|
African Population Studies
Union for African Population Studies
ISSN: 0850-5780
Vol. 11, Num. 1, 1996, pp. 29-62
|
African Population Studies/Etude de la Population Africaine, Vol. 11, October/octobre 1996
PREVENTIVE HEALTH STRATEGIES AND INFANT SURVIVAL IN ZIMBABWE
Assata ZERAI*
Code Number: ep96002
ABSTRACT
Socioeconomic and demographic variables are examined in a multilevel framework
to determine conditions influencing infant survival in Zimbabwe (1983-88) using
Demographic and Health Survey data. Community level child health inputs consistently
influence infant survival. The most unique finding is that childbearing-aged
women's average educational levels in their community exert a greater effect
on infant survival than the individual mother's educational level. This result
supports assertions that child survival is strongly impacted by mass education
(Caldwell, 1989). This research also contributes evidence to theories postulating
that widespread social development is necessary to sustain recent declines
in developing country mortality rates (Palloni, 1989).
RÉSUMÉ
Les variables socio-économiques et démographiques sont étudiées dans un
cadre à niveaux multiples pour déterminer les conditions qui influencent
la survie des enfants à bas âge au Zimbabwé (1983-88) en utilisant les données
d'une enquête démographique et de santé. Les apports communautaires à la
santé de l'enfant exercent une influence régulière sur la survie des enfants âgés
de moins de cinq ans. Le résultat le plus remarquable est que le niveau moyen
d'instruction des femmes en âge de procréation au sein de leur communauté a
un plus grand effet sur la survie des enfants de moins de cinq ans que le
niveau d'instruction de la mère prise isolément. Ce résultat conforte les
thèses selon lesquelles l'éducation populaire a un impact décisif sur la
survie de l'enfant (Calwell, 1989). Cette recherche fournit également des
preuves supplémentaires aux théories fondées sur le postulat qu'un développement
social généralisé est nécessaire pour que les baisses relevées recemment
au niveau des taux de mortalité dans les pays en développement soient durables
(Palloni, 1989).
Infant mortality levels are still high in less developed countries today.
The major causes of infant mortality are preventable and are tied to decreasing
levels of household contamination, increasing nutritional intake and improving
preventive and curative health practice (Preston, 1980). Inputs affecting infant
health can take place on a community level, and additionally community-level
infant health inputs can sometimes exert an even greater influence on infant
survival outcomes than individual-level characteristics.
I examine demographic and socioeconomic variables measured at the individual
level as well as community conditions influencing infant survival in Zimbabwe.
The theoretical model used suggests that the socioeconomic status of the household
and community conditions are mediated by proximate determinants, including
maternal fertility, personal disease control, nutrient availability, and household
contamination (Mosley and Chen, 1984).
There is not an extensive literature on community level effects on child survival
from which to chart theoretical direction for this research (notable exceptions
include: Bailey 1988; Mosley and Chen 1984). But there is a developing literature
on multilevel analysis (Entwisle 1989; Billy and Moore 1992; Tsui 1985), and
an extensive literature supportive the positive effect of the individual mother's
education on child survival (Caldwell 1986; Caldwell 1989; Das Gupta 1990;
Cleland and van Ginneken 1988; Cleland 1990). In a survey of 99 countries Caldwell
(1986:179) has found that average levels of female education is the main predictor
of child survival at the societal level. "There is... as close a correlation
between child survival and general levels of education in a community as there
is between child survival and maternal education" (Caldwell 1989:103).
A multilevel analysis which incorporates both the individual mother's level
of education with the average levels of education of women in the community
is necessary to get a better understanding of the relative importance of education
at the individual and community level for promoting child health.
Theoretically, this research is motivated by the basic premise that economic
development and unidimensional health interventions such as immunization programs
are not enough to ensure constant mortality decline in low-income countries;
widespread social development is necessary to sustain a consistent mortality
decline (Palloni 1985). There have been not only slowdowns in the apce of mortality
decline in African and Latin American countries, but also reversals in infant
and child mortality trends in such countries as Zambia (Gaisie, Cross and Nsemukila
1993:81) and Nigeria (Federal Office of Statistics 1992:78) which are witnessing
recent increases in child mortality. While medical technology and economic
development are important, programs that ensure the distribution of social
development resources throughout the country to improve the living standards
of families are necessary to sustain declines in infant mortality. Infant health
is chosen as the dependent variable because infants are the most vulnerable
population to the changes in economic and social development (Hill and Palloni
1992).
Further theoretical development is needed on many fronts in order to advance
our understanding of determinants of child health. The main analytical difficulty
has been in specifying the mechanisms of multilevel models that delineate the
pathways by which community characteristics influence individual behavior (Tienda
1991) and specifying the mechanisms through which maternal education affects
child survival (Cleland and van Ginneken 1988). Even less attention has been
paid to delineating the mechanisms through which the education of women in
the community impacts child health.
Unfortunately, this project does not solve any of those dilemmas. But it contributes
to the state of knowledge on the use of multilevel analysis and on the effect
of maternal education on child survival by incorporating both level of education
of the infant's mother and the level of education in the infant's community
into the same model to see which exerts a greater effect on infant survival.
While the mechanisms tying community and individual behavior to child survival
are still largely unknown, the average levels of education of child bearing
aged women in the community turns out to be a type of proxy for development
at the community level. Further research which incorporates participant observation
or ethnographic methodology to delineate social networks and spheres of influence
within and between networks in a community is necessary to determine the actual
mechanisms.
ZIMBABWE SOCIAL SETTING
Zimbabwe is an appropriate social setting in which to look at effects of maternal
education on child health because it is a newly independent country in which
citizens are exercising a greater sense of political and personal efficacy.
The transition in control of education and health institutions from the hands
of colonial settlers to a nascent government elected by the previously dispossessed
majority has allowed policymakers to institute progressive health strategies
motivated to benefit the majority populace.
The Zimbabwe government believes that development is not just economic in
nature. It is also social, political, and cultural. When assessing their country's
development, the people of Zimbabwe chose leadership committed to addressing
the relevant questions regarding poverty, unemployment, and inequality. Their
concern for social development grew out of their struggle for liberation from
colonialism as an leadership is that market oriented development does not give
preference to the needs of the masses. So the Zimbabwe government chose not
to follow a strategy that measured progress solely by export orientated economic
growth. Instead, they chose to couple economic schemes with social development
strategies, and to give priority to the social betterment of the majority of
its citizens. This strategy reflects the underlying belief that equitable distribution
of resources is an important part of development.
Zimbabwe is an interesting context in which to study child health because
of the substantial improvements in indicators of living conditions since political
independence (attained April 18, 1980). The country's 1960 infant mortality
rate was 113 deaths per 1000 live births; according to the ZDHS, it dropped
to 72/1000 in 1988 and the Population Reference Bureau has projected its 1994
infant mortality rate at 59/1000 (United Nations 1989; ZDHS 1989; Population
Reference Bureau 1994). Zimbabwe's Equity in Health program reflects the government's
commitment to social development in many ways. Zimbabwe has encompassed some
of the most progressive strategies for improving child health. Some of these
include integrating traditional health practitioners into the health care infrastructure,
balancing curative and preventive health measures, and a focus on redistributing
health services so that the poorest citizens in greatest need have improved
access to health services.
