosmer, D.W., Taber, S. & Lemeshow, S. (1991). The importance of assessing
the fit oflogistic regression models: a case study. American Journal of Public Health. 81
(12), 1630-1635.
Huffman, S.L. & Martin, L. (1994). Child nutrition, birth spacing and child
mortality. Annals ofthe New York Academy of Sciences. 709,236-248.

98

Kickbusch, I. (1996). Health promotion: A global perspective. In Health
promotion: An anthology (pp.14-22).Washington, DC: Pan American Health
Organization.
King, L.A. (1998). Economic growth and basic human needs. International
Studies Quarterly. 42, 385-400.
Klassen, S. (1996). Nutrition, health and mortality in Sub-Saharan Africa: Is there
a gender bias? The Journal of Development Studies. 32 (6), 913-932.
Lamontagen, J.F., Engle, P.L. & Zeitlin, M.F. (1997). Maternal employment,
child care, and nutritional status of 12-18-month-old children in Managua, Nicaragua.
Social Science and Medicine. 46 (3), 403-414.
Le, T. & Verma, V.K. (1997). An analysis of sample designs and sampling errors
of the Demographic and Health Surveys. DHS Analytical Reports No.3. Calverton,
Maryland: Macro International Inc.
Lerer, L. B., Lopez, A. D., Kjellstrom, T. & Yach, D. (1998). Health for all:
analyzing health status and determinants. World Health Statistics Quarterly Report 51 (1)
7-19.
Leslie, J. (1988). Women's work and child nutrition in the third world. World
Development. 16 (11), 1341-1362.
Leslie, J, Lycette, M. & Buvinic, M. (1988). Weathering economic crises: The
crucial role of women in health. In D. Bell & M. Reich (Eds.) Health. nutrition. and
economic crises (pp.307-348). Dover, Massachusetts: Auburn House Publishing
Company.
Macdonal, G. & Burton, R. (1992). Health promtotion, Discipline or discilplines?
In R. Bunton & G. Macdonald (Eds.), Health promotion. Disciplines and diversity.
London: Library of Congress.
MacPherson, S. (1990). Social policy and nutrition surveillance. Social Policy and
Administration. 24 (3), 254-257.
Madzingira, N. (1995). Malnutrition in children under five in Zimbabwe: Effect
of socioeconomic factors and disease. Social Biology. 42 (3-4), 239-46.
Mata, L. (1988). A public health approach to the "food-malnutrition-economic
recession" complex. In D. Bell & M. Reich (Eds.) Health. nutrition, and economic crises
(pp.265-278). Dover, Massachusetts: Auburn House Publishing Company.

99
McMurray, C. (1996). Cross-sectional anthropometry: what can it tell us about the
health ofyoung children? Health Transition Review, 6, 147-168.
Merchant, K.M. & Kurz, K.M. (1993). Women's nutrition through the life cycle:
Social and biological realities. In M. Koblinsky, J. Timyan & J. Gay, The health of
women. A global perspective, (pp.63-90). Boulder, Colorado: Westview Press Inc.
Midgley, J. (1997). Social welfare in global context. London: Sage Publications.
Midgley, J. (1995). Social development: The developmental perspective in social
welfare. Thousand Oaks, CA: Sage.
Pebley, A.R. & Goldman, N. (1995). Social inequality and children's growth in
Guatamala. Health Transition Review, 5, 1-20.
Philips, D.R. & Verhasselt, Y. (1994). Introduction: Health and development. In
D.R. Philips & Y.Verhasselt (Eds.), Health and development. London: Routledge.
Price, P. (1994). Maternal and child health strategies. In D.R. Philips & Y.
Verhasselt (Eds.), Health and development. London: Routledge.
Rifkin, S.B. & Walt, G. (1986). Why health improves: Defining the issues
concerning 'comprehensive primary health care' and 'selective primary health care'. Social
Science and Medicine, 23 (6), 107-114.
Rothman, K.J. & Greenland, S. (1998). Modem epidemiology (second edition)
Philadelphia: Lippincott-Raven Publishers.
Seers, D. (1969). The meaning of development. In D. Lehmann (Ed.)
Development theory: Four critical studies. London: Frank Cass.
Shen, T. , Habicht, J. & Chang, Y.C. (1996). Effect of economic reforms on child
growth in urban and rural areas of China. The New England Journal ofMedicine. 335 ,
400-406.
Sommerfelt, A.E. & Stewart, M.K. (1994). Children's Nutritional Status.
Demographic and Health Surveys Comparative Studies No.12.
36.

