Malnutrition and Mortality among Bolivian Children:
An Analysis of DHS Data
by
Timothy William Miller
B.A. (Princeton University) 1985
M.A. (University of California at Berkeley) 1991
DISSERTATION
submitted in partial satisfaction of the
requirements for the degree of
DOCTOR of PHILOSOPHY
in
DEMOGRAPHY
in the
GRADUATE DIVISION
of the
UNIVERSITY of CALIFORNIA at BERKELEY
Committee in charge:
Professor Ronald D. Lee, Chair
Professor Kenneth W. Wachter
Professor Sylvia R. Guendelman
1993
Abstract
This dissertation examines malnutrition and mortality among Bolivian children in the late 1980s and changes in their mortality risk before this period. Data from the Bolivian Demographic and Health Survey form the basis of this analysis. Between February and July of 1989, a total of 7,923 women were interviewed and physical measurements (height and weight) were taken on 2,682 children between the ages of 3 and 36 months. These interviews and measurements are analyzed in three chapters that examine different aspects of child health. The first examines changes in mortality rates during the last 30 years; the second examines high-altitude hypoxia, genetic inheritance, malnutrition, and child height; and the third examines neonatal and postneonatal mortality examined by cause of death.
In the first analytical chapter, application of the Brass and Feeney indirect estimation techniques to the DHS data and other surveys shows that child mortality rates have been falling quite rapidly over the last 15 years. Examination of macro-level data reveals that several commonly cited explanations of mortality decline -- urbanization, increases in GDP per capita, improvements in nutrition, expansion of primary health care and fertility changes -- are implausible explanations of the Bolivian experience. Based on the available evidence, the most likely explanation for this rapid mortality decline is that the expansion of the educational system resulted in a shift in the social composition of births toward women with higher levels of education who experience lower rates of child mortality.
Elevation data were added to the DHS data set to allow simultaneous analysis of the impact of high-altitude hypoxia, genetic inheritance, and malnutrition on child height in the second analytical chapter. Multivariate regression analysis of height Z-scores demonstrated that the slow growth of Bolivian children was principally a reflection of chronic malnutrition rather than genetic adaptation or hypoxic stress from high altitudes. However, while not being the principal cause of slow growth, the effect of high-altitude hypoxia did appear to be large enough to warrant reconsideration of the use of a single international child growth standard. In Bolivia, use of this growth standard would result in the erroneous conclusion that malnutrition is concentrated in the Altiplano region. In fact, malnutrition is widespread throughout all of Bolivia.
The final chapter analyzed neonatal and post-neonatal mortality using a proximate determinants framework. Maternal education was found to have a strong impact on both neonatal and postneonatal mortality independent of husband's education, husband's occupation, and family and geographic characteristics. It appeared that education acted through four key proximate factors to explain differences in neonatal mortality: residence in homes without sanitation facilities, lack of tetanus immunizations, lack of medical assistance at delivery, and low birth weight. However, closer examination of the impact of these bio-medical factors on specific causes of death indicated that most were proxying for some unmeasured behaviors. Educational differences in postneonatal mortality were not explained by differences in the proximate determinants as measured in the DHS. Examination of cause of death revealed that educational differentials only existed for mortality from diarrhea and not from respiratory and non-infectious causes. This proximate determinants analysis was unable to determine the bio-medical linkage between maternal education and child survival -- a causal explanation of this relationship remains elusive.