The result of the abandonment of the racially discriminatory and inequitable
colonial practice in controlling health care access has been overwhelming.
There is majority access for Zimbabwe's population to health centers and hospitals
in even the remotest areas. During the first 10 years of independence, rural
areas were targeted for the more than 500 health centers that were built or
upgraded, and more than a dozen district hospitals were completed or are under
constructions (World Bank 1992:1). Additionally, between 1980 and 1985, more
than 4000 Village Health Workers who disseminate health information and provide
preventive and simple curative services in the rural areas were trained. More
than 200 State Certified Nurses were trained to staff clinics (World Bank 1992:1).
The result has been that "millions of people have for the first time gained
access to basic care" (World Bank 1992:ix). "In 1980 (at the time
that Zimbabwe attained its independence), 44% of the publicly funding services
went to the urban-based sophisticated central hospitals serving about 15% of
the population" (Sanders and Davies 1988:724). But as of 1988, "in
Zimbabwe, nearly health of all rural women (were) within 5 kilometers of some
type of health provider, and only 12% are more than 15 km distant from one" (Wilkinson,
Njogu, and Abderrahim 1993:24). The median distance to a health center for
married women ages 15 to 49 interviewed in the nationally representative Demographic
and Heath Survey sample is 3.1 kilometers: 5,1 km in rural areas and 1 km in
urban areas (Wilkinson, Njogu, and Abderrahim 1993:26). Immunization data show
no inequality in distribution between rural and urban areas of Zimbabwe (Zimbabwe
Central Statistical Office 1989)1, although there is some inequality
by province (Zerai 1992). Additionally, health education has promoted low-risk
maternal fertility practices such as delayed child bearing, parities of four
or fewer children, and adequate spacing of births (Zimbabwe Federal Office
of Statistics 1989).
MULTILEVEL APPROACH
The multilevel approach guards against the myopia of micro level analyses
which "implicitly assume that variations in (behavior associated with
the dependent variable) are due only to individual characteristics and that
the social contexts in which people live have invariant effects..." on
the outcome variable (Billy and Moore 1992:978). The advantage of the multilevel
approach is that the researcher simultaneously estimates the individual-level
effects on the outcome variable that macro level analyses omit. "A multilevel
approach bridges the gap between the strictly macro and strictly micro orientations" (Billy
and Moore 1992:978).
While the multilevel frameworks have been widely praised for incorporating
both micro and macro level models (Billy and Moore 1992; Tsui 1985; Brewster,
Billy, and Grady 1990; Entwisle, Casterline, and Sayed 1989; etc), the approach
has been criticized for mechanically connecting the individual and contextual
models without specifying the mechanisms through which macro forces exert an
effect on micro level behavior and outcomes (Tienda, 1991).
When putting together multilevel models and infant health, the average level
of female education is an excellent independent variable to focus on because
considerable effort has been expanded in an attempt to identify mechanisms
linking maternal education to child survival (Caldwell 1986; Caldwell 1989;
Cleland and van Ginneken 1988; Mosley and Chen 1984). The problem of specifying
the mechanisms organically linking macro and micro infuences is exemplified
by the problem of specifying the mechanisms through which education affects
child health. Education is clearly relevant to the society in general even
though it is measured at the individual level. Formal education prepares the
young to fulfill the expectations of their society. Its function is largely
socialization. Unfortunately, although several mechanisms, through which maternal
education affects child health, have been proposed in the literature, we still
lack a definitive knowledge of these mechanisms (Caldwell 1989:106).
Importance of Maternal Education
A review of the literature shows that while the higher socioeconomic status
of better educated women explains about half of the magnitude of the relationship
between maternal education and child survival (Cleland and van Ginneken 1988),
the domestic health practice of individual women is probably the new most salient
mechanism in the maternal education-child mortality relationship². The fact
that mother's education is a more important determinant of child survival than
father's education shows that there is greater maternal involvement in child-health
related care (Caldwell 1989). The mother's education influences her choices
and skills in health care practices (Bailey 1988; Das Gupta 1990; Caldwell
1989). For instance, both educated and illiterate mothers recognize when their
child is sick but the educated mother more frequently will take action "without
waiting for (her) husband or mother-in-law to notice the child's condition
too" (Caldwell 1989:106). "This is partly because illiterate women
do feel a lack of capability when dealing with the modern world" (Caldwell
1989:106). Caldwell found that the educated mother is "more likely to
report back to the health center if the treatment does not seem to be affecting
a cure. Educated women see the (health process) as experimental... (And do
not feel it is an attack on the health care parctitioner to give this important
feedback)" (Caldwell 1989:106). Joshi has postulated that it is through
the acquisition of skills and identity that education impacts the health behavior
of women, but says, "while these findings are interesting, they are still
incomplete. More studies, especially longitudinal ones, are needed before these
findings can be woven into a meaningful theory" (Joshi 1994:24).
Education as a Community Phenomenon
We know even less about the reason average levels of female education in the community exert
a positive effect on infant survival. But this is an important relationship
to consider. Caldwell points that "an uneducated woman may feel more deprived
in a country where most other women are educated than one in which they are
not: nevertheless, her children stand a much greater chance of survival" (Caldwell
1989:103). The mechanisms operating in the mass education-child mortality relationship
are likely to be muldimensional. One possibility is health service advocacy.
The equitable distribution of community health services influences child health
in Kerala State India (Preston 1978). Advocacy to ensure better distribution
is imperative. Bailey (1988) found that the shortage of health facilities and
personnel affect child health in Sierra Leone. "As argued powerfully by
Caldwell (1986) the key to low mortality at the societal level may be a synergy
between mass education and egalitarian politics which leads to demands for
a health service that caters for the needs of all" (Cleland and van Ginneken
1988).
Another important possibility is the improvements in status that are accrued
to all women as a result of mass education. This impacts how women are viewed
in society and also how women view themselves. "The quality of schooling
seems relatively unimportant...; it is not so much what you learn or understand,
but how you see yourself and how others see you" (Caldwell 1989:106). "Women
who had been to school thought of the school as part of the whole modern system
which included independence, five-year plans, government programs, health centers,
modern medicine and themselves" (Caldwell 1989:106). So mass education
has an important influence on both the individual and the society.
Another result of mass education that may have important implications for
child survival is its compounding multigenerational effects (Caldwell 1989).
Wealth flows reverse as there are larger investments by families in the development
of their individual children (Caldwell 1982).
FRAMEWORK
The general theoretical question addressed examines the relationship between
development and population. The specific question asks to what extent social
development determines infant mortality. There is a debate on the causes of
mortality decline and the position argued here is that mortality decline is
sustained in less developed countries as a result of economic development,
humanistic social development, and equitable distribution of development resources
(Palloni 1985; Caldwell 1986).
Distribution of development resources to all in society ensures that the standard
of living is adequate, women and men have the freedom to choose their contribution
to society, and that children's health received a high priority. Having clean
water supplies, proper sanitation, educational facilities, and hospitals, promotes
the well-being of citizens which finally has positive effects on child health.