Stamps, T.J. (1993). Who really cares for Africa? World Health Forum, 14, 34-

Svedberg, P. (1990). Undernutrition in sub-Saharan Africa: Is there a gender bias?
The Journal ofDevelopment Studies, 26, 469-486.
Svedberg, P. (1996). Gender biases in Sub-Saharan Africa: Reply and further
evidence. The Journal ofDevelopment Studies, 32 (6), 933-943 .

100

Tabachnick, B.G. & Fidell, L.S. (1997). Using multivariate statistics (3rd ed.).
New York: HarperCollins Publishers Inc.
Thompson, B. (1996). Coping with food crises in Africa. FAO Technical Report.
52-9377.
UNICEF (1994). The Progress ofNations 1994. (Issue BriefNo.92-806-3129-7)
New York: Adamson.
UNICEF (1995). The State ofthe World's Children 1995. New York: Oxford
university Press.
UNICEF (1998). The State ofthe World's Children 1998. New York: Oxford
university Press.
UNICEF (1999). The State ofthe World's Children 1999. New York: Oxford
university Press.
Ulijasek, S.J. (1990). Nutritional status and susceptibility to infectious disease. In
G.A. Harrison & J.C. Waterlow (Eds.), Society for the study of human biology
Syiill)OSum 30. Diet and disease in traditional and developing societies. Cambridge:
Cambridge University Press.
United Nations Development Programme. (1997). Human Development Report
1997. New York: Oxford University Press
United Nations (September, 1990}. World Declaration on the Survival, Protection
and Development of Children and Plan of Action on the Survival, Protection and
Development of Children in the 1990's. United Nations: New York.
Vella, V., Tomkins, A., Borghesi, A., Migliori, G.B. & Oryem, V.Y. (1994).
Determinants of stunting and recovery from stunting in Northwest Uganda. International
Journal ofEpidemiology. 23 (4), 782-786.
Victora, C.G. (1992). The association between wasting and stunting: An
international perspective. Journal ofNutrition. 122, 1105-1110.
Victora, C.G., Huttly, S.R., Fuchs, S.C. & Olinto, M.T. (1997). The role of
conceptual frameworks in epidemiological analysis: A hierarchial approach. International
Journal of Epidemiology. 26 (1), 224-227.
Victora, C.G., Huttly, S.R., Barros, F., Lombardi, C. & Vaughan, J.P. (1992).
Maternal education in relation to early and late child health outcomes: Findings from a
Brazilian study. Social Science and Medicine. 34 (8), 899-905.

101
Wallerstein, N. & Bernstein, E. (1988). Empowerment education: Freire's ideas
adapted to health education. Health Education Quarterly. 15 (4), 379-394.
Waterlow, J.C, Buzina, R. , Keller, W., Lane, J.M., Nichaman, M.Z. & Tanner,
J.M. (1977). The presentation and use ofheight and weight data for comparing the
nutritional status of groups of children under the age of 10 years. Bulletin of the World
Health Organization. 55 (4), 489-498.
Waterlow, J.C. (1992). Protein-energy malnutrition. London: Edward, Arnold.
World Bank (1993). World Development Report 1993. Investing in Health.
Washington DC: Oxford University Press.
World Health Organization (1946). World Health Organization Constitution.
Geneva: WHO.
World Health Organization (1978). Primary Health Care. Geneva: World Health
Organization.
World Health Organization (1998). The World Health Report 1998. Life in the
21st Century a Vision for All. Geneva: World Health Organization.