TABLE OF CONTENTS
Chapter One: INTRODUCTION 1
Demographers and the health transition 1
Child malnutrition and mortality in Bolivia 3
Chapter Two: COUNTRY BACKGROUND AND DATA SOURCE 5
Bolivian geography, economy, social institutions, culture, and demography 5
Demography data set used in this dissertation 17
Chapter Three: THE HEALTH OF BOLIVIAN CHILDREN: 1960-1990 21
Introduction 21
Measurement of health: 1960-1990 22
Causes of the mortality decline 40
Conclusion 58
Chapter Four: HIGH-ALTITUDE HYPOXIA, GENETIC ADAPTATION,
MALNUTRITION, AND CHILD STUNTING
Introduction 60
Conclusion 93
Chapter Five: MATERNAL EDUCATION AND CHILD MORTALITY 97
Introduction 97
Analysis of neonatal and postneonatal mortality 110
Analysis of child mortality by cause of death 149
Conclusion 162
Chapter Six: CONCLUSION 164
Summary of thesis results and conclusions 164
Future research directions 167
BIBLIOGRAPHY 169
APPENDICES 178
Appendix 3.1: Explanation of Brass and Feeney estimation techniques 178
Appendix 5.1: Two-stage estimation results 182
Appendix 5.2: Estimation of interaction effects 188
LIST OF FIGURES
Figure 2.1. Map of Bolivia 6
Figure 2.2. Age pyramid for Quechua speakers, 1988 13
Figure 2.3. Age pyramid for monolingual Quechua speakers, 1988 14
Figure 2.4. Age pyramid for bilingual Quechua speakers, 1988 15
Figure 3.1. Percentage of children short for age (stunted), low weight
for age (underweight), and low weight for height (wasted) 24
Figure 3.2. Causes of visits to health clinic in 1970 among children under 5 27
Figure 3.3. Illnesses ever experienced by children under 5 in 1983 28
Figure 3.4. Measles and Whooping Cough cases: 1965-1991 30
Figure 3.5. Causes of death 32
Figure 3.6. Infant mortality rates: 1960-1990 34
Figure 3.7. Child mortality rate: 1960-1990 35
Figure 3.8. Infant mortality rate by maternal education, 1976 and 1988 39
Figure 3.9. GDP and GDP per capita, 1949-1988 43
Figure 3.10. Hospitals and health centers/posts 46
Figure 3.11. Doctors, trained nurses, and auxillary nurses in the public sector 47
Figure 3.12. Total fertility rate, 1965-1990 51
Figure 3.13. Years of education for women over 5 in 1950, 1976, and 1988 54
Figure 4.1. Simulated z-scores 68
Figure 4.2. Simulated z-scores with height measurement error 68
Figure 4.3. Simulated z-scores with age measurement error 68
Figure 4.4. Age distribution of 2,612 children 71
Figure 4.5. Language spoken at home by family: DHS 1989 74
Figure 4.6. Language spoken: ENPV 1988 75
Figure 4.7. Mean z-scores of children by characteristics of the child 77
Figure 4.8. Mean z-scores of children by characteristics of the mother 78
Figure 4.9. Mean z-scores of children by place of residence 79
Figure 4.10. Coefficients from linear regression predicting z-scores 82
Figure 4.11. Mean z-scores by altitude 91
Figure 4.12. Z-scores by altitude and educational groups 92
Figure 4.13. Percent of children classified as stunted, before and after
adjusting for hypoxia effect 96
Figure 4.14. Malnourished children by region: using unadjusted z-scores 97
Figure 4.15. Malnourished children by region: using altitude-adjusted z-scores 98
Figure 5.1. Relative risk of death during infancy by maternal education and region 99
Figure 5.2. Relative risks of mortality by maternal education,
trimean of 11 countries 100
Figure 5.3. Proximate determinants framework 103
Figure 5.4. Four types of equations to be estimated 109
Figure 5.5. Deaths by age 111
Figure 5.6. nQx values 113
Figure 5.7. Gross and net effects of maternal education on neonatal mortality 122
Figure 5.8. Gross and net effects of maternal education on postneonatal mortality 123
Figure 5.9. Effect of maternal education on neonatal mortality 140
Figure 5.10. Effect of maternal education on postneonatal mortality 141
Figure 5.11. Pathways by which maternal education affects neonatal mortality 146
Figure 5.12. Pathways by which maternal education affects postneonatal mortality 148
Figure 5.13. Postneonatal mortality rates by cause and maternal education 161
LIST OF TABLES
Table 3.1. Differentials in infant mortality by area, region, and education
of the mother, 1976 and 1988. 37
Table 3.2. Decomposition of change in mortality rates, 1976 and 1988 56
Table 4.1. Studies of child growth in the Andes 63
Table 4.2. Number of children: measured and unmeasured 70
Table 4.3. Linear regression predicting z-scores 81
Table 4.4. Example of creating quechua-ness variable 85
Table 4.5. Unadjusted and adjusted number of Quechua and Aymara 86
Table 4.6. Ratio of error variance to X variance 87
Table 4.7. Uncorrected and corrected parameter estimates 88
Table 4.8. Linear regressions predicting z-scores 89
Table 5.1. Description of background variables 114
Table 5.2. Description of proximate determinants 115
Table 5.3. Reduced-form estimation results for neonatal period 124
Table 5.4. Reduced-form estimation results for postneonatal period 127
Table 5.5. Proximate determinants estimation results for neonatal period 131
Table 5.6. Proximate determinants estimation results for postneonatal period 132
Table 5.7. Hybrid estimation results for neonatal period 136
Table 5.8. Hybrid estimation results for postneonatal period 138
Table 5.9. Effects of maternal education on the proximate determinants 143
Table 5.10. Impact of maternal education on neonatal mortality 144
Table 5.11. Impact of maternal education on postneonatal mortality 145
Table 5.12. Multinomial estimation of neonatal mortality 153
Table 5.13. Multinomial estimation of postneonatal mortality 156