When universal education is encouraged and high percentages of women complete
schooling, women have a more egalitarian role in society (Caldwell 1986) that
allows them to delay childbearing and encourages the lower parities conducive
to improved child health (Chen 1983). When men and women exercise their right
to choose their education and employment, living standards rise as a result
of increased household economic resources. This produces living environments
with reduced contamination that promote child health. Development resources
such as piped water and adequate sanitation promote health because they also
reduce contamination. Finally, when child health is a high priority in the
public sector, as reflected in the building of hospitals and health centers
that include prenatal care, immunization, and growth-monitoring programs, children's
survival chances improve.
The theoretical framework of child health in Zimbabwe here begins with some
of the individual-level variables in the classical proximate determinants framework
(Mosley and Chen 1984). Other crucial variables are added that operate through
the proximate determinants, such as maternal education, parents' occupation,
and community-level variable (see figure 1). The community conditions are mediated
by proximate determinants of child health. Proximate determinants operate through
specific mechanisms to determine child health outcomes.
The causal ordering is that macroeconomic, political, and social factors set
the stage for community conditions and socioeconomic realities in which families
find themselves. Community conditions and socioeconomic variables operate through
the proximate determinants of child health. Community conditions also exert
an impact on socioeconomic conditions to influence proximate determinants of
child health indirectly.
Presently, maternal fertility variables are typified by later marriage
and childbearing, lower parities, and longer birth intervals than in preceding
periods in Zimbabwe. Operating through the biological mechanism of improved
production of breastmilk and through the social mechanism of less competition
with siblings for mother's attention and household resources, children get
better nutrition and better care, thus improving their survival chances (Chen
1983).
In Zimbabwe, household contamination is still a big problem. Piped
water is provided to a minority of households. Sanitation measures are still
not adequate. But improvements in these areas since independence work through
the mechanism of less exposure of children to contamination to make them less
susceptible to disease and eventually lower mortality.
Nutrient availability has improved
due to drought feeding programs (Agere 1986), land reform and increased economic
opportunities (Sanders and Davies 1988). Operating through the mechanism
of decreased susceptibility to illness, improved nutrient availability leads
to improvements in child health (Preston 1980).
Personal disease control is typified
by better immunization coverage and increased access to treatment. The goals
set by the Ministry of Health of universal immunization of children and targeting
rural areas for building primary health care centers (Manga 1988) have improved
the timeliness of disease prevention and treatment. Timely personal disease
control is critical for improving child survival chances.
Mother's higher levels of education and increased professional and
blue collar employment of the household head lead to low child-health-risk
fertility, timely immunization of children, adequate nutritional intake by
children, and household environments with lessened contaminants. These socioeconomic
variables operate through the proximate determinants to influence child health.
Community conditions have become
more conducive to child health in Zimbabwe today than in the past. Some characteristics
include free and compulsory education (Zvobgo 1986), widespread family planning
education, land reform measures, free health services to citizens who do
not make minimum wage (Sanders and Davies 1988), prenatal care programs (Agere
1986), and postnatal care including growth monitoring. These conditions make
it easier for individual households and mothers to maximize the health of
their children.
Community conditions operate through socioeconomic variables and proximate
determinants to influence child health. For instance, higher percentages of
women receiving primary level and more education lead to improved socioeconomic
outlooks of families, which in turn influence fertility variables, personal
disease control, nutrient availability, and household contamination.
MODELS
The Zimbabwe Demographic and Health Survey (ZDHS) and ZDHS Service Availability
data are used to analyze the effect of individual and community-level determinant
on child mortality between 1983 and 1988 in Zimbabwe. To incorporate individual
level and contextual determinants in this study of infant health in Zimbabwe,
a three-tiered analysis is conducted by establishing a baseline model that
includes individual-level determinants and background characteristics and then
analyzing both a community model that includes distal determinants and finally
an expanded model that incorporates all the variables in the first two models
(see figure 2).
First, the individual level risk factors to neonatal and postneonatal survival
are evaluated for children born during 1983-88 in Zimbabwe. Background variables
are added as controls. The maternal histories collected in the Zimbabwe Demographic
and Health Surveys provide adequate data to examine effects of environmental
contamination and maternal variables on individual-level child survival3.
However, to understand child health status better, it is important to understand
the context in which child health outcomes unfold. This analysis reveals that
individual level risk factors are expectedly not the whole picture, but mediate
contextual determinants.
In the second part of the analysis, the effect of resource distributive programs
(as reflected in contextual variables) are examined. These programs were established
after Zimbabwe attained independence. The expanded model examines the joint
effects of the individual-level determinants and community conditions on neonatal
and postneonatal mortality.
DATA
The Demographic and Health Surveys (DHS) are a rich source of data on developing
countries in general, and Africa in particular. The national probability surveys
interview all or ever-married women ages 15-49 and provide a wealth of information
on child health, the proximate determinants of fertility, fertility preferences,
and other social and economic characteristics unmatched by comparative data
from developing countries to date.4 Retrospective data is collected
to provide complete birth histories, as well as more detailed information on
the five years preceding the survey (1983-88).
The sample size in the Zimbabwe Demographic and Health Survey is 4201 women.
It was fielded between June and September 1988. To prepare the data for analysis,
files with mothers, children, and finally child-months as three different units
of observation from the ZDHS data set were constructed. While the ZDHS data
give very specific information on 1983-88 they only give limited birth history
and child mortality information for previous years. Therefore, the empirical
analyses here are restricted to the 3393 children born during this short period.
Variables analyzed for all models are listed in Table 1. Descriptive statistics
for all variables are listed in Table 2. The analysis of child survival is
limited to infants under 1 year where a majority of child mortality usually
occurs. That infant mortality rates are sensitive to social development inputs
permits an assessment of the impact of proximate and distal determinants on
child health in Zimbabwe. Different sets of variables influence neonatal and
postneonatal mortality, requiring separate investigations for the latter. Analysis
of neonatal and postneonatal survival allows examinations of two important
aspects of early child health in Zimbabwe; neonatal mortality reflects "the
preexisting health conditions of the mother and the health care she and her
infant received during pregnancy and shortly after delivery" and postneonatal
mortality reflects "post-birth socioeconomic, environmental, and medical
care" indicators (Campbell 1989).
Proximate Determinants
The study used individual-level variables whose relationship to infant survival
has been well established (Chen 1983; Mosley and Chen 1984; Pebley and Millman
1986; Curtis, Diamons, and McDOnald 1993; Phuapradit et al 1990). Mother's
age at childbearing, parity, and the preceding birth interval are examined
in the form of a high risk fertility index. If any of these high risk conditions
existed at the time of the birth, a pregnancy was deemed to be at high risk
of complications, negative birth outcomes such as neonatal mortality, or problems
with postneonatal survival. If none of the conditions existed, the index child
was designated a low risk birth category5. Half (51%) of children
in the subsample were born in conditions of high risk to pregnancy, birth,
or postbirth complications. The high-risk maternal and child health index was
constructed to address the problem of collinearity among maternal fertility
variables and also in preparation for further bivariate analysis which will
treat risk as endogenous to the relationship between background and community
variables affecting infant survival.