102

Appendix A: Description of DHS Surveys
Date of
Fieldwork

Executing Agency

1986

BEMFAM *

DHS-III

1996

BEMFAM

2298

Guatemala

DHS-1

1987

IN CAP**

2207

Guatemala

DHS-III

1995

Instituto Nacional de Estadistica

4911

Mali

DHS-1

1987

948

Mali

DHS-III

1996

Centre d'Etudes et de Recherches
sur Ia Population pour le
Developpement
CPS/MSSPA, DNSI ***

4263

Uganda

DHS-1

1988/89

Ministy of Health

2297

Uganda

DHS-III

1995

3350

Zimbabwe

DHS-1

1988/89

Zimbabwe

DHS-III

1994

Ministry of Finance and Economic
Planning
Ministry of Finance and Economic
Planning
Central Statistical Office

Country

Survey

Brazil (NE)

DHS-1

Brazil (NE)

Sample
Size
644

1496
1907

*Sociedade Civil Bem-Estar Familiar no Brasil
**Institute de Nutricion de Centro America y Panama
***Centre pour la Population et la Sante, Direction National de la Statistique et de
!'Information

103

Appendix B: Logistic Regression Results
Brazil:
Stun ~
Residence
Education
pnmary
secondary
Work
Radio
Sex
Sanitation