Other variables from the classical proximate determinants model such as nutrient
availability, personal disease control, and incidence of injury were not examined
because of insufficient information on the variable itself (in the case of
incidence of injury for all infants and immunization or other indicators of
personal disease control for infants that died), and inadequate information
that would allow causal inference (such as nutrient availability and succeeding
birth intervals).
Background Characteristics
Socioeconomic variables determine the availability of nutritional resources,
which is especially important because once infants reach the age of 6 months,
they can no longer depend on nourishment from breastmilk alone. Mother's education
is important because it facilitates her integration into a society impacted
by traditional customs, colonialism, and neo-colonialism. Education heightens
her ability to make use of government and private health care resources and
it may also increase the autonomy necessary to advocate for her child in the
household and the outer world. If a mother has not completed primary level
education, her child is more likely to die.
The other socioeconomic variable measured is parents' occupation. If parents
do not have regular employment, they will be less able to supply nutritional
needs of children and consequently their children will be more susceptible
to disease and other causes of death.
Household contamination is an important indicator of child susceptibility
to contagion leading to disease. The higher the value of the household through
lack of easy access to water, modern toilet facilities, and refrigeration and
having nonporous floor material. A factor score was constructed for these variables
to address statistical problems presented by the high collinearity among them.
The number of people in the household is added as a measure of crowing in the
household. Finally rural/urban residence of the infant is included in the model
as a control variable.
Distal Determinants
The ZDHS-Service Availability (ZDHS-SA) data set was collected on 166 at 167
sample clusters6 in Zimbabwe. ZDHS-SA variables were matched to
the individual-level and child-month working files by cluster. Two of the variables
for examining the effect of community-level variables on infant survival between
1983 and 1988 are extracted from the ZDHS-SA. General categories of variables
measuring community-level inputs into child health include existence of family
planning education and contraceptive distribution programs, female education,
and health care availability7.
Clusters are the primary sampling unit used in the ZDHS. Families living in
the same cluster live in close proximity to each other -- in the same community.
In the sample of 3393 children, an average of 20 survey households is represented
by each cluster (with a standard deviation of 11). This ranges from 1 to 84
of the households in which women were interviewed. The total number of households
in a cluster is less than 500 for a majority (72%) of the communities and no
more than 2000 for all communities (Wilkinson, Njogu, and Abderrahim 1993).
The first community-level variable of interest is the average level of women's
education. A variable indicating the mean level of education for childbearing
aged women was created by calculating educational levels of all women by cluster
from the original ZDHS file. Caldwell (1986:191) points out that the percentage
of women in the community with primary schooling is an important determinant
of child health because it indicates female autonomy and the extent of egalitarianism
in the society. The more empowered childbearing-aged women are as a group,
the more likely individual women in that community will be able to promote
actively the health of their children. Children living in communities where
women on the average complete primary and secondary level education are expected
to have higher probalities of survival than those living in communities in
which women on the average do not complete primary-level education.
The second community-level variable is whether or not the locale has family
planning education and contraceptive distribution programs, which is important
because it is hypothesized to increase the likelihood that parents contracept
-- at least to space their births, to the benefit of mother and child's health.
Children who live in a community in which family planning education and contraceptive
distribution programs are available, are more likely to be healthy.
The availability of community health care facilities indicates the extent
of health services, including prenatal and postnatal care, immunization, and
growth monitoring. Hospital and clinic availability is mainly important for
personal disease control. If resources for disease prevention, such as prenatal
care, immunization, and growth monitoring are available in the community, the
child is more likely to be healthy. If resources for disease diagnosis and
treatment, such as hospitals, health clinics, and health workers are available,
the child is more likely to survive.
In Zimbabwe, most hospitals and health clinics have prenatal care, immunization,
and growth monitoring. Availability of hospitals and health centers in the
community is also an indicator of social development. The greater the development
infrastructure, the greater the access to hospitals and other health resources.
METHODS
The logistic regression and Cox proportional hazards models are used to estimate
the effect of proximate and distal covariates on child mortality8.
Neonatal mortality models are estimated with logistic regression. Postneonatal
mortality models are estimated with the proportional hazards model. The Cox
proportional hazards model allows researchers to study events, or the change
from one state to the other at a specific point in time (Allison 1984). For
the first part of this analysis the transition from the state of being alive
to the state of death is examined.
RESULTS
The model selection results appear in Table 3. The overall fit for all neonatal
and postneonatal mortality models is good. The chi-square statistic is significant
at the .001 level for proximate, community and saturated models. Chi-square
tests result in significant improvements of the expanded models over the proximate
and community models. This shows that individual and community-level variables
contribute significantly to our understanding of infant mortality in Zimbabwe.
Tables 4 and 5 list the results of the logistic regression and Cox regression
analyses of neonatal and postneonatal mortality, respectively. Odds ratios
are presented for all logit models, and hazard ratios are presented for Cox
regression models. Estimated hazard and odds ratios are relative risks --for
a one-unit change in the independent variable the instantaneous risk of death
increases by a multiplicative factor of the reported value of the hazard ratio
(Computing Resource Center 1992:212). The first column in Tables 4 and 5 list
the independent variables for all three models. The second and third columns
list the results of the individual-level model, the fourth and fifth columns
show the results of the community model, and the sixth and seventh columns
report the results of the saturated model.
Determinants of Neonatal Mortality
All children at risk of dying before the first completed month of life are
included. Of the 3393 children in the sample, 108 infants did not survive to
age 1 month (a neonatal mortality rate of 32.1 deaths per 1000 live births).
Proximate and background variables that determine whether a child survives
to age 1 month are listed in table 4.
The high-risk maternal fertility index is the first significant determinant
of neonatal mortality (& = .05). If the child's birth order is greater
than 4, if the child's preceding birth interval is 24 months or greater, or
if the mother was not the optimal age at birth, as indicated by a value of
1 on the high-risk index, the child is 1.42 times as likely to die as infants
who were born under low risk conditions. A low risk birth is one in which the
child has higher chances of survival because there has been a sufficiently
lengthy time interval after the previous birth (or because the index birth
is a first order birth). The mother is at her biologically and socially optimal
age for childbearing, and because of lessened competition for maternal resources
due to lower parities.
The only significant household contamination variable is the number of persons
residing in the household. For each additional member in the household, children
are only 89% as likely to die in the first month of life as children in smaller
households (& = .001). Whereas, originally theory predicted that the number
of people in the household could be used as a proxy for crowding and increased
susceptibility to disease through contagion (Aaby 1988), it appears the number
of household members may be a better proxy for the number of caretakers available
to children under 5 years of age. As shown in Table 6, the main compositional
difference between large and small households is that larger households contain
greater numbers of adults. This finding supports the idea that these additional
members contribute to the pool of caretakers or free the primary caretaker
to spend more time with the newborn.
Personal disease control is measured by whether or not the pregnant mother
was cared for by a nurse or doctor prior to the birth of her child. Newborns
whose mothers received prenatal care from a health professional are 35% less
likely to die as those whose parents had not obtained prenatal care from a
nurse or doctor (& = .001).
Neonates whose parents are unemployed are 2.79 times as likely to die during
the first month of life as children whose parents are working in professional
occupations. This variable indicates socioeconomic resources of the household.
Households that can afford to provide proper nutrition, personal disease control,
and highly sanitary living environments are more likely to have healthy children
residing in them.
The last variable in the individual-level model is rural/urban residence.
Newborns in urban areas are approximately half as likely to die as newborns
living in rural areas during the first month of life (& +.05). Conditions
are still more favorable to child health in urban areas of Zimbabwe.
Logistic regression analysis of the effects of community-level variables on
neonatal mortality are reported in columns 4 and 5 of Table 49.
The average educational level of women in the community significantly influences
survival of newborns. Infants living in communities where women have a secondary
level education are .33 times as likely to die and those living in communities
where women have on the average completed a primary level education are .49
times as likely to die the first month of life, as infants living in communities
where the average woman is uneducated. Both results are highly statistically
significant (& = .01).
The only shared variable in the proximate model and community model is rural/urban
residence. Similar to the residence variable in the individual model, urban
neonates are .62 times as likely to die as rural neonates.
Model 3 encompasses the individual level variables in model 1 and community-level
variables of model 2 (see columns 6 and 7 of table 4). The magnitude and direction
of effects are the same as in the results for the constrained individual and
community models. The significance is virtually the same as well.
Determinants of Postneonatal Survival
Determinants of infant survival to age 1 year are examined (see Table 5) using
Cox regression10. Of the original 3393 children observed, 108 were
deleted because they died in the first month of life, and 62 observations were
deleted because they were censored11. Of the 3223 children at risk,
68 did not survive to 12 months of age (a postneonatal mortality rate of 23.2
deaths per 1000 live births). The independent variables are the same as in
the analysis of neonatal mortality.
The significant household influence on infant mortality was the number of
members in the household. For each additional member in the household, children
were only 86% as likely to die before reaching age one as children living in
households with fewer members (& = .001). The more household members there
were, the more likely the infant to survive. The magnitude of the effect is
similar to that in the neonatal mortality model.
Infants whose mothers obtained professional prenatal care were .30 times as
likely to die in Zimbabwe as infants whose mothers had not obtained such care.
This statistic is significant at the & = 0.001 level.
Family-level socioeconomic variables are important for infant survival. Mother's
education (at least primary level) and parents' employment status significantly
influence survival of infants. Infants whose mothers had a secondary level
education were .49 times as likely to die (& = .001) and infants whose
mothers completed primary level education were only .84 times as likely to
die, as infants whose mothers did not complete a primary level education (& =
.05). While mother's individual-level education was not significant in neonatal
mortality models, it was significant in infant mortality models. In individual-level
postneonatal model, if neither parent has a regular occupation, the infants
were 1.74 times as likely to die as children whose parents work in non-manual
(including professional) occupations (& = .01).
According to individual level model results, infants living in urban areas
are .41 times as likely to die as infants living in rural areas (& = .001).
And according to results from the community model, infants living in urban
areas are .62 times as likely to die as those living in rural areas (& =
.001).
Results from the community model reveal that in Zimbabwe infants living in
communities where adult female educational attainment is on the average primary
or secondary level are .38 and .08 times as likely to die as infants living
in communities where women are largely uneducated (& = .001). Infants living
in communities with Zimbabwe National Family Planning Council (ZNFPC) contraceptive
distribution programs were .76 times as likely to die as those without this
resource in their communities (& = .001); and those living in communities
with access to hospitals or health centers are .66 times as likely to die within
the first year of life as those that do not have access to modern health care
(& = .001).
There are some differences between the results of neonatal and postneonatal
mortality community models. ZNFPC and hospitals or health care centers in the
community do not significantly influence neonatal mortality. This social determinant
is less important for neonatal than biological determinants influencing fetal
health. But the average level of adult female education in the community is
important in both neonatal and postneonatal mortality models.
Results from the expanded model reveal that the number of people in the household,
professionally administered prenatal care, mother's secondary level education,
parents' regular employment, urban versus rural residence, the average level
of women's education in the community, ZNFPC contraceptive distribution programs
and hospitals and health centers in the community all significantly influence
infant surivival (Table 5). The magnitudes of effects are similar to the proximate
model of individual-level family and household variables. The main difference
is that education of individual mothers is a weaker determinant of postneonatal
mortality in the saturated model. Whereas maternal secondary level education
was a strong determinant of child survival in the model that excluded community
variables, once community variables are added in a saturated model, the average
level of female education in the community overshadows the effect of maternal
education of infant survival. Among neonates, the effect of individual mother's
education in the saturated model is not significant in either the individual-level
or saturated model, although the level of education of women in the community
is a significant determinant of neonatal survival (refer to table 4).
DISCUSSION
The individual-level model effectively sets the stage for results to follow
in the community and expanded models. Its performance is supported by chi-square
statistics in Table 3 which report that the community model is improved by
the addition of proximate and background variables. The proximate determinants
framework yields some useful predictors, but not all variables significantly
infuenced neonatal and postneonatal mortality in Zimbabwe.
Individual-level variables that are consistently significant predictors of
child survival include the number of members in the household, whether the
mother obtained prenatal care from a doctor or nurse and parents' employment.
The household contamination variables did not yield significant results as
was expected. Household contamination, as measured by a factor of porous floor
material, refrigerators, distance to household water, and modern toilet facilities,
is not an important indicator of newborns and infants' health status once other
socioeconomic variables are taken into account. And although the number of
household members was an important indicator in all four models in which its
effect was estimated, it operated less as an crowding and increased contagion
variable and more as a possible childcare variable.
As was expected, community-level variables gave added explanatory power to
the individual-level models. High proportions of women receiving a secondary
level of education in the community was the most important community-level
variable influencing infant mortality, with access to hospitals and the existence
of ZNFPC programs being important as significant determinants of postneonatal
survival.
Educational levels of women in the infant's community proved to be more significant
than the education of an infant's own mother in neonatal and postneonatal models
that include both variables. This indicates that equitable distribution of
this resource is beneficial to infant and newborn survival (Caldwell 1986).
The proportion of young girls enrolled in primary school has declined in most
African countries since the 1980s (according to the World Bank and UNICEF),
and will probably decline further in coming years. The positive relationship
between mass schooling and infant survival will translate into lower infant
mortality in Zimbabwe only insofar as the country's population and economic
growth levels allow for increased enrollment.
CONCLUSION
While a clear delineation of the mechanisms linking education and infant survival
still remains an unresolved issue, results of this analysis show that secondary-level
education of individual mothers and of women in the community promotes child
survival. Results reveal that when maternal education and average levels of
female education in the community are both observed, the state of female educational
advancement (or lack of education) in the community overshadows the effect
of the individual mother's educational attainment on infant survival. It is
not clear what the average level of female education in the community is actually
measuring. At least, it measures the existence of a phenomenon at the contextual
level which exerts an effect on infant mortality. We need to do further research
to help us more specifically define this context in social terms (and not just
as an area demarcated by a specific geopolitical boundary) and then examine
specifically what characteristics of this context correlate to proximate causes
of health behavior and outcomes. Is it that uneducated women are mimicking
the health promoting behavior of educated women who live close to them? Or
are women who live in close proximity sharing child care information? Another
possibility is that there is a greater demand for social amenities that benefit
child health and that all benefit regardless of educational level. Ethnographic
child care practice studies are needed to get a better grasp of these mechanisms
(Feyisetan 1988).