latrine
flush

Water Source
well
piped
Immunization
other
measles
Diarrhea

Year

Coefficient

86
96

p-value

Odds Ratio

.1796
-.3137

Standard
Error
.2967
.2548

.5548
.2183

1.1968
.7308

86
96
86
96

1.1927
2.4814
.6334
1.8366

.6124
.4166
.5864
.3553

.0515
.0000
.2801
.0000

3.2958
11.9586
1.8841
6.2751

86
96
86
96
86
96

-.2027
.5498
.0812
.1273
.4864
.9136

.2274
.2380
.2095
.2282
.2060
.2224

.3726
.0209
.6982
.5769
.0182
.0000

.8165
1.7329
1.0846
1.1358
1.6264
2.4934

86
96
86
96

.8621
.5280
-3.5874
1.2403

.2683
.4550
1.0314
.4508

.0013
.2458
.0005
.0059

2.3681
1.6955
.0277
3.4568

86
96
86
96

-.1471
1.8215
-.0004
1.9364

.3334
.3879
.3584
.2943

.6590
.0000
.9991
.0000

.8632
6.1814
.9996
6.9336

86
96
86
96

.3617
1.100
-.6277
-1 .3493

.3312
.3367
.2240
.3108

.2747
.0011
.0051
.0000

1.4358
3.0042
.5338
.2594

86
96

-.0125
-.4666

.7023
.6887

.9537
.0501

.9875
.6271

;'-

104

Guatemala:
Stun ~
Residence
Education
pnmary
secondary
Work
Radio
Sex
Sanitation

latrine
flush

Water Source
well
piped
Immunization
other
measles
Diarrhea

Year

Coefficient

p-value

Odds Ratio

.0626
-.2400

Standard
Error
.1280
.0888

87
95

.6249
.0069

1.0646
.7866

87
95
87
95

1.6530
2.0752
1.0662
1.7962

.2390
.1832
.2390
.1790

.0000
.0000
.0000
.0000

5.2226
7.9664
2.9043
6.0269

87
95
87
95
87
95

.0128
.0166
.2939
.1090
.0698
.1301

.1434
.0793
.0989
.0695
.0927
.0616

.9290
.8346
.0030
.1167
.4515
.0348

1.0129
1.0167
1.3416
1.1151
1.0723
1.1389

87
95
87
95

.9637
.8322
.9074
.8561

.1807
.1333
.1621
.1158

.0000
.0000
.0000
.0000

2.6213
2.2985
2.4778
2.3540

87
95
87
95

.1700
.6566
.0805
.4250

.1960
.5595
.1771
.5560

.3857
.2406
.6496
.4447

1.1853
1.9283
1.0838
1.5295

87
95
87
95

-.6631
-.4435
-.9268
-.9167

.1279
.0986
.1071
.0729

.0000
.0000
.0000
.0000

.5153
.6418
.3958
.3998

87
95

-4.0065
-.0342

8.2066
.0705

.2383
.6277

.6254
.9664

105

Mali:
Stunting
Residence
Education
Primary
Secondary
Work
Radio
Sex
Sanitation

Latrine
Flush

Water Source
Well
Piped
Immunization
Other
Measles
Diarrhea

Year

Coefficient

87
96

p-va/ue

Odds Ratio

.0663
-.4882

Standard
Error
.2680
.0880

.8047
.0000

1.0685
.6137

87
96
87
96

-2.1876
2.0632
-2.3482
1.7891

.6344
.3709
.6456
.3828

.0006
.0000
.0003
.0000

.1122
7.8710
.0955
5.9838

87
96
87
96
87
96

4.5665
.0953
.8020
.1369
.0272
.0725

22.2431
.0698
.2150
.0737
.1916
.0692

.8373
.1718
.0002
.0634
.8873
.2946

96.2057
1.1000
2.2300
1.1467
1.0275
1.0752

87
96
87
96

4.6310
4.6224
4.1180
4.4579

9.0239
4.0421
9.0212
4.0415

.6078
.2528
.6480
.2700

102.6197
101.7367
61.4364
86.3082

87
96
87
96

.3277
-.0826
-.0205
.1599

.6178
.2677
.5975
.1904

.5958
.7576
.9726
.4011

1.3878
.9207
.9797
1.1733

87
96
87
96

.7472
-.4125
-.9938
-.9083

.3430
.0848
.3698
.0884

.0294
.0000
.0072
.0000

2.1111
.6620
.3702
.4032

87
96

-.5567
-.4991

.1933
.0741

.0040
.0000

.5731
.6071

106

Uganda:
Stun ~
Residence
Education
pnmary
secondary
Work
Radio
Sex
Sanitation

latrine
flush

Water Source
well
piped
Immunization
other
measles
Diarrhea

Year

Coefficient

88
95

p-value

Odds Ratio

-.4678
-.5580

Standard
Error
.1592
.1061

.0033
.0000

.6264
.5723

88
95
88
95

.4775
1.0511
.3799
.8551

.2073
.1584
.1967
.1434

.0213
.0000
.0534
.0000

1.6120
2.8607
1.4622
2.3516

88
95
88
95
88
95

-.0528
-.0354
.2830
.1834
.1929
.2847

.1702
.0814
.1064
.0828
.0882
.0771

.7565
.6637
.0078
.0268
.0287
.0002

.9486
.9652
1.3270
1.2012
1.2128
1.3294

88
95
88
95

.1058
-.4974
-1.7913
-.5217

.1255
.4200
.4936
.4090

.3990
.2364
.0003
.2021

1.1116
.6081
.1667
.5935

88
95
88
95

.6415
-.5379
.3513
-.7210

.6932
.2909
.6925
.2888

.3548
.0644
.6119
.0125

1.8992
.5840
1.4209
.4863

88
95
88
95

-.4373
-.4429
-.3646
-.6303

.1138
.1095
.1068
.0887

.0001
.0001
.0006
.0000

.6458
.6422
.6944
.5324

88
95

-.2088
-.3292

.0935
.0844

.0255
.0001

.8115
.7195

107
Zimbabwe:
Stun tin
Residence

Education
pnmary
secondary
Work
Radio
Sex
Sanitation

latrine
flush

Water Source
well
piped
Immunization
other
measles
Diarrhea

Year

Coefficient

88
94

p-value

Odds Ratio

-.6527
-.4229

Standard
Error
.4642
.3284

.1597
.1977

.5206
.6551

88
94
88
94

.9905
.8629
.7269
.8561

.2649
.2199
.2390
.1684

.0002
.0001
.0024
.0000

2.6925
2.3699
2.0686
2.3539

88
94
88
94
88
94

.0622
.1903
.3887
-.1144
.2540
.1119

.1391
.1287
.1488
.1372
.1287
.1282

.6551
.1392
.0090
.4046
.0484
.3828

1.0641
1.2096
1.4751
.8919
1.2892
1.1184

88
94
88
94

1.2562
.9743
.8521
.9819

.4649
.5239
.4617
.5193

.0069
.0629
.0650
.0586

3.5119
2.6493
2.3446
2.6696

88
94
88
94

-.5573
.0604
-1.2207
-.2193

.7634
.4907
.8592
.4651

.4653
.9020
.1554
.6373

.4653
1.0623
.1554
.8031

88
94
88
94

.2730
.2221
-.6881
-9.2222

.2749
.2843
.1456
7.4532

.3206
.4346
.0000
.2160

.3206
1.2487
.0000
.0001

88
94

-.4442
-.1848

.1376
.1435

.0012
.1978

.0012
.8312

108

Brazil:

Year

Coefficient

p-value

Odds Ratio

.5288
1.5101

Standard
Error
.7477
.6860

86
96

.4794
.0277

1.6969
4.5272

86
96
86
96

7.5450
7.8706
5.7935
8.3295

25.1824
24.8192
25.1824
24.8128

.7645
.7512
.8180
.7371

1891.193
2619.1335
328.1715
4144.2325

86
96
86
96
86
96

-.5611
2.2238
-.1423
2.9648
-.7544
1.8574

.4734
1.1193
.4829
.7422
.4763
.7157

.2359
.0469
.7683
.0001
.1132
.0095

.5706
9.2427
.8674
19.3902
.4703
6.4072

86
96
86
96

.8576
.7352
-6.9156
-.3572

.6554
1.0223
20.6614
1.0620

.1907
.4720
.7378
.7366

2.3574
2.0859
.0010
.6997

86
96
86
96

-.8062
-7.6197
-.3914
2.2573

.7822
58.6430
.8548
.7762

.3027
.8966
.6470
.0036

.4466
.0005
.6761
9.5570

86
96
86
96

2.2608
2.0095
-1.0952
-.9305

.5248
.8110
.6540
.8481

.0000
.0132
.0940
.2726

9.5904
7.4597
.3345
.3944

86
96

-.9618
-2.2705

.4632
.6683

.0379
.0007

.3822
.1033

~

Residence
Education
pnmary
secondary
Work
Radio
Sex
Sanitation

latrine
flush

Water Source
well
piped
Immunization
other
measles
Diarrhea

109

Guatemala:
Underweight
Residence
Education
pnmary
secondary
Work
Radio
Sex
Sanitation

latrine
flush

Water Source
well
piped
Immunization
other
measles
Diarrhea

Year

Coefficient

p-value

Odds Ratio

.1908
-.5747

Standard
Error
.1378
.1059

87
95

.1661
.0000

1.2102
.5629

87
95
87
95

1.5713
4.0896
.9460
3.7620

.3968
.7121
.3953
.7111

.0001
.0000
.0167
.0000

4.8131
59.7160
2.5755
43.0326

87
95
87
95
87
95

-.1185
-.1581
.2321
.1681
-.0835
.0680

.1569
.0852
.1022
.0718
.0988
.0658

.4498
.0634
.0232
.0192
.3984
.0192

.8882
.8537
1.2612
1.1830
.9199
1.0703

87
95
87
95

2.1545
.8772
2.0107
.7998

.2949
.1709
.2835
.1580

.0000
.0000
.0000
.0000

8.6232
2.4042
7.4683
2.2251

87
95
87
95

.0017
.2388
-.1239
.3226

.2065
.6777
.1906
.6748

.9933
.7246
.5157
.6326

1.0017
1.2698
.8835
1.3807

87
95
87
95

-.3444
-.3686
-.9030
-.3593

.1288
.1041
.1223
.0736

.0075
.0004
.0000
.0000

.7087
.6917
.4054
.4251

87
95

3.2557
-.3593

13.5025
.0736

.8095
.0000

25 .9372
.6982

110

Mali:

Year

Coefficient

p-value

Odds Ratio

-.1065
-.2017

Standard
Error
.2180
.0781

87
96

.6253
.0098

.8990
.8173

87
96
87
96

-.8746
.7684
-.9837
.4179

.5529
.1811
.5569
.1989

.1137
.0000
.0773
.0357

.4170
2.1563
.3739
1.5188

87
96
87
96
87
96

5.3328
.1307
.4196
.2466
.0802
-.0713

22.2415
.0645
.1781
.0692
.1550
.0640

.8105
.0427
.0185
.0004
.6049
.2657

207.0122
1.1397
1.5213
1.2796
1.0835
.9312

87
96
87
96

5.1971
.2017
4.6760
-.1059

9.6270
.4432
9.6250
.4370

.5893
.6491
.6271
.8085

180.7392
1.2234
107.3349
.8995

87
96
87
96

.1916
-.0016
-.0227
.1579

.4181
.2338
.4017
.1535

.6468
.9945
.9550
.3037

1.2112
.9984
.9776
1.1710

87
96
87
96

-.1865
-.4678
-.6947
-.4933

.2639
.0823
.2444
.0769

.4797
.0000
.0045
.0000

.8298
.6264
.4992
.6106

87
96

-.8438
-.5954

.1570
.0704

.0000
.0000

.4301
.5514

~

Residence
Education
pnmary
secondary
Work
Radio
Sex
Sanitation

latrine
flush

Water Source
well
piped
Immunization
other
measles
Diarrhea

111

Uganda:

Year

Coefficient

88
95
88
95
88
95

p-value

Odds Ratio

-1.0935
-.7546

Standard
Error
.2935
.1364

.0002
.0000

.3351
.4702

1.1134
1.3253
.5240
1.1008

.3482
.2178
.3429
.2061

.0014
.0000
.1265
.0000

3.0447
3.7634
1.6888
3.0066

~

Residence
Education
Primary
secondary

Radio
Sex
Latrine
flush
Water Source
Well
piped
Immunization
Other
measles
Diarrhea

.

·..

·"·

<'

>•

;

..

··

Work

Sanitation

'i

88
95
88
95
88
95

-.6872
-.0072
.6441
.3859
.2087
.3838

.2209
.0924
.1552
.0957
.1133
.0880

.0019
.9380
.0000
.0001
.0653
.0000

.5030
.9928
1.9042
1.4709
1.2321
1.4679

88
95
88
95

.0334
-1.1350
-1.0814
-1.4550

.1487
.4517
1.0395
.4410

.8223
.0120
.2982
.0010

1.0340
.3214
.3391
.2334

88
95
88
95

.5044
1.5532
.0311
1.4132

1.0823
.6103
1.0829
.6099

.6412
.0109
.9771
.0205

1.6561
4.7265
1.0316
4.1090

88
95
88
95

-.4558
-.1907
-.5971
-.3136

.1405
.1217
.1396
.1002

.0012
.1170
.0000
.0018

.6340
.8264
.5504
.7308

88
95

-.6879
-.6397

.1164
.0922

.0000
.0000

.5026
.5275

:'.)

"

·•····.

··.··•

.

••••

.

...

'·•

(

..·

>

.
i

•
'

....
······

'· .
.

112
Zimbabwe:
Underweight
Residence
Education
pnmary
secondary
Work
Radio
Sex
Sanitation

latrine
flush

Water Source
well
piped
Immunization
other
measles
Diarrhea

Year

Coefficient

88
94

p-value

Odds Ratio

3.6380
.7806

Standard
Error
1.0445
.4455

.0005
.0797

38.0148
2.1827

88
94
88
94

10.2387
2.2232
7.8387
1.4733

33.3700
.3613
33.3702
.3292

.7590
.0000
.8143
.0000

27964.268
9.2367
2536.9779
4.3637

88
94
88
94
88
94

.5689
.2798
2.3216
.4353
-.1867
.8854

.3651
.1793
.6996
.2049
.3095
.1855

.1191
.1187
.0009
.0336
.5464
.0000

2536.268
1.3229
10.1920
1.5455
.8297
2.4240

88
94
88
94

10.5266
5.1250
9.8851
4.4999

23 .7896
1.0118
23.7875
1.0009

.6581
.0000
.6777
.0000

37294.937
168.1806
19635.864
90.0114

88
94
88
94

9.8839
-2.2110
.9705
-1.5467

221.3835
.6737
223.1783
.6308

.9644
.0010
.9965
.0142

19612.704
.1096
2.6393
.2129

88
94
88
94

.0847
.3594
-2.4016
-4.4763

.5740
.3658
.5593
1.0041

.8827
.3285
.0000
.0000

1.0884
1.4325
.0906
.0114

88
94

-1.6019
-1.0510

.3272
.1833

.0000
.0000

.2015
.3496