Figure 1:Causal ordering of factors affecting child health in Zimbabwe
Figure 2:Proximate and distal determinants of survival for infants born
between 1983 and 1988 in Zimbabwe
Table 1:Variable names and values for all models
Variable type and name |
Description |
Value |
DEPENDENT VARIABLES |
Neonatal mortality |
Mortality before completing first
month of life |
1= Died, 0= Survived |
Postneonatal mortality |
Time varying: Mortality after surviving
the first month and before completing the 1st year of life |
1= Child died in observation
month,
0= Child survived the observation month |
INDEPENDENT VARIABLES |
1. High-risk maternal fertilty
index |
Index accounting for the presence
of at least one high risk variable: Parity greater than 4, Insufficient
preceding birth interval: interval of <24 months for second or high
parity child; or Maternal age under 18 or over 36 |
1= At least one high risk
factor
0= None |
SOCIOECONOMIC STATUS |
|
|
Educational level of mother |
Recoded to 3 dummy variables |
|
2. Incomplete primary |
(omitted category) |
1= yes, 0= no |
3. Complete primary
4. Secondary |
-
- |
1= yes, 0= no
1= yes, 0= no |
Parent's usual occupation |
Father's occupation measured; mother's
occupation substituted if father unemployed or absent |
|
5. Missing |
Father's and mother's occupational
data missing |
1= yes, 0= no |
6. None |
Parent has no usual occupation |
1= yes, 0= no |
7. Agricultural |
Includes farmworkers and farmers |
1= yes, 0= no |
8. Manual labor |
Non-agricultural |
1= yes, 0= no |
9. Nonmanual labor |
(omitted category) |
|
HOUSEHOLD CONTAMINATION |
10. Household contamination factor |
Factor score coefficient comprised
of presence of refrigerator in household; toilet on premises; porous floor
material; and distance to household water supply (ordinal codes: 1 = on
premises, 2= less than 5 meters from premises, 3= 6 to 100 meters away,
4= 101 to 500 meters away,
6= 501 meters to 1 km away,
7= 1-3 kilometers away,
8= 3-5 kilometers away, to
9= greater than 5 kilometers away). |
mean= 0; s.d.=1 |
11. Number of household members |
Number at time of survey |
Continuous |
12. Prenatal care
13. Residence |
Care from a medical professional
Place of residence |
1= yes, 0= no
1= urban, 0= rural |
Table 1:Variable names and values for all models (continued)
Variable type and name |
Description |
Value |
COMMUNITY CONDITIONS
Average level of females'
education in the community
14. Incomplete primary
15. Complete primary
16. Secondary education
17. ZNFPC community based
distribution program
18. Hospital or health center
in the community |
Ordinal variable recoded
to three
dichotomous variables
(omitted category)
-
-
-
- |
1= incomplete primary
1= complete primary
1= secondary
1= yes, 0= no
1= yes, 0= no |
Source : ZDHS data (1989).
Table 2:Descriptive statistics on all variables in models
Variables |
N |
Mean |
Standard
division |
Neonatal mortality
Postneonatal mortality |
3393
3223 |
0.03
0.02 |
0.14
0.14 |
PROXIMATE VARIABLES |
1. High risk fertility
2. Mother under age 18
3. (Mother 18-36)
4. Mother over age 36
5. Preceding birth interval <24 months
6. Birth order >4
7. Number of household members
8. Professional prenatal care |
3390
3393
3393
3393
3388
3393
3393
3393 |
0.51
0.04
0.82
0.14
0.19
0.36
7.41
0.84 |
0.50
0.19
0.38
0.35
0.39
0.48
3.37
0.37 |
BACKGROUND CHARACTERISTICS |
9. (Mother did not complete
primary level education)
10.Mother completed a primary level of education
11.Mother completed secondary level or more education
12.Parents unemployed
13.Agricultural occupation
14.Manual occupation
15.(Non-manual occupation)
16.Missing information on occupation |
3393
3393
3393
3393
3393
3393
3393
3393 |
0.19
0.63
0.18
0.02
0.26
0.50
0.17
0.05 |
0.39
0.48
0.38
0.14
0.44
0.50
0.38
0.22 |
TYPE OF PLACE OF RESIDENCE |
17.Rural/urban residence |
3393 |
0.27 |
0.44 |
COMMUNITY CONDITIONS
(Average level of female's education in the community) |
18.(Incomplete primary)
19.Complete primary
20.Secondary education
21.ZNFPC distribution program in locale
22.Hospital or health center in the community |
3393
3393
3393
3387
3387 |
0.04
0.82
0.14
0.66
0.81 |
0.19
0.38
0.35
0.47
0.39 |
Source: ZDHS data (1989)
Note: Residual categories in parentheses.
Table 3:Mortality model selection
Model |
Overall model significance |
-2*Log Likelihood |
Degrees of freedom |
Chi-square (comparing constrained
and saturated models) |
Neonatal mortality: |
Base (proximate model)
Comunity model
Expanded (saturated model) |
82.86 ***
23.86 ***
95.41 *** |
1450.38
1517.92
1440.48 |
4
10
- |
9.90 *
77.44 ***
--- |
Postneonatal mortality: |
Base (proximate model)
Community model
Expanded (saturated model) |
544.41 ***
278.19 ***
669.70 *** |
12618.82
12883.52
12490.48 |
4
10
- |
128.34 ***
393.04 ***
--- |
Source: ZDHS data (1989)
Notes: *p<.05, **p<.01, and
***p<.001.
Negative log likelihoods are reported in tables 4 and 5.
Table 4:Determinants of neonatal mortality
Variables |
Individual model |
Community model |
Saturated model |
Odds Ratio |
z Value |
Odd
Ratio |
z
Value |
Odds
Ratio |
z
Value |
PROXIMATE VARIABLES |
1. High risk maternal fertility
2. Number of household members
3. Household contamination factor
4. Professional prenatal care |
1.42*
0.89***
0.89
0.35*** |
2.246
-4.430
-0.953
-6.591 |
|
|
1.42*
0.86***
0.88
0.35*** |
2.170
-4.227
-0.978
-6.596 |
BACKGROUND VARIABLES |
5. Mother completed secondary
or more education
6. Mother completed primary level education
7. (Mother did not complete primary level education)
8. Parents unemployed
9. Agricultural
10.Manual
11.Missing
12.(Non manual)
13.Rural/urban residence |
0.73
0.89
2.79*
1.11
1.00
1.73
0.53* |
-1.082
-0.659
2.388
0.391
0.035
1.499
-2.497 |
0.62*
|
-2.063 |
0.86
1.02
2.85
1.09
1.00
1.96
0.54* |
-0.496
0.086
2.426
0.317
0.018
1.330
-2.165 |
COMMUNITY VARIABLES |
14.Secondary level education
15.Completed primary level education
16.(Incomplete primary level education)
17.ZNFPC distribution program in locale
18.Community hospital or health center |
|
|
0.33**
0.49**
0.76
0.97 |
-2.715
-2.568
-1.697
-0.169 |
0.41*
0.58
0.73
1.04 |
-2.022
-1.737
-1.846
0.243 |
Source: ZDHS data (1989)
Note: Logistic regression analysis
of neonatal mortality; *p<.05, **p<.01, and *** p<.001. Omitted
category in parentheses.
Table 5:Determinants of postneonatal mortality
Variables |
Individual model |
Community model |
Saturated model |
Hazard Ratio |
z Value |
Hazard Ratio |
z Value |
Hazard Ratio |
z Value |
PROXIMATE VARIABLES |
1. High risk maternal fertility
2. Number of household members
3. Household contamination factor
4. Professional prenatal care |
1.10
0.86***
1.02
0.30*** |
1.268
-10.973
0.330
-16.483 |
|
|
1.08
0.87***
0.99
0.31*** |
0.991
-10.380
-0.208
-16.222 |
BACKGROUND VARIABLES |
5. Mother completed secondary
or more education
6. Mother completed primary level education
7. (Mother did not complete primary level education)
8. Parents unemployed
9. Agricultural
10.Manual
11.Missing
12.(Non manual)
13.Rural/urban residence |
0.49***
0.84*
1.74**
0.92
0.90
1.50*
0.41*** |
-4.517
-2.031
2.639
-0.720
-0.959
2.206
-6.553 |
0.62***
|
-4.044
|
0.71*
1.10
1.80**
0.85
0.85
1.33
0.54*** |
-2.096
0.989
2.426
0.317
0.018
1.330
-2.165 |
COMMUNITY VARIABLES |
14.Secondary level education
15.Completed primary level education
16.(Incomplete primary level education)
17.ZNFPC distribution program in locale
18.Community hospital or health center |
|
|
0.08***
0.38***
0.76***
0.66***
|
-9.284
-8.274
-3.432
-5.178
|
0.11***
0.45***
0.72***
0.74***
|
-2.022
-1.737
-1.846
0.243
|
Source: ZDHS data (1989)
Note: Cox regression analysis of
postneonatal mortality; *p<.05, **p<.01, and *** p<.001. Omitted
category in parentheses.
Table 6:Household composition by household size
|
All households [n=2228] |
Number of people
in household |
|
|
Six or fewer [n=1035] |
More than six [n=1193] |
Average number of children
under age 5 in household
Average number of children born to index mother 1983-88 |
1.81
1.52
|
1.36
1.47
|
2.20
1.57
|
Percentage of households
with children born between 1983 and 1988 not residing in household
Percentage of households with children under age 5
in addition to those born between 1983 and 1988 |
17.00
32.40 |
24.20
16.60 |
10.70
46.20 |
Average number of members
over age 5
Average number of eligible women in household |
5.49
2.00 |
3.23
1.00 |
7.45
2.00 |
Average household size |
7.30 |
4.59 |
9.66 |
Source: ZDHS data (1989)
REFERENCES
- Aaby, P. (1988). Overcrowding and intensive exposure: Major determinants
of variations in measles mortality in Africa. In Mortality and Society
in Sub-Saharan Africa, ed. Etienne van de Walle, Gilles Pison and Mpembele
Sala-Diakanda. Oxford: Clarendon Press.
- Agere, S.T. (1986). Progress and problems in the health care and delivery
system. In Zimbabwe: the Political Economy of Transition 1980-1986,
ed. Mandaza. Berkeley: Coderia Book Series.
- Allison, P.D. (1984). Event History Analysis: Regression for Longitudinal
Event Data. Sage University Paper series on Quantitative Applications
in the Social Sciences, series n° 07-046. Beverly Hills, CA: Sage Publications.
- Bailey, M. (1988). Individual and Environmental Influences on Infant
and Child Mortality in Rural Sierra Leone: a Multivariate Analysis. Journal
of Population Studies 12:155-185.
- Billy, J.O.G. & Moore, D.E. (1992). A Multilevel Analysis of Marital
and Nonmarital Fertility in the U.S. Social Forces 70:977-1011.
- Brewster, K.L., John O.G.Billy & Grady, W.R. (1990). A Multilevel
Analysis of Adolescent Contraceptive Use at First Nonmarital Intercourse.
Paper presented to PAA, Toronto.
- Brockerhoff, M. & De Rose, L.F. (1994). Parental education and
child survival: can the DHS tell us anything new? Health Transition Review 4:192-196.
- Caldwell, J.C. (1982). Toward a re-statement of demographic transition
theory. Theory of Fertility Decline. New York: Academic Press.
- Caldwell, J.C. (1986). Routes to low mortality in poor countries. Population
and Development Review 12:171-220.
- Caldwell, J.C. (1989). Mass Education as a Determinant of Mortality
Decline. In Selected Readings in the Cultural, Social and Behavioural
Determinants of Health, ed. Caldwell and Santow. Canberra, Australia:
Health Transition Center, The Australian National University.
- Campbell, G.R. (1989). The Political Epidemiology of Infant Mortality:
A Health Crisis Among the Montana American Indians. American Indian Culture
and Research Journal 113:105-148.
- Chen, L.C. (1983). Child Survival: Levels, Trends, and Determinants, Determinants
of Fertility in Developing Countries. New York: Academic Press.
- Cleland, J.G. and van Ginneken, J.K. (1988). Maternal Education and
Child Survival in Developing Countries: The Search for Pathways of Influence. Social
Science and Medicine 27:1357-68.
- Cleland, J.G. (1990). Maternal Education and Child Survival: Further
Evidence and Explanations. In What we know about health transition: the
cultural, social and behavioural determinants of health, ed. John Caldwell,
et al. Canberra, Australia: Health Transition Centre, Australian National
University.
- Computing Resource Center (1992) Stata Reference Manual. Santa
Monica, CA: Computing Resource Center.
- Curtis, S.L., Diamond, I. & McDonald, J.W. (1993). Birth Interval
and Family Effects on Postneonatal Mortality in Brazil. Demography 30:33-43.
- Das Gupta, M. (1990) Death Clustering, Mothers' Education and the Determinants
of Child Mortality in Punjab, India. Population Studies 44:489-505.
- Entwisle, C. and Hussein A.A. Sayed. (1989). Village as Contexts for
Contraceptive Behavior in Rural Egypt. American Sociological Review 54:1019-34.
- Federal Office of Statistics (1992) Nigeria Demographics and Health
Survey 1990. Lagos: Federal Office of Statistics.
- Feyisetan, B.J. (1988). Understanding the Relationship between Child
Care Practices and Child Mortality: The need for ethnographic approaches
in child mortality studies. Proceedings of African Population Conference
Dakar, Senegal 3:7.2.37-55.
- Gaisie, K. Cross, A. & Nsemukila, G. (1993). Zambia Demographic
and Health Survey 1992. Lusaka: University of Zambia.
- Hill, K. & Palloni, A. (1992). Demographic Response to Economic
Shocks: The Case of Latin America. IUSSP Peopling of the Americas Conference
Proceedings 3:411-438.
- Joshi, A.R. (1994). Maternal schooling and child health: preliminary
analysis of the intervening mechanisms in rural Nepal. Health Transition
Review 4:1-28.
- Kalbfleisch, J.D. & Prentice, R.L. (1980). The Statistical Analysis
of Failure Time Data. New York: John Wiley and Sons.
- Mandaza, I. (1986). Zimbabwe: The Political Economy of Transition
1980-1986. Berkeley: Coderia Book Series.
- Manga, P. (1988). The transformation of Zimbabwe's health care system:
a review of the white paper on health. Social Science Medicine 27:1131-1138.
- Morgan, S.P. & Teachman, J.D. (1988). Logistic regression: Description,
examples and comparisons. Journal of Marriage and the Family 50:929-936.
- Mosley, W.H. and Lincoln C. Chen. (1984). An analytic framework for
the study of child survival in developing countries. Population and Development
Review. Supplement to Volume 10.
- Mugabe, R.G. (1990). African countries must increase support for family
planning programmes. Populi 17:5-11.
- Norusis, M.J. (1990). SPSS Advanced User's Guide. Chicago: SPSS
Inc.
- Palloni, A. (1985). An epidemio-demographic analysis of factors in
the mortality decline of 'slow decline' developing factors. Paper presented
to IUSSP International Population Conference, Florence, 1985.
- Palloni, A. and Tienda, M. (1986). The effects of breastfeeding and
pace of childbearing on mortality at early ages. Demography 23:31-52.
- Pebley, A.R. and Millman, S. (1986). Birthspacing and Survival. International
Family Planning Perspectives 12:71-79.
- Phuapradit, W.; Pongthai, S.; Sudhutoravut, S.; Chaturachinda, K. & Benchakan,
V. ( ). Implementation of Risk Approcah in Maternal Health. Helen M. Wallace
and Kanti Giri (ed.), Health Care of Women and Children in Developing
Countries. Oakland, CA: Third Party Publishing Company.
- Population Reference Bureau (1994). World Population Data Sheet. Washington,
DC: Population reference Bureau.
- Preston, S.H. (1980). Causes and consequences of mortality declines
in less developed countries during the twentieth century. In Population
and Economic Change in Less Developed Countries, ed. Richard A. Easterlin.
Chicago: University of Chicago press.
- Qunta, C. (1987). Women in Southern Africa. New York: Allison
and Busby.
- Sanders, D. and Davies, R. (1988). The economy, the health sector and
child health in Zimbabwe since independence. Social Science and Medicine 27:723-731.
- Seers, D. (1969). The meaning of development. International Development
Review 11(4):2-6.
- Stoneman, C. & Cliffe, L. (1989). Zimbabwe: Politics, Economics
and Society. New York: Pinter Publishers.
- Tienda, M. (1984). Community characteristics, women's education, and
fertility in Peru. Studies in Family Planning 15(4):162-169.
- Tienda, M. (1991). Poor People and Poor Places: Deciphering Neighborhood
Effects on Poverty Outcomes. In Macro-Micro Linkages in Sociology,
ed. Joan Huber. Newbury Park, CA: Sage.
- Tsui, A.O. (1985). Community Effects on Contraceptive Use. In The
Collection and Analysis of Community Data, ed. John B. Casterline.
Vooburg, Netherlands: International Statistical Institute.
- United Nations. (1989). World Population Prospects 1988. New
York: United Nations.
- Wilkinson, M.I., Wamucii Njogu, and Noureddine Abderrahim. (1993). The
Availability of Family Planning and Maternal and Child Health Services.
Columbia, MD: Macro International.
- World Bank. (1992). Zimbabwe: Financing Health Services. Washington:
The International Bank for Reconstruction and Development.
- Zanamwe, L. (1988). The relationship between fertility and child mortality
in Zimbabwe. IUSSP African Population Conference Proceedings Dakar, Senegal 3(3):35-47.
- Zerai, A. (1992). Inequality in Child Health Status and Resources in
Zimbabwe by Province. Unpublished paper.
- Zerai, A. (1994). Preventive Health Strategies and Child Survival in
Zimbabwe. Ph.D. Dissertation. University of Chicago.
- Zimbabwe Central Statistical Office and Institute for Resource Development/Macro
Systems, Inc. 1989. Zimbabwe Demographic and Health Survey Country Report.
Harare: Central Statistical Office.
- Zimbabwe National Family Planning Council and Westinghouse Public Applied
Systems. (1985). Zimbabwe Reproductive Health Survey. Harare: Ministry
of Health.
- Zvobgo, R.J. (1986). Education and the Challenge of Independence. In Zimbabwe:
The Political Economy of Transition 1980-1986, ed. Mandaza. Berkeley:
Coderia Book Series.
NOTES
- According to the ZDHS, among all children 12-23 months, 79,8% living
in rural areas had health cards with immunization data recorded on them
and 71,3% of children living in urban areas had health cards. Among children
with health cards, 98% received the BCG vaccine in rural areas and 96%
received
the vaccine in urban areas; 96% of children living in rural areas received
their measles vaccine and in rural areas this number was 92% of children
aged 12-23 months (Zimbabwe Central Statistical Office 1989:87).
- There is supporting evidence for this in a recent article by Arun
R. Joshi "Maternal Schooling and Child Health: Preliminary Analysis
of the Intervening Mechanisms in Rural Nepal" (1994).
- Values of socioeconomic variables collected at the time of the
survey are used as a proxy for socioeconomic measures at the time the child
was
exposed to risk.
- Some of the limitations of the Demographic and Health Survey data
set pointed out by Brockerhoff and De Rose (1994) that need to be aknowledged
include the fact that there is no information on childbearing history.
This analysis assumes that all children live with their birth mothers.
Another
limitation is that there is insufficient observation of morbidity and
health interventions --educated mothers are more likely to report health
interventions.
The result in this analysis is that the effect of prenatal care is possibly
underestimated. Finally, there are at least two instances in which the
effect of educational levels of infant survival might be underestimated.
Because
the birth history data is collected retrospectively and mothers with
greater levels of education are probably less likely to omit births that
end in death
than are mothers with less education, the effect of education on infant
survival may be underestimated. And because maternal education is related
to maternal
mortality, and birth history information is not collected on mothers
who died in the locale, the effect of education on infant survival might
again
be underestimated.
- The combined index more closely aligns this analysis with policy
prescriptives that identify high risk pregnancies than with theoretical
concerns with the distinctive impact of fertility limitation, birth spacing,
and maternal
age on infant survival.
- Often these clusters were located in independent communities.
- Since information collected on community health facilities within
the Zimbabwe Demographic and Health Survey consisted of data from only
the nearest health
facility, these data are not nationally representative.
- These were estimated with the State corporation statistical analysis
software.
- My assumptions for analysis of multilevel models include the following:
that contextual variables are not picking up effects of correlated,
excluded individual-level variables; that mothers have not selected their
contexts
by perceived levels of education, existence of contraceptive distribution
programs, or access to health facilities; and that mothers' community
of residence at the time of the survey are where the children were exposed
to
the risk of mortality in earlier years.
- In this section postneonatal mortality, the probability of death
from ages 1 month through the twelfth month of life, is estimated.
- These children were less than 1 month old because they were born
during the month the survey was being taken.
Copyright 1996 - Union for African Population Studies.
|